Keywords: Adaptive control design
Abstract: Learning and adaptation are fundamental to the modeling and control of uncertain dynamical systems, and they are also cornerstones of intelligent systems. While statistical learning has underpinned much of the progress in machine learning, its theoretical foundations often rest on idealized assumptions about data, such as independent and identically distributed (i.i.d.) samples. These assumptions are typically not valid for dynamical systems, especially those under feedback control, where the input-output data exhibit strong temporal correlations and inherent nonstationarity. This motivates the development of a more general theoretical framework that can account for the complex data characteristics intrinsic to dynamical systems. This talk addresses these challenges by presenting some of our recent advances in learning and adaptation for stochastic dynamical systems. We introduce general and verifiable data conditions that can guarantee global stability and estimation performance for adaptive learning and filtering algorithms with either diminishing or non-diminishing adaptation gains. Topics covered include gradient-based and Newton-type adaptive methods, distributed learning and filtering in networked systems, integrated offline-online learning frameworks. We also consider practical challenges such as saturated or limited-value measurements, finite data size, input-state constraints, and time-varying system parameters. The lecture concludes with a discussion of emerging research at the nexus of control and AI.

Keywords: Electrical protection and fault diagnosis
Abstract: Large Language Models (LLMs) have rapidly emerged as tools of interest across engineering disciplines, and Process Systems Engineering (PSE) is no exception. This survey provides a systematic review of LLM applications in PSE, organizing the literature into six categories: (1) process design and engineering, (2) molecular design and synthesis, (3) process modeling and simulation, (4) optimization and scheduling, (5) process control, and (6) fault detection and diagnosis. For each category, we summarize the state of the art, identify common methodological approaches, and critically assess demonstrated capabilities versus aspirational claims. We find that LLMs show genuine promise for tasks involving natural language, including querying documentation, synthesizing unstructured knowledge, and enabling flexible human-machine interaction. However, applications requiring real-time performance or formal guarantees remain challenging. We conclude by identifying open problems and productive research directions for the PSE community.

Keywords: Advanced process control
Abstract: Fault diagnosis is a critical link in ensuring the safe operation of industrial systems. Traditional time-series data diagnosis models typically output abstract results, such as anomaly scores or fault categories, but they cannot answer key questions like “why the fault occurred” or “how to perform maintenance.” Although large language models (LLMs) show great potential for fault diagnosis, they face the challenge of a semantic gap when processing time-series industrial signals; that is, continuous temporal data are difficult to encode into discrete tokens that language models can effectively process. Differing from the traditional “signal-to-category” paradigm in fault diagnosis, we propose a novel explainable fault diagnosis framework, namely the “Signal-to-Semantics” (S2S) fault diagnosis framework. Our research replaces the original paradigm of mapping abstract time-series data to abstract diagnostic results, and instead outputs reasoning processes and diagnostic texts that are comprehensible and verifiable by human experts, establishing a new generation of intelligent diagnosis frameworks for industrial equipment.

Keywords: Advanced process control

Keywords: Advanced process control
Abstract: Fault recovery in process plants still relies heavily on plant operators, especially when faults fall outside predefined supervisory logic. Operators interpret alarms, procedures, P&IDs, interlocks, and process trends, then decide how to move the plant to a safe operating mode without triggering a shutdown. This paper examines how Large Language Model (LLM) agents can support such recovery decisions. The proposed framework treats the LLM as a constrained supervisory planner. It uses plant-specific knowledge to propose recovery actions, and every proposal is checked by an external validator, either symbolic or simulation-based, before actuation. The paper develops three design dimensions for applying the framework: the recovery patterns for which LLM agents are useful, the validation strategies that separate admissible from inadmissible proposals, and the deployment constraints imposed by latency, knowledge engineering, safety integration, and model lifecycle management. To make the framework directly usable, two openly available executable Python environments are provided. Both re-implement established case studies, a modular mixing module and a continuous stirred-tank reactor, extended with configurable faults and defined interfaces for custom recovery and validation methods.

Keywords: LLMs for modeling and control, Development of assistant systems for manufacturing systems, LLM-enhanced human-in-the-loop
Abstract: Large model-driven robotic embodied intelligence systems have achieved breakthrough progress in various tasks, thanks to the powerful cross-modal information processing and semantic understanding capabilities of large models. However, in the more traditional process industry and discrete manufacturing systems, research and applications of large model-based technologies are just in their infancy, and industrial embodied intelligence has thus become a key development direction for the future. This paper first attempts to provide a generalized definition and key components of industrial embodied intelligence. On this basis, a generalized architecture of industrial embodied intelligence is proposed, and the existing research and technical progress are elaborated in detail from three aspects: general multi-modal large models and industrial large models, large model-driven industrial data perception and knowledge extraction, and large model-driven task decision-making and optimal control. Finally, the key technical challenges in realizing industrial embodied intelligence are presented, providing guidance and reference for theoretical research, technological breakthroughs, and practical applications in this field.

Keywords: Linear system identification, Active learning and experiment design
Abstract: We address optimal trajectory design for mobile sensors in distributed-parameter systems. The problem is formulated as an optimal control program that minimizes a Fisher-information–based criterion over sensor initial positions and controls, while enforcing motion, domain, and separation constraints. Non-convex constraints are handled via an exact mixed-integer reformulation, and gradients are computed from a linearized sensitivity–adjoint scheme. The proposed framework is illustrated using a two-dimensional advection–diffusion system characterized by a parametric initial condition and diffusivity field.

Keywords: Linear system identification, Estimation and filtering
Abstract: This paper studies the sparse identification problem of stochastic dynamical systems with infinite parameters. We first use a least squares (LS) algorithm to obtain the parameter estimates, where the dimension of parameters gradually increases with time. Based on the estimate, we propose a loss function with L_1 regularization term, by minimizing which we obtain an algorithm to estimate the unknown sparse infinite parameters. We establish the almost sure convergence result of the sparse algorithm, and further give the finite-time convergence of the set of zero elements. Our theoretical results are obtained without requiring the regression vectors to be independent and identically distributed (i.i.d.) or to satisfy the persistent excitation (PE) condition. A simulation example is given to verify the effectiveness of the theoretical results.

Keywords: Linear system identification, Estimation and filtering, Kalman filtering
Abstract: Prediction error and maximum likelihood methods are powerful tools for identifying linear dynamical systems and, in particular, enable the joint estimation of model parameters and the Kalman filter used for state estimation. A key limitation, however, is that these methods require solving a generally non-convex optimization problem to global optimality. This paper analyzes the statistical behavior of local minimizers in the special case where only the Kalman gain is estimated. We prove that these local solutions are statistically consistent estimates of the true Kalman gain. This follows from asymptotic unimodality: as the dataset grows, the objective function converges to a limit with a unique local (and therefore global) minimizer. We further provide guidelines for designing the optimization problem for Kalman filter tuning and discuss extensions to the joint estimation of additional linear parameters and noise covariances. Finally, the theoretical results are illustrated using three examples of increasing complexity. The main practical takeaway of this paper is that difficulties caused by local minimizers in system identification are, at least, not attributable to the tuning of the Kalman gain.

Keywords: Linear system identification, Gaussian process, Probabilistic and Bayesian methods for system identification
Abstract: This paper addresses the challenge of systematically determining the optimal parameters for Continuous-Time Predictor-Based Subspace Identification (CT-PBSID) to maximize the accuracy of the identified model while significantly reducing the computational time with respect to grid search. The median of the Root Mean Square Error (RMSE) of the outputs in cross-validation is used as the objective in a Bayesian optimization framework, which efficiently converges to its minimum, thereby yielding the most accurate identified model. Simulations in a test example demonstrate the effectiveness and robustness of the proposed algorithm. In addition, the advantage in terms of computational time with respect to grid search is shown, suggesting that the method is effectively transferable to future industrial applications.

Keywords: Linear system identification, Kalman filtering, Time/parameter varying system identification
Abstract: Many dynamical systems operate under unknown periodic disturbances, which degrade the performance of fault diagnosis and control algorithms if left untreated. In this paper, we propose a simple recursive subspace method for estimation and rejection of time-varying harmonic components in outputs of a system generated by a stochastic linear time-invariant plant and a deterministic linear time-varying harmonic subsystem. The method is validated on a toy example of a mechanical system, illustrating its effectiveness.

Keywords: Linear system identification, Learning methods for control
Abstract: Given a reaction-diffusion equation with unknown right-hand side, we consider the nonlinear inverse problem of estimating the associated leading eigenvalues and initial condition Fourier coefficients from a finite number of non-local noisy measurements. We define a reconstruction (i.e., estimation) criterion and, for small enough noise, we prove existence and uniqueness of the desired estimates. We derive closed-form expressions for the first-order condition numbers and bounds for their asymptotic behavior in a regime when the number of measured samples is fixed and the inter-sampling interval length is arbitrarily large. When computing the sought estimates numerically, our simulations show that the exponential fitting algorithm ESPRIT is first-order optimal, since its first-order condition numbers have the same asymptotic behavior as the analytic condition numbers in the considered regime.

Keywords: Linear system identification, Nonlinear adaptive control, Learning methods for control
Abstract: This paper presents a new composite adaptive trajectory tracking controller for fully actuated torque-driven robotic manipulators. The proposed approach integrates two powerful parameter identification techniques—namely, the dynamic regressor extension and mixing (DREM) methodology and the learning-based methodology—and exploits their combined benefits. As a first stage, the system parametrization is obtained using the power balance equation parametrization (PBEP), which yields a simpler and less computationally demanding regressor, thereby reducing the total computational cost of the proposed algorithm compared with the classical parametrization for mechanical systems. Compared with classical adaptive controllers, the proposed methodology guarantees exponential convergence to zero of the position, velocity, and parameter estimation errors—that is, the difference between the true and estimated parameters—without requiring verification of the persistent excitation condition in the regressor. Moreover, compared with the original learning-based controllers, the proposal removes the stringent requirement of verifying the invertibility of the regressor matrix, which further reduces computational cost and enhances its feasibility for implementation. Finally, the performance of the proposed controller is validated through experiments on a two-degree-of-freedom robotic manipulator, which support the claims presented.

Keywords: Linear system identification, Nonlinear system identification, Adaptive observer design
Abstract: Recently, an exponentially convergent estimator was proposed in Liu et al. (2024) for frequency estimation of unknown continuous-time multi-tone sinusoidal signals. However, since this estimator employs a standard gradient algorithm in its parameter adaptation law, the fixed adaptation gain limits its ability to handle the measured signal with varying amplitudes. To overcome this limitation, we propose a new parameter adaptation law based on a normalized gradient algorithm. The resulting estimator features a time-varying adaptation gain that dynamically adjusts according to the measured signal amplitudes. Comparative simulations demonstrate the superior performance of the proposed estimator over the existing one, achieving reduced fluctuation and faster convergence.

Keywords: Linear system identification, Statistical analysis, Machine and deep learning for system identification
Abstract: Quantifying the estimation error of a model estimate is a key problem in system identification for a given data record with finite sample size N. There are mainly two routes to address this problem: the large sample asymptotic theory based method and the non-asymptotic theory based method. However, the existing results are not very effective for quantifying the estimation error when N is not large, the model order n is not small, and n/N is not too small (e.g., n/N=0.5). In this paper, we revisit the asymptotic theory of the FIR model estimation with white noise input and measurement noise by the least squares (LS) method but under a more realistic asymptotic setup: let both N, nrainfty with n/Nragammain(0,1). We first derive the asymptotic variance and then establish the Central Limit Theorem for the squared estimation error of the LS method. Based on the obtained theoretical results, we provide two types of quantification for the estimation error of the LS method. Monte Carlo simulation demonstrates that the provided two types of quantification are more accurate than the classic ones, especially when gamma is not too small.

Keywords: Linear system identification, Time series modeling
Abstract: This paper presents an errors-in-variables identification method for autoregressive (AR) models in the presence of additive noise. The approach exploits the properties of the dynamic Frisch scheme and employs a loss function based on the discrete spectral distance between the power spectral density (PSD) of the noisy measurements and that of the estimated noisy AR model. The performance of the proposed identification algorithm is evaluated through Monte Carlo simulations and compared with existing methods, focusing on robustness to observation noise and spectral estimation accuracy. Simulation results demonstrate that the method is effective for both narrowband and broadband processes and achieves superior spectral estimation performance for narrowband signals. This is a valuable feature for applications such as fault diagnosis, biomedical signal processing and speech analysis.

Keywords: Machine and deep learning for system identification, Nonlinear system identification
Abstract: Due the complexity of modern power systems, modeling based on first-principles becomes increasingly difficult. As an alternative, dynamical models for simulation and control design can be obtained by black-box identification techniques. One such technique for the identification of continuous-time systems are neural ordinary differential equations. For training and inference, they require initial values of system states, such as phase angles and frequencies. While frequencies can typically be measured, phase angle measurements are usually not available. To tackle this problem, we propose a novel structure based on augmented neural ordinary differential equations, learning latent phase angle representations on historic observations with temporal convolutional networks. Our approach combines state-of-the art deep learning techniques, avoiding the necessity of phase angle information for the system identification. Results show, that our approach clearly outperforms simpler augmentation techniques.

Keywords: Nonlinear system identification, Active learning and experiment design
Abstract: The quality of an estimated nonlinear model highly depends on the data quality that was used for the system identification. By using a Gaussian Process-based optimal input design approach, a so-called space-filling dataset can be generated in the feature space of the system model. The design method is applicable for a broad type of signals and models and also incorporates information measures through optimality criteria into the signal design. However, the resulting input design can be costly to apply to the real system. The goal of this paper is to propose a space-filling input design that can minimize the experimentation cost in terms of a user defined measure, while still guaranteeing a prescribed level of space-fillingness. Through a Monte Carlo simulation study we demonstrate that the proposed method can appropriately shape the excitation signal to significantly reduce the experimental cost while the identified model performance remains adequate.

Keywords: Nonlinear system identification, Adaptive observer design
Abstract: Many technological process models contain nonlinear functions with parameters that cannot be isolated or appear in an affine form after representation. This paper proposes a method for adaptively estimating systems with these non-separable nonlinear parameterizations by transforming the problem into an observation of an augmented state of a linearly parameterized nonlinear descriptor system. We propose a new adaptive observer design within this framework. The effectiveness of the developed method is shown through academic examples.

Keywords: Nonlinear system identification, Data-driven control theory, Control under communication constraints
Abstract: This paper develops a data-driven method for designing state-feedback controllers for nonlinear discrete-time dynamic systems in the presence of time-varying feedback delays. We first develop a Koopman autoencoder that learns linear latent representations of the nonlinear model directly from state measurements. Thereafter, we design a state-feedback controller in the Koopman-lifted space that is robust to the worst-case feedback delay. The two designs are illustrated using a power system model with wind power integration that contributes towards the system nonlinearity. The simulation results verify that the delay-robust Koopman-based controller can improve the control performance over a wide range of delays, thereby outperforming conventional delay-agnostic data-driven control approaches, which are shown to fail under realistic delay conditions.

Keywords: Nonlinear system identification, Data-driven control theory, Realization theory
Abstract: Understanding when linear immersions of nonlinear dynamical systems exist is important since such immersions allow us to leverage the rich tools of linear system theory to analyze nonlinear dynamics. Recently, Liu et al. 2023 showed that continuous-time dynamical systems that admit countably many but more than one omega-limit sets cannot be immersed into finite dimensional linear systems with a one-to-one and continuous mapping. In this paper, we extend these results to discrete-time dynamics and show that similar obstructions exist also in discrete time. We further consider a generalization involving alpha-limit sets. Several examples are provided to demonstrate the results.

Keywords: Nonlinear system identification, Filtering and smoothing, Physics informed and grey box model identification
Abstract: A novel method is presented for the direct continuous-time model identification of nonlinear systems subject to output measurement noise. The approach combines delayed state-variable filters with an instrumental-variable (IV) estimation scheme to remove the dominant stochastic bias present in least-squares-based formulations. The analysis further reveals residual modeling errors arising from output interpolation and imperfect commutation between delayed filtering and nonlinear mappings. Although these deterministic contributions cannot be removed by IV estimation, the imperfect commutation error can be mitigated through appropriate dSVF design and cutoff-frequency tuning. Monte Carlo simulations demonstrate robustness to high output measurement noise levels and to variations in the filter cutoff frequency.

Keywords: Nonlinear system identification, Linear system identification, Time/parameter varying system identification
Abstract: Polynomial systems arise naturally in control theory and related areas, yet their nonlinear structure often prevents direct analysis. This paper investigates the notion of polynomial constructibility, where the solution of a nonlinear system can be recovered as a polynomial function of the solution of a linear system. Our main results provide sufficient conditions for polynomial constructibility, formulated in terms of skeleton graphs and depth decompositions. In particular, we show that a large class of super-linearizable systems are polynomially constructible.

Keywords: Nonlinear system identification, Machine and deep learning for system identification
Abstract: Approximating nonlinear dynamics with sharp transitions remains challenging in many engineering and control modeling problems. We propose COPNet, a compositional orthogonal polynomial network that uses the structure of orthogonal polynomials without explicitly constructing high degree polynomial expansions. COPNet is built from a learned second order recurrence inspired by classical orthogonal polynomial relations. Through multiplicative feature coupling and a two back recursive connection, COPNet forms a fixed width architecture that remains compact while developing expressive features across depth. In physics informed learning, COPNet can be paired with different polynomial families according to the structure of the target problem: Chebyshev based models for bounded spatiotemporal problems with sharp transitions, Hermite based models for localized Gaussian like behavior on large domains, and Legendre based models for bounded elliptic problems on uniform domains. We evaluate COPNet on the Burgers, Allen--Cahn, harmonic oscillator heat, and two dimensional Poisson equations. Across these benchmarks, COPNet achieves accurate solutions with compact architectures. On the main comparison problems, COPNet attains lower relative L^2 errors than reported PINN baselines while using fewer interior collocation points and narrower networks. These results support COPNet as an effective recurrence based architecture for efficient physics informed nonlinear modeling.

Keywords: Nonlinear system identification, Machine and deep learning for system identification
Abstract: The availability of equidistant sampled data is a starting assumption for most identification approaches. However, in some scenarios, only non-equidistant sampled data is available, e.g. due to sensor imperfections or event-triggered sampling. This paper introduces Irregular SUBNET, tailored to identify continuous-time nonlinear state-space models starting from data sampled at irregular intervals. This approach introduces two main changes compared to the previously introduced continuous-time SUBNET identification approach: a sample interval aware encoder function for the estimation of the initial state, and the use of a variable length ODE integration to propagate the state information forward in time. The effectiveness of the proposed approach is validated on a simulation example and on the EMPS benchmark.

Keywords: Nonlinear system identification, Machine and deep learning for system identification, Learning methods for control
Abstract: We present an end-to-end neural network approach to estimate the largest Lyapunov exponent (LLE) directly from time series. We address the research gap regarding system-agnostic generalization by training a Long Short-Term Memory (LSTM) on two structurally different maps: the logistic and Hénon maps. Results show that a jointly-trained network matches the accuracy of system-specific models (R² ≈ 0.984), suggesting the network internalizes the underlying estimation algorithm rather than memorizing system-specific features.

Keywords: Nonlinear system identification, Physics informed and grey box model identification
Abstract: This work addresses the online parameter identification of planar flexible two-link manipulators modeled by an assumed modes formulation. We exploit the linear-in-parameters structure of this model to perform online estimation of physically meaningful parameters, such as the motor torque constants, viscous and Coulomb friction coefficients, and structural damping parameters, using a recursive least squares scheme with an adaptive forgetting factor. This is in contrast to previous approaches that either require offline identification of friction parameters or neglect unknown parameters such as structural damping and motor torque constants. To handle the nonlinear dependence of the flexible modes on the second joint angle, three regressor constructions are proposed and compared: a model linearized around a nominal second joint angle, the full nonlinear model, and a lookup table approximation based on offline solutions of a configuration-dependent eigenvalue problem. Simulation results show that the linearized scheme does not provide reliable convergence, whereas both nonlinear variants substantially reduce the parameter errors. Importantly, the lookup table-based scheme closely matches the accuracy of the full nonlinear estimator while requiring significantly less computation time per step, which makes it a promising candidate for real-time parameter identification.

Keywords: Nonlinear system identification, Probabilistic and Bayesian methods for system identification
Abstract: In this paper, a Maximum Likelihood estimation algorithm for a Wiener system with a piecewise linear approximation to model the output non-linearity is developed. We propose a methodology to construct the probability density function associated with the piecewise linear function by using a Gaussian mixture indicator approximation. An Expectation-Maximization algorithm is proposed to estimate both the linear system model and the piecewise linear function parameters, obtaining closed-form expressions for the parameter estimators. The benefits of our proposal are illustrated via numerical simulations.

Keywords: Time series modeling, Linear system identification, Estimation and filtering
Abstract: In both control and signal processing, there has been a decades long interest in constructing vector auto-regression (VAR) estimators with guaranteed stability. We revisit some classic work from the late 1970s and find a fatal flaw - namely that initiating estimators in forward and backward recursions are not guaranteed to be stable. Then, focussing on the VAR(1) case, we develop a new approach which yields a closed from estimator with the properties claimed for the classic estimator.

Keywords: Time series modeling, Linear system identification, Realization theory
Abstract: We introduce a new spectrum approximation framework for multivariate stationary processes based on a transportation-entropy formulation. Classic rational covariance extension methods allow to impose covariance constraints on the spectrum and, in the scalar case, additional constraints of cepstral coefficients to control the spectral zeros. However, extending the cepstral coefficients to the multivariate setting has remained challenging, since the standard matrix logarithm does not yield a tractable dual problem. Building on recent developments in optimal transport for Gaussian processes, we define a new class of cepstral-type quantities, called W-cepstral coefficients, derived from a spectral factor and compatible with a transportation-entropy functional. This leads to a well-posed convex optimization problem whose dual formulation can be explicitly characterized. We show that the proposed approach successfully identifies the spectrum of a multivariate stationary process from a finite set of covariance lags and W-cepstral coefficients, and we validate the theory through numerical simulations.

Keywords: Adaptive control of multi-agent systems, Model reference adaptive control, Multi-agent systems
Abstract: This paper presents the application of a Distributed Model Reference Adaptive Control (DMRAC) strategy for robust multi-agent synchronization of a network of drones. The proposed approach enables the development of controllers that can accommodate differences in real-life model parameters among agents, thereby enhancing overall network performance. We compare the performance of adaptive control laws with that of classical PID controllers for the reference tracking task. Each follower drone has a model reference adaptive controller that continuously updates its parameters based on real-time feedback and reference model information. This adaptability ensures adequate performance, which, compared to conventional non-adaptive techniques, can reduce the amount of energy required and consequently increase the operating duration of the drones. The experimental results, particularly in vertical velocity control, underscore the effectiveness of the proposed approach in achieving synchronized behavior.

Keywords: Adaptive observer design, Nonlinear adaptive control
Abstract: Trigonometric-polynomial disturbances are among the most commonly encountered disturbances in practice, as they can approximate nearly any periodic signal with an unknown period. The most effective method for asymptotically rejecting this class of disturbances is through a dynamic compensator known as an internal model, which transforms the disturbance rejection problem into a stabilization problem for an augmented system. However, existing internal model design approaches rely heavily on the properties of the solution to the regulator equations. An effective internal model can only be constructed when this solution exhibits specific characteristics, such as being polynomial in the exogenous signal. For complex nonlinear systems, especially nonautonomous ones, solving the disturbance rejection problem using traditional methods remains challenging. In this paper, we propose a novel framework for disturbance rejection in a class of nonautonomous nonlinear systems affected by matched trigonometric-polynomial disturbances. The core of our approach is the design of a canonical internal model that directly converts the disturbance rejection problem into an adaptive stabilization problem for an augmented system. Unlike conventional methods, this internal model is synthesized directly from the given nonlinear plant and the knowledge of the exosystem, without relying on the solution of the regulator equations. This makes the approach applicable to a significantly broader class of nonautonomous nonlinear systems. Furthermore, we develop an adaptive disturbance observer comprising the canonical nonlinear internal model, a Luenberger-type state observer, and a parameter adaptation law. This observer ensures global asymptotic convergence of the disturbance estimate to the true disturbance without requiring persistent excitation (PE). Under the PE condition, both the disturbance estimation error and the parameter estimation error converge exponentially. By incorporating the disturbance estimate as a feedforward compensation signal, we establish sufficient conditions for achieving global trajectory tracking and asymptotic disturbance rejection.

Keywords: Autonomous marine systems and vehicles, Marine robotics, Marine system guidance, navigation and control
Abstract: Autonomous navigation for Unmanned Surface Vehicles (USVs) in cluttered harbors presents a dual challenge: the rigorous constraints of hydrodynamic inertia and the complexity of perceiving diverse surface and underwater hazards amidst wave clutter. While recent neuro-symbolic planners excel on ground robots, they often fail in maritime settings due to kinematic mismatches and sensitivity to environmental noise. This paper presents a Predictive and Inertia-Aware Neuro-Symbolic framework designed to bridge this gap. First, to address the heterogeneity of maritime perception, we replace raw sensor inputs with a high-level representation based on Potential Collision Areas (PCAs). By aggregating object detections from multiple modalities (e.g., LiDAR, Sonar) and predicting their future trajectories, we construct dynamic 2D bounding boxes in the navigation plane that encapsulate future collision risks. This object-level abstraction unifies diverse sensor data and effectively filters out transient wave clutter that confuses standard planners. Second, we propose a physics-informed domain adaptation strategy where the neural policy is trained via constrained imitation learning. By subjecting the expert demonstrator to strict acceleration limits, the network implicitly internalizes hydrodynamic inertia, learning to initiate avoidance maneuvers well in advance. Validation in high-fidelity simulations demonstrates that our method successfully compensates for drift, handles multi-modal obstacle data, and proactively avoids dynamic threats, achieving safe navigation where standard kinematic baselines fail.

Keywords: Autonomous marine systems and vehicles, Marine system guidance, navigation and control
Abstract: In large-scale marine monitoring and inspection, coordinated fleets of unmanned surface vehicles (USVs) must perform efficient task allocation and path planning in dynamic environments characterized by ocean currents, obstacles, and potential system faults. To address these challenges, we propose a hierarchical framework (DSOM-GA) that combines a dual-layer self-organizing map for task partitioning with genetic algorithm-based task sequencing. The framework incorporates a coarse-to-fine, flow-aware path-cost evaluation scheme and a bounded reallocation mechanism for fault recovery. Extensive simulation-based mission-level evaluations show that the proposed framework reduces the normalized mission cost and improves the robustness of post-fault task reallocation relative to representative heuristic baselines.

Keywords: Autonomous marine systems and vehicles, Marine system guidance, navigation and control, Dependability in marine systems
Abstract: This paper presents a practical adaptive PD depth controller for torpedo-style autonomous underwater vehicles (AUVs), integrating a modified model reference adaptive control (MRAC) update law into a conventional PD architecture. The controller is evaluated on a REMUS 100 6-DOF nonlinear AUV model across multiple payloads and surge speeds. Results demonstrate that the adaptive scheme maintains accurate depth tracking with minimal overshoot and consistent rise times, while reducing control effort compared to a baseline PID controller. The adaptive gains evolve smoothly with payload and setpoint variations, providing robustness to dynamic changes. The approach offers a simple, modular method to enhance AUV performance without the complexity of fully adaptive controllers.

Keywords: Autonomous marine systems and vehicles, Marine system guidance, navigation and control, Modelling, identification and control in marine systems
Abstract: This paper proposes a Lyapunov-constrained Soft Actor-Critic (LC-SAC) controller for dynamic positioning (DP) of unmanned surface vehicles (USVs). Due to the persisting disturbances from waves, wind, and ocean currents, as well as the resulting complex hydrodynamics, it is difficult to obtain an accurate USV model. To address this issue, a lightweight random Fourier feature (RFF) learning method is used to learn a unified model of USV dynamics and environmental disturbances. Considering the stringent stability and steady-state accuracy specifications in DP control, a Lyapunov-based stability constraint is integrated into the SAC framework via a primal-dual optimization scheme, in which a Lagrange multiplier enforces the stability condition during policy learning. Simulation results show that the proposed LC-SAC achieves faster convergence, smaller steady-state error, and stronger disturbance rejection than adaptive PID, Krasovskii-constrained RL, and standard SAC controllers.

Keywords: Autonomous marine systems and vehicles, Marine system guidance, navigation and control, Modelling, identification and control in marine systems
Abstract: Actuator amplitude and rate saturation (A&RSat), together with the associated windup problem, have long been recognised as challenges in control systems. Anti-windup (AW) schemes have been developed over the past decades and can generally be categorised into two main groups: classical and modern anti-windup (CAW and MAW) approaches. Classical methods have provided simple and effective solutions, primarily addressing amplitude saturation. In contrast, modern approaches offer powerful and theoretically sound frameworks capable of handling both amplitude and rate saturation. However, the derivation of MAW schemes often imposes restrictive conditions and can be complex to apply in practical engineering problems. Nevertheless, the literature has paid limited attention, if not largely ignored, to the potential of CAW schemes that can operate in the presence of both A&RSat. This paper revisits this issue and proposes modifications to two well-known controllers: PID and LQI. The results obtained, benchmarked on the REMUS AUV yaw control problem and compared with constrained MPC, indicate that these classical techniques can still provide simple yet effective solutions with comparable performance, at least for SISO systems. These findings may stimulate further research into solutions that achieve comparable performance with only one (or a limited number of) additional tuning parameters and enable straightforward implementation.

Keywords: Autonomous marine systems and vehicles, Multi-vehicle systems, Simulation and digital-twin in marine systems
Abstract: Sustainable deep-sea mining requires effective monitoring of sediment plumes to safeguard vulnerable marine ecosystems. This paper presents a collaborative, technique for tracking sediment plumes without the explicit calculation of gradients for a swarm of realistic six-degree-of-freedom nonlinear unmanned underwater vehicles (UUVs). This vehicle model takes into account all hydrodynamic effects including thrust allocation. The suggested hybrid control framework combines the proposed control architecture with realistic UUV dynamics. We use Lyapunov-based stability analysis for the parameter selection and for theoretical stability. Numerical simulations validate the method, showing that it can coordinate swarms well and accurately localize the plumes. These results show that deploying cooperative UUV swarms for autonomous deep sea plume monitoring is a feasible option that will make marine operations safer and more environmentally friendly.

Keywords: Cyber security networked control, Control over networks
Abstract: This paper introduces a cloud-assisted closed-loop anaesthesia control system that preserves patient data privacy through homomorphic encryption. Sensor measurements are encrypted using the homomorphic CKKS scheme, and controller computations are performed directly on ciphertexts, maintaining confidentiality. Since homomorphic encryption cannot perform non-linear operations such as saturations and clampings, the paper analyses how such limitations affect the performance of a controller with anti-windup mechanism. The study is carried out by means of numerical simulations concerning the practical scenario of a patient whose level of anaesthesia must be regulated.

Keywords: Cyber security networked control, Resilient networked control systems, Fault detection and diagnosis
Abstract: This paper addresses the problem of designing a resilient vector current control strategy for grid-connected Inverter-Based Resources (IBRs) under stealthy cyberattacks. We investigate scenarios where sophisticated adversaries aim to compromise the safety of the grid-connected IBR by injecting false data into control input channels, specifically designed to bypass existing bad data detection mechanisms in the system. The proposed approach introduces a safety-preserving control framework that can maintain the safe operation of the grid-connected IBR even when subjected to such stealthy attacks. The proposed controller is a solution to a convex optimization problem. Simulation results demonstrate the effectiveness of the proposed approach in ensuring continued stable operation under stealthy attacks, thereby enhancing the cybersecurity posture of critical IBR interfaces in modernized power systems.

Keywords: Distributed control and estimation
Abstract: Situational awareness for connected and automated vehicles describes the ability to perceive and predict the behavior of other road-users in the near surroundings. However, pedestrians can become occluded by vehicles or infrastructure, creating significant safety risks due to limited visibility. Vehicle-to-everything communication enables the sharing of perception data between connected road-users, allowing for a more comprehensive awareness. The main challenge is how to fuse perception data when measurements are inconsistent with the true locations of pedestrians. Inconsistent measurements can occur due to sensor noise, false positives, or unmodeled disturbances. This paper employs set-based estimation with constrained zonotopes to compute a confidence metric for the measurement from each sensor. Estimated sets and their confidences are then fused using hybrid zonotopes. This method can account for inconsistent measurements, enabling reliable and robust fusion of the sensor data. The effectiveness of the proposed method is demonstrated in experiment.

Keywords: Marine renewable energy systems, Modelling, identification and control in marine systems, Simulation and digital-twin in marine systems
Abstract: The performance of wave energy converters (WECs) relies on both device design and control strategies. This paper presents the modelling and simulation of a non-causal optimal control strategy for an onshore hinged WEC developed by Eco Wave Power Ltd (EWP). A standalone linear non-causal optimal control (LNOC) algorithm is implemented to improve the energy capture efficiency of the WEC. A motor-driven hydraulic pump is used within the hydraulic power take-off (HPTO) system to deliver the control torque. We compare the WEC control performance based on two PTO designs respectively: an existing HPTO design employed by EWP only allowing uni-directional power flow tailored for passive damping control and a modified HPTO design supporting bi-directional power flow suitable for active controller implementation. The HPTO system and the WEC are controlled in a hierarchical control architecture. Simulation results demonstrate that with the new HPTO design the LNOC controller can improve the energy capture output by 19.65% over a well-tuned passive controller. The proposed control scheme is also robust against nonlinear PTO dynamics. The enhanced power absorption shows a 0.57% to 0.63% deviation between the linear and nonlinear PTO models.

Keywords: Marine system guidance, navigation and control, Autonomous marine systems and vehicles, Decision and support in marine systems
Abstract: This paper presents a trajectory prediction method for marine vessels based on optimal planning. Crude initial trajectories respecting static obstacles are first generated using A*-search to provide a feasible warm start. In the second step, a numerical optimizer is used to ensure COLREG compliance. The prediction problem is posed as sequential trajectory planning from the perspective of each surrounding vessel, requiring only their current positions, velocities, and intended destinations as input. As the latter is included in AIS messages, this enables faster predictions than learning-based methods that typically require longer data histories. The proposed method is validated using real-world scenarios constructed from AIS data.

Keywords: Marine system guidance, navigation and control, Autonomous marine systems and vehicles, Marine robotics
Abstract: Predefined time control (PdTC) schemes enable control designers to pre-specify the time in which system trajectories should converge to the origin of the state-space, thereby enabling user-defined control of the system. However, PdTC schemes are comparatively rarer when cascade control structures (CCSs) are used for controller design especially for trajectory tracking problems in marine surface vessels (MSVs). Even when used for CCSs especially in conjunction with sliding mode control (SMC) methods, the controller design is limited to each loop in the CCS. In this paper, a PdTC scheme for an MSV trajectory tracking problem is proposed wherein a predefined time unified sliding surface is designed across both loops of the CCS, which to the best of the author's knowledge, has never been presented before for trajectory tracking problems in MSVs. The resulting controller structure is simple in the relevant literature and needs few parameters to be tuned. The stability of the closed-loop system is shown via Lyapunov theory and numerical simulations are presented to support the proposed approach.

Keywords: Marine system guidance, navigation and control, Nonlinear and optimal automotive control, Autonomous marine systems and vehicles
Abstract: This paper addresses the trajectory-tracking problem for Unmanned Surface Vessels (USVs) in the presence of actuator input limitations. To handle these constraints, we propose a bounded backstepping control strategy that ensures the control inputs remain within predefined limits while preserving the structural advantages of the classical backstepping design. The stability of the closed-loop system is analyzed using Lyapunov theory, allowing the derivation of a control law that guarantees convergence of the tracking errors to zero. The proposed controller is validated through numerical simulations on a fully actuated USV model incorporating nonlinear damping terms and is compared with a classical backstepping approach.

Keywords: Modelling, identification and control in marine systems, AI and embodied-AI in marine systems
Abstract: The objective of this paper is to evaluate the capability of Physics-Informed Neural Networks (PINNs) for parameter estimation and current-induced disturbance identification in a 3-degree-of-freedom (3-DOF) marine vessel model. The proposed approach uses a neural network to approximate the system states, where training is performed by minimizing the residuals of the governing differential equations. The developed algorithm is validated using simulation data.

This work first compares the proposed method with classical parameter estimation techniques for first-order models, both with and without delay. Then, the proposed method is applied to a 3-DOF marine vessel model with unknown parameters and drifting current disturbances. The obtained results demonstrate the performance of the proposed approach.

Keywords: Modelling, identification and control in marine systems, Marine robotics, Autonomous mobile robots
Abstract: This paper introduces a data-driven method for modeling surface vehicle dynamics by integrating a deep learning framework with the Koopman Operator. Ship dynamics involves strong nonlinearities due to hydrodynamic forces and actuation effects, making conventional approaches less effective. In order to address these nonlinearities, the proposed method learns a finite-dimensional linear time-invariant predictor in an observable space, while using monotonic rational-quadratic splines to capture nonlinear input effects. It identifies an end-to-end dynamics model without requiring prior knowledge of the system. Validation on turning and zigzag maneuvers shows high prediction accuracy for surge, sway, and yaw velocities.

Keywords: Multi-agent systems, Control over networks, Nonlinear adaptive control
Abstract: This paper develops a real-time graph learning framework for nonlinear multi-agent systems (MASs) subject to unknown inter-agent interaction dynamics. Unlike prior methods that rely on linear approximations or batch processing, we approximate the unknown interaction dynamics online using a graph neural network (GNN) for a MAS performing trajectory tracking and formation control. To identify the network topology, we leverage memory regressor extension (MRE), which uses a history stack of data to relax standard persistence of excitation conditions. Structural properties of the graph adjacency matrix are enforced by embedding the adaptive update law into a projected dynamical system.

Keywords: Multi-agent systems, Control under communication constraints
Abstract: Addressing the intractability of high-dimensional multi-attacker multi-defender (MAMD) reach-avoid differential graphical (RADG) games, this paper presents a hierarchical control framework for defending a line target against multiple attackers using nonlinear heterogeneous agents. We explicitly decouple the global game into tractable single-attacker multi-defender (SAMD) sub-problems to mitigate the curse of dimensionality. A key innovation is the value-based greedy coalition (VBGC) strategy, which supersedes traditional geometric heuristics by dynamically allocating defenders based on learned game-theoretic value functions, thereby capturing the true heterogeneous dynamics of the team. To ensure consistency between layers, we introduce the concept of admissible partitions, providing rigorous topological constraints that guarantee the existence of a solution for each sub-game. At the execution layer, a distributed solver utilizing approximate dynamic programming (ADP) is developed to generate control policies without requiring persistence of excitation. Numerical simulations demonstrate the framework’s superior performance in coordinating heterogeneous agents compared to distance-based baselines.

Keywords: Multi-agent systems, Randomized algorithms in stochastic systems
Abstract: In this work we consider a group of agents interacting randomly in order to learn the best action. At the beginning of the process, all the agents independently observe a realization of a random signal corresponding to some action, with the best action being the most probable. The initial observations lead to the assignment of initial actions for the agents. Next, the agents observe in discrete time the action of a random agent in the network, and they update their own action. We consider the cases when the set of agents is fixed (closed network) and when the set of agents is time-varying (open network). In both situations we are analyzing if a majority of agents is able to learn the most probable realization of the state of nature. In the closed network case (or when the entry/exit rate is very small), the herding behavior hampers any learning of changes in the state of nature. On the other hand, it is shown that, under suitable conditions, open networks are able to learn the most probable realization even when this realization changes in time. Numerical simulations illustrate our theoretical results.

Keywords: Multi-agent systems, Stochastic control, Control under communication constraints
Abstract: We study self-triggered two-player stochastic games on Piecewise Deterministic Markov Processes (PDMPs), where each agent decides when to observe and which open-loop action to hold. Augmenting the state with clocks and committed controls yields flow regions (both hold) and trigger surfaces (at least one updates). The framework covers both blind simultaneous (Nash) timing and observable sequential (Stackelberg) commitments; the former leads to coupled, intractable QVIs, while the latter admits a nested Hamilton–Jacobi–Bellman quasi-variational inequality and a tractable dynamic-programming decomposition. We outline a computational scheme based on implicit differentiation of the follower’s fixed point. A pursuit–evasion example illustrates the strategic timing interaction.

Keywords: Neural and fuzzy adaptive control
Abstract: Fuzzy neural networks (FNNs) combine the interpretability of fuzzy systems with the self-learning ability of neural networks, yet they struggle with high-dimensional unstructured data, facing challenges of rule explosion and computational collapse. Inspired by cognitive neuroscience, this paper proposes BIHIC, a Brain cognition inspired Interpretable High-definition Image Classification model based on a pre-trained StyleGAN and an FNN. The model transforms StyleGAN's high-dimensional latent codes into low-dimensional disentangled features via a transformation network, mitigating fuzzy rule explosion. The improved FNN utilizes these low-dimensional features for interpretable classification, enhanced by fuzzy rule visualization and feature visualization methods. Experiments on the CelebA-HQ face dataset show that our method maintains high classification accuracy while providing human-intuitive explanations.

Keywords: Nonlinear adaptive control, Model reference adaptive control
Abstract: Despite decades of progress in continuous--time adaptive control, deploying such controllers on digital processors frequently requires altering their adaptive structure, thereby risking the loss of core stability guarantees. To address this challenge, this paper introduces a unified Lyapunov--based framework that establishes a direct correspondence between continuous--time and discrete--time nonlinear affine adaptive control systems. A transformation is developed that maps a wide class of continuous--time adaptive laws to their discrete--time realizations without modifying their structural form. Numerical results demonstrate that the proposed approach achieves robust tracking even at computation rates as low as 10 Hz.

Keywords: Aerial and space robotics, Guidance, navigation and control of aircraft and spacecraft, Guidance, navigation and control for AVs
Abstract: This paper presents a robust nonparametric output regulation framework for altitude-guided navigation and station-keeping of super-pressure balloons. Unlike extremum seeking control, which relies on local gradient estimation, the proposed regulator ensures robust output tracking under uncertain wind dynamics. A bearing–distance objective function is employed to minimize drift and maintain the balloon within a target region. Simulation studies using real atmospheric and wind data demonstrate improved tracking accuracy, reduced sensitivity to disturbances, and an improvement of 5–39 percent in time spent within the station-keeping zone compared to extremum seeking control.

Keywords: Aerial and space robotics, Learning and adaptation in autonomous vehicles, Space exploration and transportation
Abstract: This paper presents NS-TD3, a null-space reinforcement learning framework for trajectory planning of a 7-DOF free-floating space manipulator performing repetitive module transport under abrupt inertia changes. The one-dimensional kinematic redundancy is parameterized by a scalar alpha, and a TD3 agent learns the optimal alpha(t) policy that restores base attitude across full forward-and-homing cycles-a task beyond any instantaneous strategy when payload attachment shifts the inertia coupling matrix mid-cycle. Without explicit knowledge of the inertia-change event, NS-TD3 reduces terminal attitude error by 67% over the minimum-norm (MN) baseline and 42% over a receding-horizon QP (RH-QP) using the same inertia model with explicit mode switching, while satisfying joint constraints and maintaining end-effector accuracy at pseudoinverse-level computational cost.

Keywords: AI for aircraft and spacecraft navigation, guidance and control, Aerospace mission control and operations, Aerial and space robotics
Abstract: This paper presents an autonomous cooperative control framework based on Expert-Guided Proximal Policy Optimization (EG-PPO) for unmanned aerial vehicle (UAV) diamond formation maintenance in cluttered environments. In the proposed framework, an artificial potential field-proportional-integral-derivative (APF-PID) controller is first designed to generate high-quality expert demonstrations, which are used to guide the initial policy learning process. Then, a cooperative reward function that jointly considers formation maintenance and obstacle avoidance is constructed to support multi-objective optimization. To focus on high-level cooperative decision-making, the problem is formulated in a two-dimensional planar environment with constant flight altitude, where the learned policy outputs velocity commands for each UAV. Simulation results show that the proposed EG-PPO method achieves faster convergence, smaller formation-maintenance errors, and smoother trajectories than baseline PPO. The framework also demonstrates good scalability in formation assembly tasks with different swarm sizes.

Keywords: Guidance, navigation and control of aircraft and spacecraft
Abstract: This paper presents a Fuzzy Adaptive Kalman Filter (FAKF) for anomaly and spoofing detection on a 3-DoF helicopter platform. Unlike classical Kalman filters with fixed noise assumptions, the proposed method dynamically updates process and measurement covariances using fuzzy logic driven by innovation statistics and residual consistency measures. An LQI controller regulates the helicopter’s elevation, travel, and pitch, while additive and drift-type spoofing attacks are injected into the sensor channels to evaluate robustness. Simulation results show that the FAKF effectively suppresses corrupted measurements, enhances anomaly sensitivity, and maintains stable state estimation under non-stationary and adversarial conditions.

Keywords: Guidance, navigation and control of aircraft and spacecraft
Abstract: Missions involving rapid and large-angle attitude maneuvers have been conceived for astronomical and Earth observation satellites in recent years. Since the rotational motion of a spacecraft in such missions is nonlinear, it will be required to design an attitude control system that takes into account nonlinear motion. In light of the necessity for an actuator capable of generating large torque, it is imperative to consider the characteristics of an actuator when designing a control system. Actuators capable of generating large torques include the reaction control system (RCS). RCS provides an on/off input by using the reaction force of fuel injection from the thrusters, it can generate a large torque. In addition, the control system of current application satellites normally uses both RCS and reaction wheel (RW) conventionally used for attitude control. For the above reasons, this paper considers large angle attitude maneuver of spacecraft by a combination of RCS and RW. First, characteristics of RCS and RW are defined, and a design model for controller design is derived based on the relative motion equation of the spacecraft. Next, we design a nonlinear tracking controller using the integral sliding mode control (ISMC) method to ensure that the switching function remains bounded. Then, we propose a method to appropriately change RCS injection threshold according to spacecraft attitude by solving an optimization problem.

Keywords: Guidance, navigation and control of aircraft and spacecraft
Abstract: The paper presents a linear parameter varying (LPV) controller design in the induced L2-framework for a tail-free airship. The considered airship is a technology demonstrator developed in the European Innovation Council project IPROP, which ultimately aims to design a high-altitude airship propelled by a novel ionic thruster system. The demonstrator considered in the present paper still uses conventional propellers driven by electric motors, but it already lacks traditional aerodynamic control surfaces and an an empennage. The latter increases aerodynamic efficiency and simplifies the structure but leads to a loss of natural stability. As a novelty, the stabilization and attitude control of the airship is purely achieved by differential thrust allocation. The LPV controller is scheduled with the airship's airspeed which enables a significantly larger flight envelope compared to a linear time invariant controller. The performance and robustness of the controller are evaluated in a high-fidelity nonlinear simulator.

Keywords: Guidance, navigation and control of aircraft and spacecraft, Aerial and space robotics, Aerospace mission control and operations
Abstract: Tilt-wing unmanned aerial vehicles (UAVs) combine the vertical takeoff and landing capability of multi-rotors with the high-speed cruise efficiency of fixed-wing aircraft, but their transition phase involves strong aerodynamic coupling and time-varying control effectiveness. To improve robustness under uncertainties, this paper proposes a robust optimal transition trajectory optimization method for tilt-wing UAVs. Unlike existing deterministic optimization approaches, the proposed method explicitly accounts for stochastic uncertainties in the initial state, propeller thrust coefficients, and aerodynamic parameters. Correlated uncertainties commonly observed in coupled flight dynamics are efficiently decoupled using the Gram–Schmidt transformation, avoiding the need to construct new orthogonal polynomial bases. Moreover, a sinusoidal transformation of control inputs and an extended penalty function are introduced to convert the constrained optimization into an unconstrained formulation, simplifying numerical computation while ensuring control feasibility. Simulation results demonstrate that the proposed method significantly enhances the robustness of the transition flights.

Keywords: Guidance, navigation and control of aircraft and spacecraft, Aerospace mission control and operations
Abstract: This paper proposes an optimal control framework for designing avoidance maneuvers to protect a target satellite from intentional jamming by a non-cooperative spacecraft. The framework integrates J2-perturbed orbital dynamics with a link-budget-based jamming-to-signal ratio model and employs a smoothed antenna gain representation suitable for direct collocation. The resulting optimal control problem simultaneously minimizes jamming exposure and maneuver cost while enforcing recovery of the primary orbital elements after the avoidance maneuver. In the numerical case study, the proposed method reduces the maximum J/S ratio from 21.39 dB in the no-maneuver case to 0.72 dB and decreases the duration above the jamming-risk threshold by approximately 85%. Compared with a reactive baseline, the proposed method reduces the total delta-V by approximately 95%, demonstrating its ability to avoid jamming while efficiently recovering orbital geometry.

Keywords: Guidance, navigation and control of aircraft and spacecraft, Aerospace mission control and operations, Aerial and space robotics
Abstract: This work presents a control scheme for trajectory tracking of multirotor UAVs featuring finite-time convergence of both position and attitude. The attitude error is expressed in the Lie algebra via the logarithmic map of SO(3), which transforms geodesics on the rotation manifold into straight lines in the Lie algebra, thereby providing the most natural and effective representation of attitude error. Both the position and the attitude controllers are designed based on a unified framework of linear-nonlinear sliding-mode control where the coexistence of the linear and fractional-exponent terms induces an adaptive two-phase dynamic behavior: a smooth decay is governed by the linear component when the error is relatively large, while the nonlinear fractional term accelerates convergence as the state approaches the neighborhood of the origin. In both regions, the corresponding dominant term ensures that the tracking error continues to converge rapidly toward zero. Computer simulations were conducted to validate the approach, and the preliminary results demonstrated the effectiveness of the proposed method supporting its feasibility in the UAV applications.

Keywords: Guidance, navigation and control of aircraft and spacecraft, Flight dynamics modelling and identification, Aerospace mission control and operations
Abstract: This paper addresses simultaneous attitude stabilization and angular momentum management for spacecraft operating in Very Low Earth Orbits (VLEOs), where aerodynamic disturbance torque is significant. Conventional momentum unloading methods often rely on auxiliary aerodynamic surfaces or specialized mechanisms, increasing system mass and complexity. In contrast, this work exploits naturally available aerodynamic torque while using reaction wheel torques as the sole commanded actuators. A two-loop adaptive sliding mode controller is developed, where the adaptive switching-gains in both loops provide robustness to aerodynamic model uncertainty. The control law is developed based on a nominal aerodynamic torque model derived from spacecraft symmetry, whereas the aerodynamic torque in simulation is generated from high-fidelity Direct Simulation Monte Carlo (DSMC) data for a realistic spacecraft geometry, so that closed-loop robustness is assessed under deliberate model–plant mismatch. Numerical simulations at 300 km altitude demonstrate asymptotic convergence of the attitude error and angular momentum over multiple orbital periods without any additional actuators or dedicated aerodynamic surfaces.

Keywords: Guidance, navigation and control of aircraft and spacecraft, Guidance, navigation and control for AVs, Trajectory and path planning for AVs
Abstract: This paper presents a new Enhanced Finite-Time Convergent Spatiotemporal Constrained Guidance Law (EFTG), which is achieved by augmenting an exact spatiotemporal guidance baseline with faster time-error convergence and saturation-aware compensation mechanisms. Rather than redefining the precise underlying time-to-go predictor, this study retains the exact baseline angle/time coordination structure and introduces a nonlinear finite-time time-control term, an auxiliary anti-saturation compensator and an H_infty-inspired dynamic output-feedback fusion strategy. Simulation results show that the EFTG method maintains accurate impact-time and impact-angle coordination while enhancing robustness, command smoothness and terminal precision during aggressive manoeuvres.

Keywords: Guidance, navigation and control of aircraft and spacecraft, Space exploration and transportation
Abstract: The exploration of a planet often requires a spacecraft to enter a low circular orbit for detailed scientific observation. Aerobraking is an orbital maneuver which aims to lower the apoapsis by passing multiple times through the planet’s atmosphere. Aerobraking enables propellant mass savings but requires that the altitude at periapsis is properly maintained within a safe corridor to ensure sufficient drag for orbit reduction while preventing destructive heating. This paper proposes a periapsis altitude control strategy for Mars aerobraking using nonlinear model predictive control (NMPC) and continuous low-thrust propulsion. The designed NMPC optimizes a finite-horizon trajectory by predicting the altitude at periapsis and computing optimal thrust profile to ensure that the spacecraft’s altitude at periapsis is maintained within the prescribed safe corridor with minimal use of low-thrust actuators. Quantization of thrust values computed by the NMPC is handled by a pulse-width modulation (PWM) scheme to generate on-off commands to low-thrust actuators. Simulations show that the spacecraft’s altitude at periapsis is successfully maintained within the safe corridor despite atmospheric variations and uncertainty, thereby demonstrating that the proposed control strategy enhances robustness and autonomous capabilities for aerobraking maneuvers with minimal propellant consumption and enables more science.

Keywords: Guidance, navigation and control of aircraft and spacecraft, Trajectory tracking and path following for AVs, Space exploration and transportation
Abstract: Scanning the Earth surface with space-based imagers often involves aligning the sensor axis towards a fixed orientation with respect to a local-horizontal-local-vertical frame. The direction along the sensor plane is typically constrained to align towards the spacecraft velocity vector to minimize smear and obtain consistent overlapping images along a path on the surface. Because the surface may be non-Euclidean, we show that a more precise approach during imaging is to constrain the sensor plane towards the tangent velocity vector at the point where the sensor line-of-sight intersects the surface. We discuss prerequisites for claiming a PD-like attitude tracking controller with feedforward is stabilizing. We present strategies and numerical results for handling this time-varying attitude control problem given a non-Euclidean surface and an orbit.

Keywords: Guidance, navigation and control of aircraft and spacecraft, Vehicle dynamic systems, Flight dynamics modelling and identification
Abstract: This paper examines when planar landing-guidance approximations become unreliable for reusable rocket landing under off-nominal initial conditions. A full six-degree-of-freedom (6--DoF) nonlinear optimal control problem is transcribed with Dymos and solved using the Interior Point OPTimizer (IPOPT), and its solutions are compared against a 2D--composed 3D baseline obtained by combining two decoupled planar optimizations. Rather than proposing a new optimization algorithm, the paper aims to quantify the regime in which planar guidance remains adequate and the regime in which full 6--DoF optimization becomes necessary. Numerical results show that the planar composition can provide an effective warm start, but it tends to underestimate fuel usage and exhibits larger dynamics defects in strongly off-nominal, non-planar cases.

Keywords: Information processing and decision support in transportation, Artificial intelligence in transportation, AI for aircraft and spacecraft navigation, guidance and control
Abstract: Low-cost unmanned aerial vehicle (UAV) swarms create coupled threat and response-urgency assessment challenges for air defense systems. Many existing methods provide useful baselines, yet they rarely encode semantic feature groups, swarm-coordination synergy, and threat-urgency coupling in one model. This paper proposes the Hierarchical Group Threat Attention Network (HGTAN), which combines a group-wise feature encoder, a Synergy Attention Module, and a dual-task decoder. A 16-indicator UAV swarm benchmark is organized into individual, swarm, adversarial, and environmental groups. Experiments on controlled synthetic scenarios show that HGTAN achieves strong dual-task performance and interpretable group-level attention patterns, with 0.923 threat macro-F1 and 0.886 urgency macro-F1.

Keywords: Information processing and decision support in transportation, Intelligent transportation systems, Automatic control, optimization, real-time operations in transportation
Abstract: This survey presents a comprehensive study of V2X-supported vehicle fault diagnostics, with a specific emphasis on its transformative applications in increasingly complex automotive networks. With the rapid evolution of connected and autonomous vehicles, conventional in-vehicle diagnostic systems face limitations in providing proactive and holistic fault detection. We systematically review recent research that explores how Vehicle-to-Everything (V2X) communication facilitates enhanced fault identification, predictive maintenance, and real-time anomaly resolution by enabling seamless integration with external services. Key areas of discussion include architectural paradigms for secure data exchange, distributed diagnostic processing leveraging cloud-based platforms, and the critical role of robust V2X connectivity for real-time vehicular electronic health monitoring. This work synthesizes emerging applications and identifies pivotal research challenges for practical deployment, underscoring the significant potential of these integrated approaches to elevate vehicle safety and operational efficiency.

Keywords: Intelligent transportation systems, Automatic control, optimization, real-time operations in transportation, Control architectures in automotive control
Abstract: This paper proposes a resilient control algorithm to enhance the security of Cooperative Adaptive Cruise Control in vehicular platoons with unidirectional communication and heterogeneous dynamics subject to actuator attacks. The proposed framework employs a model reference adaptive control scheme to drive the heterogeneous platoon toward a reference homogeneous platoon using a distributed structure. Stability and convergence are established through Lyapunov analysis using a virtual platoon concept, which is employed solely for analysis and does not interact with the actual system. Simulation results demonstrate the effectiveness of the proposed algorithm against actuator attacks.

Keywords: Intelligent transportation systems, Modeling and simulation of transportation systems
Abstract: In this paper, we develop a hybrid prediction framework for accurate electric vehicle (EV) charging time estimation, a capability that is critical for trip planning, user satisfaction, and efficient operation of charging infrastructure. We combine a physics-informed gradient boosting model with a reinforcement learning (RL) approach. The physics-informed component captures the nonlinear constant-current/constant-voltage (CC–CV) charging dynamics and explicitly models state-of-health (SoH)–dependent capacity and power fade, providing a reliable baseline when historical data are limited. Building on this foundation, we introduce an RL component that progressively refines charging-time predictions as operational data accumulate, enabling improved long-term adaptation. Both models incorporate SoH degradation to maintain predictive accuracy over the battery lifetime. We evaluate the framework using 5,000 simulated charging sessions calibrated to manufacturer specifications and publicly available EV charging datasets. Our results show that the physics-informed gradient boosting model achieves coefficient of determination R2 =98.5% and mean absolute percentage error MAPE=2.1%, while the RL model further improves performance to R2 =99.2% and MAPE=1.6%, corresponding to a 23% accuracy gain over the physics-informed model and 35% improved robustness to battery aging compared to a linear baseline.

Keywords: Intelligent transportation systems, Modeling and simulation of transportation systems, Automatic control, optimization, real-time operations in transportation
Abstract: In this paper, we investigate traffic signal control in a network of interconnected intersections, aiming to balance lane-level vehicle densities through optimal green-time allocation. We develop a two-lane traffic flow model that explicitly captures lane-specific propagation dynamics, addressing key limitations of conventional road-level formulations. The proposed model offers a more granular and flexible representation of urban traffic, enabling controllers to react more accurately to lane-specific congestion patterns. Building on this model, we design a distributed model predictive control (MPC) framework and integrate it with the efficient alternating direction method of multipliers (ADMM) to enhance scalability and real-time performance. To accommodate time-varying traffic conditions, we further introduce a data-driven method for forecasting dynamic split ratios. Comprehensive VISSIM simulations on a six-intersection network in Dalian, China, demonstrate that the proposed approach outperforms existing signal control strategies in both traffic efficiency and computational speed, showing its promise for real-time deployment.

Keywords: Intelligent transportation systems, Transportation logistics, Information processing and decision support in transportation
Abstract: In this paper, we investigate cooperative platoon formation and benefit allocation in mixed-energy truck fleets composed of both electric and fuel-powered trucks. The central challenge arises from the platoon-size constraint, which limits the number of trucks permitted in each platoon and introduces combinatorial coupling into the search for optimal platoon formation structures. We formulate this problem as a coalitional game with bounded coalition sizes and derive a closed-form characterization of the optimal coalition structure that maximizes the fleet-wide platooning benefit. Building on this structure, we develop a type-based least-core payoff allocation scheme that guarantees stability within the coalition-structure core (CS-core). For cases in which the CS-core is empty, we compute the least-core radius to determine the minimal relaxation required to achieve approximate stability. Through numerical studies, we demonstrate that the proposed framework consistently achieves the highest total platooning benefit among all feasible formation configurations while providing stable benefit allocations that outperform existing baseline methods.

Keywords: Planning, management and security in transportation, Electric and solar vehicles, Trajectory and path planning for AVs
Abstract: This study proposes an integrated electric vehicle (EV) route planning framework that extends battery lifespan by jointly considering the effects of 3D road topography and road-induced vibration. By formulating a combined cost function that mathematically unifies elevation-based energy consumption with vibration-induced thermal stress, the research evaluates trade-offs between travel distance, energy efficiency, and long-term capacity decay on a complex road network. Minimizing elevation changes and vibration produces much smoother energy usage. Specifically, the minimum-elevation route strikes the optimal balance, achieving the slowest capacity fade and the best final State of Charge (SoC), highlighting the fact that incorporating both mechanical and electrical stressors into routing decisions is essential for enhancing long-term battery health.

Keywords: Planning, management and security in transportation, Trajectory and path planning for AVs, Mission planning and decision making for AVs
Abstract: This paper presents a visibility-aware motion planner for autonomous driving in multi-occlusion scenarios. Critical blind spots are identified, and a set-based estimation method is used to infer the states of hidden traffic participants. A visibility cost, formulated from the temporal evolution of these state sets, guides lateral offset planning to actively enlarge the field of view of the ego vehicle. The proposed hierarchical planner incorporates this visibility cost into a sampling-based lateral path generator, followed by Hybrid A ∗ -based longitudinal speed planning. Simulations in unsignalized intersections, pedestrian dart-out, and two-way overtaking scenarios demonstrate that the proposed planner improves driving efficiency and reduces the time required to expose blind spots, while maintaining safety feasibility and ride comfort.

Keywords: Rail transportation modelling and control systems, Artificial intelligence in transportation, Automatic control, optimization, real-time operations in transportation
Abstract: Heavy-haul yards need minute-scale forecasts for arrival, unloading, and departure to manage queues and resources. In this context, the literature is largely passenger-oriented and does not provide multi-stage heavy-haul forecasts that use both operational and temporal information. This study addresses that gap by developing an interpretable, multi-stage forecasting pipeline and quantifying gains over operator-fixed targets. Using about one million industrial timestamps, we reconstruct event-level durations and train stage-wise gradient-boosted models with operational, temporal, and congestion features; evaluation uses an unseen-train split by train name. The models achieve mean absolute error of 65.3, 16.1, and 114.6 minutes for arrival, unloading, and departure, respectively, with a 51 to 80 percent reduction compared with fixed stage targets. Feature-importance analysis indicates operational signals dominate. The forecasts enable earlier crew calls and equipment staging and support a large-scale simulation toward a cognitive digital twin of yard operations.

Keywords: Rail transportation modelling and control systems, Control architectures in automotive control, Nonlinear and optimal automotive control
Abstract: This paper presents a hierarchical MPC architecture for large-scale systems, motivated by railway control applications. The approach decomposes the problem into a high-level linearized MPC for global coordination and a low-level layer ensuring feasibility under the true nonlinear dynamics. Coupling constraints and costs are handled at the high level, which provides reference trajectories to the low level. The low level further corrects high-level model approximations online, improving feasibility and optimality. The resulting design bridges centralized and decentralized control, offering a scalable and close-to-optimal control strategy. Its effectiveness is validated through simulations of a multi-train scenario, showing improved coordination and energy-efficient operation.

Keywords: Control and optimization for sustainability and energy systems, Wind power, Control and management of energy systems
Abstract: Airborne Wind Energy (AWE) farms exhibit intrinsic power oscillations due to the cyclic traction–retraction operation of individual units. Previous work has addressed layout optimization and energy storage sizing under constant wind assumptions; however, realistic operation involves time-varying wind speeds that introduce additional dynamic variability in aggregate power output. This paper proposes a Model Predictive Control (MPC) framework for real-time charge–discharge management of a battery energy storage system to ensure smooth power injection into the grid under variable wind conditions. A dynamic model of the AWE farm power output coupling wind-dependent farm power generation and battery state-of-charge evolution is considered, including operational and safety constraints. The MPC controller minimizes power tracking error while enforcing state-of-charge and power limits. Preliminary simulation results indicate that predictive energy management significantly attenuates wind-induced fluctuations. The work opens a discussion on hierarchical smoothing strategies and predictive control architectures for large-scale airborne wind energy integration.

Keywords: Control of complex systems, Analytic design
Abstract: System-level disturbance suppression in mechatronic systems can be achieved using a feedthrough (FT) control framework that leverages interconnections between subsystems. This paper explicitly incorporates disturbance models to overcome fundamental performance limitations of conventional FT control methods.

Keywords: Control of hybrid systems, Sampled-data/digital control, Learning methods for optimal control
Abstract: This study examines the two-sided looped functional-based synchronization problem for semi-Markov jump neural networks via an intelligent event-triggered control technique. Specifically, an intelligent event-triggered control is formulated to reduce the communication cost and mini-batch gradient-based machine learning algorithm is deployed to optimize the threshold parameter in event-trigger control. Further, a two-sided looped type Lyapunov functional is configured and significant advantage of the suggested method is its elimination of the traditional requirement for a standard positive definite matrix in Lyapunov formulation. From there, sufficient requirements for ensuring synchronization are delineated within the context of linear matrix inequalities. To substantiate the validity and utility of the proposed control law, numerical example along with the graphical illustrations are offered.

Keywords: Control of multi satellite systems, Aerospace mission control and operations, Guidance, navigation and control of aircraft and spacecraft
Abstract: This paper addresses the problem of multi-satellite constellation control for allocating shared service areas among satellites, taking into account their orbital motion, illumination conditions, and battery charge levels. A corresponding modified assignment problem for task allocation is formulated. A general formulation of the unbalanced assignment problem with additional constraints is considered. A relaxation scheme to a linear programming problem through a series of auxiliary problems is proposed. A model problem of continuous-coverage satellite constellation control is considered as a case study.

Keywords: Cyber-physical and human systems (CPHS)
Abstract: Work engagement is the state where employees are energetic, enthusiastic, and absorbed in their work. The Job Demands-Resources Model (JD-R Model) examined the effects of work environment, worker psychology to work engagement. The purpose of this study is to control work engagement through designing a control system based on the JD-R Model. The system was designed using Matlab/Simulink and verified by simulations. By designing a feedback control system using a PI controller, it was possible to track work engagement to the target value. Future prospects are to verify the usefulness through experiments and to expanded the system from an individual to multiple persons.

Keywords: Cybersecurity in smart grids, Power systems stability, Cyberphysical security in processes
Abstract: In this work, we introduce a new class of attack in the power-grids, where an intruder suddenly isolate a single or multiple bus-bars from the grid during the normal operation. This type of attack is referred to as bus-bar isolation attack. The impact of such attack is studied here through a standard IEEE-5 bus power-grid. To bring resiliency against such type of attacks, we propose a novel feedback control architecture and its design procedure. For this, we use the descriptor form model (linearized model of original differential-algebraic equation model) of power-grid, and show that the traditional (transformed) ordinary state space model of power-grid does not provide sufficient insight to design such resilient feedback control.

Keywords: Data-driven control theory, Learning methods for control
Abstract: This study investigates discrete-time nonlinear systems with symbolic inputs and outputs and proposes a data-driven approach for controller design based solely on observed input-output sequences. Here, “Symbolic” means that ordinal or distance relationships among symbols are unknown a priori and must be implicitly reconstructed from time-series data. The control objective is symbolic output tracking to a specified reference symbol. The controller is trained through a supervised learning procedure using a feedforward neural network. The results of numerical simulations based on benchmark nonlinear systems demonstrate that the proposed approach is effective.

Keywords: Disturbance rejection and input-to-state stability, Lyapunov methods, Saturation and discontinuity
Abstract: Two-axis clinostats simulate microgravity through time-averaged simulated microgravity (taSMG) by continuously reorienting the gravity direction vector. Although the reciprocal sinusoidal angular velocity profile achieves near-ideal kinematic uniformity, hardware experiments reveal a persistent gravity offset exceeding the 10-3 G threshold, attributable to a position-dependent load torque asymmetry on the outer-axis motor. This paper formalizes the dynamic origin of this residual bias through perturbation analysis of the nonlinear rotational dynamics, and proposes a cascaded algorithmic framework comprising: (1) a perturbation-derived bias-compensated reciprocal profile; (2) a torque-aware adaptive angular acceleration limiter; and (3) an integral feedback drift cancellation loop with Lyapunov-certified asymptotic stability. The proposed extensions require only software modification and one-time offline calibration, without additional actuators or structural redesign, offering a hardware-agnostic upgrade pathway for high-precision long-duration simulated microgravity experiments.

Keywords: Distributed control and estimation, Filtering and smoothing, Probabilistic and Bayesian methods for system identification
Abstract: We study the problem of reproducing a given multivariate power spectral density via a distributed system, where each node generates a component and exchanges minimal information to match cross-correlations. We introduce the notion of graphical spectral factorization, which decomposes the process into independent innovations and a structured communication network among nodes. In the non-causal case, conditional uncorrelation defines a spectral graphoid, enabling an acyclic structure of the network. For real-time, causal implementations, each process is split into past and present components to preserve causality while reasoning about sparsity. This framework provides a systematic approach for designing distributed stochastic systems that realize a target power spectral density with structured, sparse communication.

Keywords: Data-driven control theory, Multi-agent systems
Abstract: We propose an input-output data-driven framework for certifying the stability of interconnected multiple-input-multiple-output linear time-invariant discrete-time systems via QSR-dissipativity under a structured class of supply rates motivated by the considered interconnection structure. That is, by using measured noise-free input-output trajectories of each subsystem, we verify QSR-dissipativity and extract local channel-wise passivity indices. These passivity indices are then used to derive conditions under which the equilibrium of the interconnected system is stable. In particular, the framework identifies how the lack of passivity in some subsystems can be compensated by surpluses in others. The proposed approach enables a compositional stability analysis by combining subsystem-level conditions into a criterion valid for the overall interconnected system. We illustrate via a numerical case study, how to compute channel-wise passivity indices and infer stability guarantees directly from data with the proposed method.

Keywords: Industrial artificial intelligence, Industry X.0 for production and logistics, Intelligent manufacturing systems
Abstract: Assessment of automotive paint quality in real production environments remains a multifaceted challenge influenced by numerous interacting factors. Even minor deviations in application techniques or material properties may propagate into measurable defects. Automatization of paint quality evaluation is the current trend in this area; however, not all paint quality control processes can be fully automated with the required accuracy and confidence. This issue is significantly more important when manufacturing premium car models. Therefore, it is necessary to align the automated and semi-manual evaluation of paint quality to minimise human influence. To address this issue, we have proposed a data driven evaluation platform that integrates evaluation data from multiple systems to discover new knowledge. The implemented platform utilises statistical analysis and machine learning approaches required for the complex evaluation of the quality of the paint structure. The platform was deployed in the real production environment and provides an improved approach to evaluating the quality of the paint structure and was integrated into the quality evaluation processes of the company.

Keywords: Resilient networked control systems, Learning methods for control, Control over networks
Abstract: This paper studies optimal control over networks under packet losses that may be adversarial, such as denial-of-service attacks. Existing stochastic models of packet losses do not capture strategic attacks, while designing controllers for worst-case packet-loss patterns is often overly pessimistic. To address this issue, we formulate the networked optimal control as an online control problem that aims to minimize regret. Here, regret measures the performance loss relative to the best controller in hindsight. Then, we design an online controller that achieves an widetilde{O}(sqrt{T}) regret over a control horizon T . The guarantee holds for any packet-loss sequence satisfying certain assumptions on its frequency and success-to-failure ratio, and remains valid even in the presence of adversarial disturbances and unknown cost functions.

Keywords: Industrial artificial intelligence, Intelligent manufacturing systems
Abstract: This work explores the use of artificial intelligence in mobile robotics to achieve autonomous detection and pose estimation of load carriers for automated pickup. A deep neural network is designed to recognize predefined landmarks on the carrier from RGBD data; these landmarks are then used to compute the carrier’s pose. The network operates directly on RGBD images to estimate landmark positions, which form the basis for determining the carrier’s location. The approach is validated in extensive experiments and comprises both software and hardware implementations.

A deep learning–based framework is presented to detect load carriers and estimate their pose for use with autonomous logistics vehicles. Our method uses a convolutional neural network to identify characteristic reference points on the carrier from RGBD input and computes its pose by combining these inferred landmarks with prior geometric knowledge. Experiments show that the resulting accuracy is sufficient for reliable load carrier detection in industrial environments, confirming the suitability of the method for autonomous intralogistics applications.

Keywords: Multi-agent systems, Linear system identification
Abstract: In this work we evaluate the excitation and measurement patterns (EMP) for networks with tree topology. We investigate guidelines for the selection of the minimal EMPs, i.e. those with the least number of excited and measured nodes combined, for which the accuracy obtained, in terms of the trace of the asymptotic covariance matrix, is optimal. We introduce the concept of partial information matrix as a means to systematically obtain the information matrix for any dynamic network. For a specific tree class, called cross, we show that the accuracy of a particular module depends on the magnitude of the parameters to be estimated. Furthermore, when all factors are equal, it is best to excite. We extend a topological condition for branches under which the accuracy of a particular module of the network is independent of the other parameters from the tree. We provide a numerical analysis showing that our guidelines could be used as a selection tool for minimal EMPs for tree networks.

Keywords: Diagnosis of discrete event and hybrid systems, Discrete event modeling and simulation, Event-based control
Abstract: This work investigates how the control parameter influences the triggering behaviors of event-triggered control (ETC) systems, focusing on its role in changing the feasibility of the inter-event time (IET) transitions. To this end, the feasibility condition is reformulated as an equivalent linear feasibility problem. This reformulation transforms the analysis to a geometric problem of determining when the null space intersects the fixed polyhedral cone. As the control parameter varies, the null space changes accordingly. A control parameter value is identified as a critical value if, at that value, the null space and the polyhedral cone switch between having an intersection and being disjoint. Consequently, any changes in the feasibility of the IET transition can occur only at such critical values. To capture all possible changes, we construct a candidate set that contains all possible critical values. This result enables the exact identification of all critical values, providing a complete and analytically tractable characterization of how the control parameter governs the triggering behaviors of ETC systems.

Keywords: Event-based control, Control over networks
Abstract: By configuring a laser interferometer among three spacecraft in orbit, a large-scale gravitational wave telescope can be constructed, which enables observations of low-frequency gravitational waves. To establish and maintain such a space interferometer, the laser must be tracked to the counterpart spacecraft with high accuracy. This study focuses on the problem of laser tracking control under limited inter-spacecraft communication for three-spacecraft formation flying in generalized circular orbits. In particular, we propose an event-triggered control for regulating the laser emission angles so that the interferometry laser link is consistently maintained. A numerical example illustrates that the proposed controller effectively reduces the communication load while maintaining the laser link among spacecraft.

Keywords: Event-based control, Control over networks, Kalman filtering
Abstract: This paper addresses bumpless transfer in event-based cloud control of a three-tank pilot plant. The proposed architecture coordinates a cloud-hosted nonlinear model predictive controller (NMPC) and local PID loops running on ABB System 800xA and a Raspberry Pi+Arduino edge stack. The main methodological novelty is a state-of-the-art ring-buffer mechanism that provides precise clock synchronization among the three device controllers and the Unscented Kalman Filter, enabling more advanced bumpless switching under heterogeneous communication delays. Two practical mechanisms—(i) UKF-based state alignment and (ii) a short command blend with PID integrator pre-initialization—enable seamless switching between controllers under realistic round-trip delays (150 ´ms cloud, 50–60ms local). A physics-based orifice model is used within NMPC, while plant state is estimated with an Unscented Kalman Filter (UKF). On a comprehensive scenario with cloud outages and latency inflation, the supervisor enforces bumpless handovers, preserves tight level tracking, and maintains low integral absolute error (IAE), as evidenced by smooth actuator trajectories and output levels. Compared with single-controller operation, the event-based supervisory scheme exploits the cloud when available and safely falls back to local control otherwise, without inducing detrimental bumps. The study demonstrates a reproducible path to deploy NMPC-in-the-cloud for industrial Internet of Things (IIoT) systems using an OPC DA/UA bridge and ABB integration, and provides an implementation blueprint suitable for replication on pilot rigs. The results are analyzed and discussed.

Keywords: Event-based control, Distributed control and estimation, Multi-agent systems
Abstract: This paper investigates a neighbor-to-neighbor event-triggered consensus control problem for a class of strict-feedback multi-agent systems under a directed communication graph, aiming to address the conflict between the requirements of distributed feasible consensus control and the limitations of communication resources in networked control systems. First, a triggering module is developed for each follower, integrating a neighbor-to-neighbor event-triggering mechanism to ensure that only sampled states are transmitted among neighboring agents. Then, a control module composed of multiple filter banks and a controller component is designed. Each filter bank contains two first-order low-pass filters connected in cascade, which take the sampled neighboring states as inputs and generate continuous and differentiable outputs used in the corresponding distributed dynamic surface controller. Consequently, the proposed approach avoids the non-differentiability of virtual control signals and can be implemented without relying on global information of the communication topology. The closed-loop stability analysis of the proposed scheme is presented and the Zeno behavior is excluded. Finally, simulation results demonstrate the effectiveness and applicability of the proposed strategy.

Keywords: Event-based control, Supervisory control and automata
Abstract: We investigate the combined observation of events and states in supervisory control of a discrete-event system. We compute the set of all state-event pairs such that the appearance (as transitions) of any such pair in an uncontrolled system and the nonexistence of a supervisor, no matter which controllable event set it uses, are equivalent.

Keywords: Reachability analysis, verification and abstraction of hybrid systems, Supervisory control and automata, Control under communication constraints
Abstract: Networked control systems (NCS) are widely used in safety-critical applications, but they are often analyzed under the assumption of ideal communication channels. This work focuses on the synthesis of safety controllers for discrete-time linear systems affected by unknown disturbances operating in imperfect communication channels. The proposed method guarantees safety by constructing ellipsoidal robust safety invariant (RSI) sets and verifying their invariance through linear matrix inequalities (LMI), which are formulated and solved as semi-definite programming (SDP). In particular, our framework simultaneously considers controller synthesis and communication errors without requiring explicit modeling of the communication channel. A case study on cruise control problem demonstrates that the proposed controller ensures safety in the presence of unexpected disturbances and multiple communication imperfections simultaneously.

Keywords: Fault detection and diagnosis
Abstract: To address the challenges of modality heterogeneity and cross-device generalization in fault diagnosis, this paper proposes a Device-Agnostic Modality-Adaptive Perception Embedding (MAPE) and a Universal Time-Frequency Aggregation Transformer (TFAformer) for unknown-domain fault diagnosis. MAPE adaptively extracts unified time, frequency, and time–frequency features from multi-modal signals, regardless of modality type or quantity. TFAformer aggregates global and local representations via cross-attention and self-attention, capturing both private and shared characteristics. Extensive experiments on 10 models show MAPE outperforms existing embeddings by 7.23%, and the combined framework achieves average 90.20% accuracy on industrial robot, wind turbine, and bearing datasets, demonstrating strong generalization across devices and domains. The proposed framework is open-sourced at https://github.com/qtchen0730/TFAformer.

Keywords: Fault detection and diagnosis, Data-driven control theory, Machine and deep learning for system identification
Abstract: Fault detection and identification (FDI) is critical for maintaining the safety and reliability of systems subject to actuator and sensor faults. In this paper, the problem of FDI for nonlinear control-affine systems under simultaneous actuator and sensor faults is studied. We model fault signatures through the evolution of the probability density flow along the trajectory and characterize detectability using the 2-Wasserstein metric. In order to introduce quantifiable guarantees for fault detectability based on system parameters and fault magnitudes, we derive upper bounds on the distributional separation between nominal and faulty dynamics. The latter is achieved through a stochastic contraction analysis of probability distributions in the 2-Wasserstein metric. A data-driven FDI method is developed by means of a conditional flow-matching scheme that learns neural vector fields governing density propagation under different fault profiles. To generalize the data-driven FDI method across continuous fault magnitudes, Gaussian bridge interpolation and Feature-wise Linear Modulation (FiLM) conditioning are incorporated. The effectiveness of our proposed method is illustrated on a spacecraft attitude control system, and its performance is compared with an augmented Extended Kalman Filter (EKF) baseline. The results confirm that trajectory-based distributional analysis provides improved discrimination between fault scenarios and enables reliable data-driven FDI with a lower false alarm rate compared with the augmented EKF.

Keywords: Fault detection and diagnosis, Gaussian process, Nonlinear system identification
Abstract: In this paper we address the problem of detecting differences or anomalies in a dynamical system, based on historical data of nominal operations. This problem encompasses quality control, where newly manufactured systems are tested against desired nominal operations, and the detection of changes in the dynamics due to degradation or repairs. We propose a model free approach based on Gaussian processes (GPs). The idea is to train offline a GP based on nominal data, which is then deployed online to detect whether measurements of the system's state are compatible with nominal operations or if they deviate. Detection is made more challenging by the presence of process and measurement noise, which might obfuscate deviations in the dynamics. The detection then is based on a threshold that ensures a specific false positive rate. We showcase the promising performance of the proposed method with two systems, and highlight several interesting future research questions.

Keywords: Fault detection and diagnosis, Learning methods for control, Adaptive observer design
Abstract: This paper develops a fault detection and identification (FDI) method for nonlinear control-affine systems under simultaneous actuator and sensor faults. We adopt a geometric approach to study the isolability of faults in the sense of the principal angles between subspaces corresponding to each actuator and sensor fault. As for the fault identification, a hybrid estimator that consists of a Luenberger-like observer with contraction guarantees is developed. Moreover, neural networks are embedded in the mentioned observer to estimate actuator and sensor faults. Considering that the training dataset for neural networks cannot be representative of every fault scenario, the last layer of each network is adapted using mirror descent-based laws. The mirror descent-based adaptive laws impose isolabilty conditions for fault channels and do not assume a quadratic parameter estimation space to preserve the geometry of the fault subspaces. A Lyapunov-based analysis establishes that the state and parameter estimation errors are uniformly ultimately bounded. The effectiveness of our proposed FDI method is illustrated on 3-axis attitude control system of a spacecraft.

Keywords: Fault detection and diagnosis, Learning methods for control, Nonlinear system identification
Abstract: In this paper, we propose a method to learn a manifold representation of data covariance Hankel matrices using an autoencoder, and demonstrate its use for fault detection in dynamical systems. The autoencoder is trained in two stages to capture both a low dimensional latent manifold representation and a complementary representation of Hankel matrices constructed from nominal data. We illustrate that when a Hankel matrix built from test data does not lie on a manifold obtained on data collected from a nominal reference system, a fault occurs. To detect it, we propose a simple residual which is tested in a classical hypothesis testing framework. For linear systems, this residual is closely related to the classical subspace fault detection residual based on the null space of the Hankel matrix. The performance of the proposed detection scheme is validated on linear and weakly nonlinear mechanical system.

Keywords: Advanced process control
Abstract: Industrial control system oscillations incur substantial economic costs through energy waste, accelerated equipment wear, and compromised product quality. Traditional diagnostic methods, reliant on manual expert scrutiny of data from thousands of control loops, present a critical scalability bottleneck. This tutorial presents a novel, scalable framework that integrates Large Language Models (LLMs) with specialized oscillation detection toolboxes via a Retrieval-Augmented Generation (RAG) architecture. The system features a programmatic command-line interface, decoupling analysis from graphical user interfaces and enabling automation. Our domain-specific RAG pipeline dynamically contextualizes LLM responses by retrieving relevant real-time analytical outputs and structured technical knowledge. This allows plant personnel to conduct investigations using natural language queries. The framework further enhances diagnostic reliability by applying LLM reasoning to interpret the results of signature algorithms, such as triangle-like shape detection for valve stiction. Rigorous validation on industrial datasets—including refinery process data, the International Stiction Database, and the Tennessee Eastman Process benchmark—demonstrates the system's robust performance, achieving excellent classification accuracy and correlation values. This work effectively democratizes high-fidelity oscillation analysis, transitioning it from a specialist-centric task to a scalable, accessible operational resource.

Keywords: Machine and deep learning for system identification, Data-driven control theory, Learning methods for control
Abstract: Robustness certification for recurrent neural networks (RNNs) is important in safety-critical control settings. We propose a relaxation-based method that formulates RNN layer interactions as a semidefinite program (SDP) to compute certified upper bounds on the Lipschitz constant. Evaluations on a synthetic multi-tank system compare certified and empirical estimates, showing reasonably tight bounds even for long sequences, further improved by incorporating input constraints. The results also highlight the influence of initialization errors, relevant for frequently reinitialized models such as in model predictive control.

Keywords: Machine and deep learning for system identification, Data-driven control theory, Physics informed and grey box model identification
Abstract: Accurate prediction of the internal states of Lithium-Ion Batteries is critical for the safety and efficiency of Battery Management Systems (BMS), particularly for advanced indicators such as the State of Available Power (SOAP). While physics-based electrochemical models like the Doyle-Fuller-Newman (DFN) or Pseudo-Two-Dimensional (P2D) model offer superior fidelity compared to empirical Equivalent Circuit Models (ECMs), their high computational complexity typically precludes their use in real-time embedded applications, specially when high mesh resolution and multiple model calls within specialized algorithms are needed. This paper proposes a novel Physics-Informed Machine Learning (PIML) approach, specifically utilizing Physics-Informed Neural Networks (PINNs), to generate a Reduced Order Model (ROM) of the battery cell dynamics. By embedding physical equations directly into the neural network loss function, we derive a data-efficient ROM capable of capturing the electrochemical behavior defined by the DFN and Single Particle Model (SPM) framework. This ROM is subsequently integrated into a SOAP estimation algorithm. We demonstrate that the PINN-based approach outperforms classical numerical solvers in terms of execution speed while maintaining a high degree of physical consistency. The results suggest that physics-embedded machine learning provides a viable pathway for implementing next-generation, physics-based power prediction algorithms on standard BMS hardware.

Keywords: Machine and deep learning for system identification, Iterative and repetitive learning control, Learning methods for control
Abstract: We derive the coupled dynamics between generator and discriminator in continuous-time and discretized data space. We show that under fast-discriminator, adversarial learning and gradient-play yield a replicator dynamics for the generator, enabling Lyapunov stability analysis of the generator probability distribution. We derive local stability conditions for a reduced aggregate model and validate them through simulation using Electronic Health Records to synthesize realistic patient records.

Keywords: Machine and deep learning for system identification, Kalman filtering, Time series modeling
Abstract: Superconducting radio frequency cavities with a high quality factor enable energy-efficient accelerator operation but are very sensitive to mechanical disturbances that detune their resonance. Accurate detuning estimation is therefore essential for efficient resonance control and stable beam conditions. This paper introduces Kalman-Inspired Neural Decomposition (KIND), a data-driven estimator that fuses a Dynamic Mode Decomposition model for stationary modal behavior with a Transformer-based predictor for transient dynamics. KIND further outputs learned uncertainty signals that indicate regime changes, enabling anomaly detection. Using operational cavity data, we compare KIND with a classical Kalman filtering baseline and discuss its potential as a foundation for future uncertainty-aware, forecast-based control.

Keywords: Machine and deep learning for system identification, Learning methods for control, Statistical analysis
Abstract: Federated learning (FL) enables collaborative training of machine learning models without centralized access to raw data, but suffers from optimization instability and client drift under heterogeneous (non-IID) data distributions. These challenges are exacerbated for Generative Adversarial Networks (GANs), whose adversarial training dynamics are notoriously fragile even in centralized settings. In this work, we propose Roughness-Informed Federated Generative Adversarial Learning (RI-FedGAN), a new framework that stabilizes federated GAN training by explicitly measuring and controlling the local loss landscape roughness on each client.

We introduce the Roughness Index (RI) as a per-client diagnostic that captures the local variability of the adversarial loss along random one-dimensional projections. Building on this, we develop RI-FedGAN-Prox, a proximal variant of federated GAN training where both the generator and discriminator on each client are regularized toward the global model with a strength that is adaptively scaled by the client's RI. Clients exhibiting highly oscillatory or unstable adversarial dynamics are thus automatically constrained, while smoother clients retain more freedom to explore. We further define a Full RI-FedGAN variant that combines RI-scaled proximal updates with an RI-weighted aggregation scheme, down-weighting unstable clients in the global update.

Empirically, on MNIST and CIFAR-10 under various IID and non-IID partitioning schemes, RI-FedGAN-Prox and Full RI-FedGAN consistently improve classification score, Frechet Inception Distance (FID), Earth Mover's Distance (EMD), and robustness to mode collapse compared to standard FedGAN and fixed-proximal baselines. Our results demonstrate that roughness-aware regularization and aggregation is a promising ingredient for stabilizing generative modeling in federated environments.

Keywords: Machine and deep learning for system identification, Time/parameter varying system identification
Abstract: The identification and modeling of time-varying systems is a fundamental challenge in signal processing and system identification. To address this challenge, we propose a class of time-varying state-space model (SSM) based neural networks in which the neurons' states are governed by time-varying dynamics. The proposed model provides the learnable time-varying dynamics through a dictionary of basis functions, where each basis function evolves differently over time. We evaluate the proposed approach on both synthetic data from switching systems and a speech denoising task where real audio is corrupted with switching dynamics noise. The results show that the proposed time-varying model consistently outperforms its time-invariant counterparts while maintaining comparable computational complexity. Our investigations also reveal which aspects of the time-varying dynamics of the data most need to be captured by the proposed time-invariant models, how the additional freedom provided by time-varying basis functions should be allocated across model components, and to what extent larger models can compensate for time-invariant limitations.

Keywords: Event-based control, Nonlinear adaptive control
Abstract: This paper proposes a dynamic encoding-decoding event-triggered control for a class of strict-feedback nonlinear systems. A dynamic encoding-decoding event-triggered scheme is designed from the viewpoint of signal encoding-decoding process. It makes only an L-length codeword be transmitted for each communication between the control and actuator boxes, where L can be user-specified according to communication bandwidth. At the same time, our proposed event-triggered scheme allows the triggering threshold to be dynamically updated based on control input signal itself rather than relying on Lyapunov stability requirement. The resulted dynamic encoding-decoding event-triggered control can guarantee asymptotic tracking performance without Zeno behavior. Finally, the effectiveness of the proposed control scheme is illustrated by simulation results.

Keywords: Estimation and filtering, Distributed control and estimation, Kalman filtering
Abstract: Joint localization and target tracking are fundamental capabilities for multi-robot systems. Existing distributed approaches either neglect the cross-covariances between robots and targets, which leads to degraded estimation accuracy and increase of computational cost, or require transmitting full covariances, which incurs prohibitive communication costs. To address these challenges, we propose a distributed joint localization and tracking framework in which each robot maintains the local cross-covariance between its own state and the target states, while neighboring robot pairs disseminates only the marginal mean-covariance pairs for the robot and each target separately, deliberately omitting the cross-covariances. This design fully exploits the inherent correlations, enabling locally optimal estimation and eliminating the computational overhead arising from optimizing unknown cross-covariances. Additionally, owning to the selective transmitting mechanism, it achieves an efficient balance between accuracy and communication efficiency. Moreover, a dimensionally adaptive Inverse Covariance Intersection (ICI)-based approach is developed to fuse non-common state estimates and nonlinear measurements, achieving consistent and less conservative results. Extensive simulations demonstrate that it outperforms existing distributed methods in terms of both accuracy and efficiency.

Keywords: Adaptive control of multi-agent systems
Abstract: This paper studies the leader-follower formation control problem where followers can only obtain noisy bearing measurements of their local neighbors. We first develop a distributed localization algorithm to estimate the relative position with the leader, where a modified stochastic gradient (SG) algorithm with a compensation term is used. Based on the estimates, we design an adaptive control law to drive all agents to the desired formation where a decaying excitation signal is introduced to solve the issue caused by the excitation conditions required by the localization algorithm and the stability of formation control. We show that under certain non-persistent excitation (PE) conditions, the convergence of the distributed localization algorithm can be guaranteed. Furthermore, by verifying that the control laws satisfy these excitation conditions, we prove that the proposed control algorithm enables all agents to achieve the desired formation. The effectiveness of the theoretical results is validated by numerical simulations.

Keywords: Multi-agent systems, Control over networks, Markov decision process
Abstract: In this work we study how heterogeneous banks strategically operate interbank lending networks as a dynamic game. We model how links emerge from liquidity needs and evolving trust among borrowers, lenders, and intermediaries. Our main contributions are to use entropy-regularized constrained Markov perfect equilibrium (ECMPE) to analyze the steady-state behavior of such network and to give sufficient conditions for its existence. As a necessary step, we first integrate network formation, liquidity dynamics, and trust evolution into a unified dynamic-game model. Finally, we develop an efficient iterative algorithm for computing the ECMPE numerically and demonstrate its convergence in simulation.

Keywords: Big data and machine learning applied to smart cities, IoT for cities
Abstract: Presently, information intensive cities are confronted with challenges related to traffic congestion, urban design, increased polluting emissions, and a lack of information systems that facilitate management to conventional traffic monitoring systems. To address this problem, this article proposes the implementation of a modular urban traffic digital twin (DT) focused on urban mobility environment monitoring and simulation. This work presents a structured methodology for digital twin platform development, constructed by perception layer, data management phase, simulation function, and tridimensional visualization. The proposed platform integrates real-time and real world data acquisition, enabling the synchronization of physical and virtual environment, thus providing a modern instrument for the analysis of urban traffic data, thereby facilitating decision-making and the implementation of strategies to enhance urban issues and develop traffic management solutions. The results demonstrate the applicability of the proposed development methodology by validating the platform under a real-time connectivity environment, by measuring data transmission latency and platform functional evaluation. The focus of a modular design establishes a scalable foundation for advanced urban traffic management and data-driven decision-making in smart city applications.

Keywords: Distributed optimization and control for smart cities, Transportation networks, Interconnected city networks
Abstract: In this article, we present a long-duration autonomy approach for the control of connected and automated vehicles (CAVs) operating in a transportation network. In particular, we focus on the performance of CAVs at traffic bottlenecks, including roundabouts, merging roadways, and intersections. We take a principled approach based on optimal control, and derive a reactive controller with guarantees on safety, performance, and energy efficiency. We guarantee safety through high order control barrier functions (HOCBFs), which we ``lift'' to first order CBFs using time-optimal motion primitives. This yields a set of first-order CBFs that are compatible with the control bounds. We demonstrate the performance of our approach in simulation and compare it to an optimal control-based approach.

Keywords: Consensus, Distributed optimization, Multi-agent systems
Abstract: This paper addresses the challenge of solving distributed nonconvex nonsmooth optimization (DNNO) problems. While most existing works focused on DNNO with composite structures, only a limited number of studies tackled general DNNO problems. However, these studies primarily provided asymptotic convergence analysis for their proposed algorithms. In this paper, we advance the field by presenting a non-asymptotic convergence analysis for solving general DNNO problems. To measure the convergence performance of algorithms in DNNO, we introduce the concept of Goldstein stationarity. Building on this, we propose a distributed zeroth-order algorithm tailored for DNNO and establish an mathcal{O}(d^frac{3}{8}T^{-frac{1}{4}}) convergence rate for achieving a Goldstein stationary point. Finally, we validate the efficacy of the proposed algorithm through numerical experiments.

Keywords: Distributed optimization, Distributed control and estimation, Multi-agent systems
Abstract: This paper studies online stochastic nonconvex optimization in multi-agent networks under some constraint sets. In this setting, each agent makes decisions from a feasible set using only locally available stochastic gradients from previous steps and information exchanged with its neighbors. This paper develops a projection-free distributed optimization algorithm, which replaces costly projection steps with efficient conditional gradient updates. To further mitigate the effects of stochastic gradient noise, the algorithm integrates recursive variance-reduced gradient estimators into its update process. Theoretically, this paper establishes that the proposed algorithm achieves a high-probability sublinear regret bound for online stochastic nonconvex optimization problems. Numerical experiments are conducted to demonstrate the efficiency of the proposed algorithm.

Keywords: Multi-agent systems, Control of networks, Distributed optimization
Abstract: A budget-balancing incentive mechanism for pseudo-gradient-based noncooperative dynamical systems is developed for achieving a largest admissible social welfare with unincentivizable agents. In the proposed approach, the system manager transfers utilities only among incentivizable agents and tries to move the Nash equilibrium of the entire game to maximize the social welfare function as much as possible. With and without the explicit knowledge of Nash equilibrium mapping, we present a primal method and a dual method to update the incentive functions. Sufficient conditions are derived under which agents’ state converges to the optimal target state as a Nash equilibrium. A couple of numerical examples are presented to illustrate the efficacy of our results.

Keywords: Model predictive control of hybrid systems
Abstract: This paper investigates the feedback optimization of linear time-invariant systems with unknown additive disturbances and state/input constraints. We propose a controller that combines a feedback optimization algorithm with a tracking model predictive control (MPC) algorithm. The feedback optimization algorithm generates a reference input signal based on real-time gradient measurements of the economic objective at the plant’s real-time output and control input, eliminating the need for the analytical form of the gradient function. The tracking MPC algorithm drives the actual system input to follow this reference signal while enforcing state and input constraints. The control input is obtained by solving a quadratic programming (QP) problem. We establish, under mild assumptions, recursive feasibility of the MPC problem, satisfaction of all state and input constraints, and input-to-state stability with respect to disturbance changes. If the disturbance change converges to zero, the state and input converge to their optimal trajectories. An economic optimization case study is used to demonstrate the method’s effectiveness.

Keywords: Modeling and simulation of transportation systems, Automatic control, optimization, real-time operations in transportation, Vehicle dynamic systems
Abstract: This paper studies the optimization of stochastic one-way car-sharing networks, where the network topology and station roles are governed by a Markov process. To regulate the network flow, we introduce a pricing-based car-sharing model on a directed network. The model is formulated as a positive linear system, and system performance indices are used for mathematically rigorous congestion-oriented design. To address the nonconvexity and computational complexity of the resulting optimization problems, we develop a DC (difference-of-convex) programming framework. By exploiting the log--log convexity of posynomial functions, the design problems are reformulated as DC programs. Simulation results demonstrate the effectiveness of the proposed method.

Keywords: Modeling and simulation of transportation systems, Electric and solar vehicles, Automatic control, optimization, real-time operations in transportation
Abstract: Accurate estimation of the State of Charge (SoC) is essential for the reliable and safe operation of lithium-ion batteries in Electric Vehicles (EVs). Conventional SoC estimation methods such as open-circuit voltage method and Coulomb counting have limitations including the accumulation of arbitrary errors over time and susceptibility to dynamic operating conditions. Additionally, battery systems exhibit highly non-linear characteristics. To overcome these limitations and handle the non-linearity, model-based techniques have been increasingly adopted in recent years. Hence, identification of accurate battery model parameters is very essential. The main objective of this work is to select suitable optimization algorithm for identifying the equivalent circuit model(ECM) parameters of 100Ah lithium-ion battery. Initially, a Hybrid Pulse Power Characterization (HPPC) test is conducted to identify the initial parameters of a Two RC-network Equivalent Circuit Model (ECM) for the 100Ah prismatic cell under discharging condition. The experiments are carried out on a fully charged battery. For HPPC test, discharging from 100% SoC to 0% SoC with step changes of 10%, 0.5C constant discharging and dynamic discharging with variable steps while recording the battery voltage. The ECM parameters are identified for all the 10 regions of battery voltages obtained through HPPC. Further, the ECM parameters, are optimized using Simplex, Nonlinear Least Squares (NLS), and Pattern Search (PS) algorithms. Finally, the performances of the optimization algorithms are compared by estimating battery voltage with optimized parameters against 0.5C constant discharging and dynamic discharging conditions.

Keywords: Design methods for data-based control, Applications of optimal control
Abstract: This article presents the adaptation of the Virtual Reference Feedback Tuning (VRFT) for a control system with multiple parallel branches, since this data-driven method was originally conceived for single-branch structures. We present two possible approaches - the joint matrix tuning and the sequential tuning - and apply them to tune the controllers of a virtual power plant (VPP) composed of four photovoltaic plants, focusing on the regulation of the system’s active power. Simulated VPP results show that the obtained responses closely reproduce the specified desired behavior.

Keywords: Design methods for data-based control, Decentralized control, Adaptive control design
Abstract: A performance-oriented decoupling learning control method is developed for multi-variable systems subject to unknown disturbances, which only uses the system outputs. The key is to devise a tailored residual neural network (TRNN)-based equivalent-input-disturbance (EID) estimator to handle the overall influence of coupling and unknown disturbances and then to design the resultant system. First, for a performance-oriented learning, an intermediate index of decoupling control is adopted to guide the backpropagation training of the feedforward neural network of the TRNN. Then, multiple simple residual terms of the TRNN are designed for the sake of closed-loop stability. Further, the resultant decoupled system is designed. Last, a case study of a pilot distillation column validates the superior decoupling performance over other methods.

Keywords: Design methods for data-based control, Linear systems, Controller constraints and structure
Abstract: This paper investigates the static output feedback (SOF) stabilization problem for discrete-time linear time-invariant (DLTI) systems using finite noise-free input-output data only. Within the behavioral framework, systems are described by input-output representations. A necessary and sufficient Lyapunov-based condition for SOF stabilization is first established in the model-based setting. An exact identification method is then proposed to reconstruct a minimal lag input-output model via the solution of a single matrix equation. Building on these results, a cone complementarity linearization (CCL)-based linear matrix inequality (LMI) algorithm is developed for data-driven SOF stabilization. The proposed method naturally accommodates structural constraints on the SOF gain, and its effectiveness is illustrated by numerical examples.

Keywords: Design methods for data-based control, Optimal control theory, Stability of nonlinear systems
Abstract: In this paper, we combine a data-driven system representation with a framework to systematically construct (approximate) solutions to nonlinear optimal control problems. By immersing the unknown dynamics into an extended state space, solutions are characterised via purely data-dependent algebraic conditions. This allows us to design dynamic state-feedback controllers with local stability and performance guarantees for unknown nonlinear, input-affine systems directly using data, without explicitly identifying the dynamics.

Keywords: Design methods for data-based control, Positive linear systems
Abstract: We formulate a data-driven stabilization problem for discrete-time linear systems in the framework of linear programming (LP). It is known that if a polyhedral cone is invariant, referred to as K-positivity, then stability analysis reduces to solving an LP. Building on this fact, we establish a necessary and sufficient condition for the existence of a state feedback gain imposing both closed-loop K-positivity and stability, given a proper polyhedral cone. We also propose a data-driven design method of such a state feedback gain.

Keywords: Design methods for data-based control, Switching linear systems, Lyapunov methods
Abstract: This paper proposes a data-driven control framework for discrete-time periodic piecewise linear systems with unstable subsystems. Unlike conventional approaches that strictly require the stabilization of every subsystem, our framework allows some subsystems to remain unstable under the action of the controller. An equivalent data representation is established for the closed-loop periodic piecewise linear systems, with a corresponding data collection strategy. Sufficient conditions ensuring exponential stability are developed through analysis using piecewise Lyapunov functions. To address the computational challenges caused by the resulting non-convex bilinear matrix inequalities and to enforce physical constraints on the controller gain, an iterative algorithm is developed to solve these conditions. These theoretical results are formulated as solvable linear matrix inequalities. The effectiveness and engineering applicability of the proposed approach are verified through numerical simulations and a practical case study involving a DC-DC Boost converter. The results validate that the proposed data-driven framework can achieve exponential stability for general periodic systems with unstable subsystems.

Keywords: Observer design, Distributed nonlinear control
Abstract: This paper addresses the state estimation problem for a class of interconnected nonlinear multi-agent systems (MAS). The block--triangular structure inherent to platoons results in highly coupled complex error dynamics. These coupled dynamics are difficult to analyze directly, presenting a significant challenge for observer design. The innovation of this work lies in the quadratic form of the decomposed error that enables the transformation of the coupled quadratic forms from the Lyapunov analysis into a tractable convex optimization problem, that can be solved by a set of Linear Matrix Inequalities (LMIs). The obtained result provides a constructive design of a distributed high--gain observer that ensures exponential convergence while providing lower bounds on all observer gains. The effectiveness of the proposed design is validated through simulations on a vehicle platoon model.

Keywords: Output feedback nonlinear control, Robust estimation, Stability of nonlinear systems
Abstract: This paper presents an observer-based feedback control strategy for nonlinear coupled vehicle dynamics affected by state-dependent measurement uncertainties. The proposed observer structure accounts for the nonlinear distortions introduced by imperfect sensors, while the control law ensures both stability and accurate trajectory tracking of the vehicle. The stability analysis relies on the convexity principle and the mean value theorem, enabling the transformation of the nonlinear terms into linear representations through Jacobian matrices evaluated at specific operating points. The stability of the resulting augmented system is established via a set of Linear Matrix Inequalities~(LMIs), derived under Lipschitz continuity and convex polytopic descriptions of the Jacobians. The proposed framework guarantees asymptotic stability and provides a systematic procedure for designing both the observer and the controller gains. Simulation results on a nonlinear single-track vehicle model demonstrate the effectiveness of the approach in reconstructing unmeasured states and maintaining robust stability despite state-dependent sensor uncertainties.

Keywords: Robust estimation, Nonlinear observers and filters, Application of nonlinear analysis and design
Abstract: In this paper, the state of a class of Lipschitz nonlinear systems with an observable linear part is estimated using a modulating functions-based recursive finite-memory Volterra state estimator. To this end, the linear part of the system is first transformed into an observable canonical form. Then, by applying the generalized modulating functions method, the system state is implicitly characterized by a set of nonlinear Volterra integral equations of the second kind over a sliding time window. This formulation naturally attenuates measurement noise through integral smoothing. For discrete-time noisy output measurements, a numerical quadrature scheme is used to approximate the integral equations, leading to a recursive state reconstruction algorithm that does not require knowledge of the initial conditions. Then, the estimation errors are analyzed. In particular, error bounds are derived for the recursive estimation errors. Finally, a simulation example demonstrates that the proposed method provides robust and non-asymptotic estimation of both the state and disturbances.

Keywords: Observer design
Abstract: The weighted Kalman filter (WKF) is a recently introduced variant of the extended Kalman filter that replaces the standard Jacobian-based linearization with Gaussian-weighted integration. Until now, this reliance on integration has represented a major obstacle to the applicability of the WKF to general nonlinear systems. This paper shows that recent developments in probability theory, specifically those concerning the analytical evaluation of expectations involving nonlinear functions of Gaussian random vectors and higher-order moments, open new avenues for deriving a fully analytical and computationally efficient formulation of the WKF. The proposed approach eliminates the need for multidimensional numerical integration, thereby substantially reducing the computational complexity of the WKF. Simulation results obtained for a nonlinear quadratic system and for a nonlinear system with non-polynomial nonlinearities illustrate the effectiveness and applicability of the proposed approach.

Keywords: Observer design, Nonlinear observers and filters, Nonlinearity learning from data
Abstract: This paper proposes a hybrid estimator for vehicle systems with unmodeled dynamics. A generalized unknown input observer provides bounded physics-based estimates when classical rank conditions are relaxed through output derivatives. These estimates supervise a neural adaptive observer that learns a state-dependent approximation of the unknown input. LMI conditions certify mathcal{H}^{1} and mathcal{L}_{2} estimation bounds, and vehicle simulations show improved trajectory tracking and tire-force residual reconstruction.

Keywords: Adaptive control design, Observer design, Application of nonlinear analysis and design
Abstract: A common assumption when applying the Kalman filter is a priori knowledge of the system parameters. These parameters are not necessarily known, and this may limit the real-world applicability of the Kalman filter. The well-established Model Reference Adaptive Controller (MRAC) utilizes a known reference model and ensures that the input-output behavior of a potentially unknown system converges to that of the reference model. We present KalMRACO, a unification of Kalman filtering and MRAC leveraging the reference model of MRAC as the Kalman filter system model, thus eliminating, to a large degree, the need for knowledge of the underlying system parameters in the application of the Kalman filter. We also introduce the concept of blending estimated states and measurements in the feedback law to ensure stability during the initial transient. KalMRACO is validated through simulations and lab trials on an underwater vehicle. Results show superior tracking of the reference model state, observer state convergence, and noise mitigation properties.

Keywords: System identification and adaptive control of distributed parameter systems
Abstract: This paper investigates how to distinguish ensemble systems from their collective behavior using statistical methods in reproducing kernel Hilbert spaces (RKHS). We develop a nonparametric framework for comparing ensemble systems by computing the maximum mean discrepancy (MMD) between their aggregated measurements, without requiring any prior knowledge of the underlying dynamics. The framework naturally extends to clustering multiple unknown ensembles solely from their aggregated observations. Numerical experiments demonstrate the reliability and robustness of the proposed approach across a variety of ensemble models with distinct dynamical structures.

Keywords: Distributed parameters port Hamiltonian systems, Lagrangian and Hamiltonian systems, Interconnected nonlinear systems
Abstract: A nonlinear continuous model for vocal fold dynamics based on the port-Hamiltonian systems (PHS) framework is presented. The model couples two-dimensional elastodynamics with a diffusion equation for fluid concentration, where local tissue deformation drives changes in concentration, producing volumetric swelling. This formulation enables the simulation of diffusion-driven tissue hydration and inflammation, and is further coupled to aerodynamic loading via a one-dimensional Bernoulli pressure model, while mechanical contact between the folds is incorporated through a nonlinear damping-based penalty method. A contact-induced swelling criterion, based on dissipated power, is tested to evaluate the onset and progression of inflammation under sustained phonation. Numerical experiments illustrate the capability of the model to capture the complex interplay between tissue mechanics, airflow-induced forces, contact phenomena, and diffusion-driven swelling, providing an energy-based framework for studying pathological vocal fold behaviors.

Keywords: Distributed parameters port Hamiltonian systems, Boundary control of distributed parameter systems, Lagrangian and Hamiltonian systems
Abstract: This work extends previous 1D irreversible port-Hamiltonian system (IPHS) formulations to boundary-controlled ND distributed parameter systems describing conduction–diffusion fluid phenomena. Within a unified and thermodynamically consistent framework, we show that conduction and diffusion can be represented through a single coherent structure that preserves global energy balance and ensures a correct characterization of entropy production. The resulting formulation provides a foundation for the systematic modeling and control of complex multi-physical processes governed by coupled transport mechanisms in N-dimensions. In the longer term, this framework opens the door to structure-preserving numerical schemes capable of enforcing thermodynamic principles directly at the discretized level.

Keywords: Distributed parameters port Hamiltonian systems, Boundary control of distributed parameter systems, Lagrangian and Hamiltonian systems
Abstract: The Irreversible Port-Hamiltonian Systems (IPHS) framework is extended to the modelling of non-isentropic fluids with viscous dissipation in the Eulerian description. Building on earlier IPHS formulations for diffusion-driven and non-convective distributed systems, it is shown that convective transport can be consistently encompassed by the framework by modifying the underlying differential operators. After revisiting the constitutive relations of non-isentropic fluids in both Eulerian and Lagrangian coordinates, it is demonstrate how these systems fit within an extended IPHS formulation. Furthermore, an extended parametrisation of the boundary port variables which ensures that the first and second laws of Thermodynamics are fulfilled allows to define a general class of boundary controlled IPHS.

Keywords: Distributed parameters port Hamiltonian systems, Lagrangian and Hamiltonian systems, Lyapunov methods
Abstract: Rayleigh beam models are of significant practical interest for the simulation and control of flexible structures due to their balance of accuracy and complexity. Unlike previous approaches based on kinematic constraints, this work derives the Rayleigh beam by degenerating the shear elastic energy of a Timoshenko beam model. This results in an equivalent reduced-state descriptor port-Hamiltonian system on a Stokes–Lagrange structure. Following a different approach, a second non-descriptor Stokes–Lagrange representation is presented, more closely related to an Euler–Bernoulli formulation with differential constitutive relations. Finally, a mixed finite element discretization is proposed that mirrors the degeneracy mechanism to transform a non-descriptor infinite-dimensional Timoshenko beam model into a descriptor Rayleigh beam approximation by driving the shear strain interpolation functions to zero.

Keywords: Nonlinear time-delay systems, Lyapunov methods, Stability of nonlinear systems
Abstract: For mechanical systems in the Lienard canonical form of the model with constant delays, the problem of robust stability analysis is considered. It is assumed that all nonlinearities are homogeneous functions of different degrees. For this class of systems, stability conditions are obtained, first, for a delay-free model, and next, they are developed to the delayed case. To this end, different Lyapunov--Krasovskii functionals are constructed. Applying the averaging theory, the influence of time-varying perturbations is taken into account. The results are illustrated by simulations for a mechanical system.

Keywords: Linear time-delay systems, Robust controller synthesis, Control of complex systems
Abstract: We consider in this paper a class of single-input single-output delay systems of neutral type with transfer functions with one delay inducing a chain a poles clustering the imaginary axis from left of right. For a large class of subsystems, we provide coprime factorizations over H_infty which guarantee H_infty-stabilization properties and are a first step towards their robust H_infty-stabilization.

Keywords: Nonlinear time-delay systems, Stability of nonlinear systems
Abstract: We study the logistic equation when its two terms each feature a delay. The two delays, sigma and tau, are the only parameters of this nonlinear delay differential equation, and we show that its bifurcation diagram in the (tau, sigma)-plane features regions with complicated dynamics, including transitions to chaotic attractors. This system arose, after rescaling, as a conceptual model for the Atlantic Meridional Overturning Circulation. However, it can also be seen as a generalization of the well-known Hutchinson-Wright equation, and we show how all complicated dynamics `vanishes' as tau is decreased to zero.

Keywords: Nonlinear time-delay systems, Adaptive control design
Abstract: This paper investigates the control problem of a class of nonlinear systems subject to unknown time-varying delays, input saturation, and external disturbances. To effectively handle the challenges introduced by the unknown time delay, a novel Lyapunov-Krasovskii functional (LKF) is constructed. A set of static gain functions is developed to compensate for the delayed states, while a auxiliary subsystem is employed to address the input saturation. By integrating these techniques within a backstepping framework, n static gain functions are systematically designed to guarantee the stability of the closed-loop system. It is proven that all closed-loop signals remain bounded and ultimately converge to a small neighborhood around the origin, while the output tracking error approaches zero. Finally, simulation results are provided to demonstrate the effectiveness and robustness of the proposed control scheme.

Keywords: Nonlinear time-delay systems, Optimal control theory, Applications of optimal control
Abstract: This paper develops a delay differential equation model for Thailand’s disability labor market and examines the effect of delayed policy implementation. A fixed delay is used as an aggregate representation of the implementation lag in employment support. Based on this model, an optimal control problem is formulated in which a time-dependent intervention is used to reduce unemployment while balancing policy cost. The model is calibrated using Thai disability labor market data from 2022--2025, and numerical simulations are performed to illustrate the resulting dynamics. The results suggest that, under the calibrated parameter set, a front-loaded intervention strategy can reduce unemployment more effectively than a static policy over the planning horizon. These findings provide a quantitative perspective on delayed disability employment policy and may support further discussion of policy design in Thailand.

Keywords: Linear time-delay systems, Robust time-delay systems
Abstract: In the analysis of dynamical systems, it is a common and convenient assumption that infinitesimally small noise has a negligible effect on deterministic behaviour. However, the present work highlights that for systems with slow dynamics - slowly oscillating between stable and unstable states - the application of a purely deterministic approach is fundamentally flawed. Our research explains the sharp contradiction between idealised analytical calculations and experimental or professional numerical simulation results. Any numerical integration method, even purely due to finite machine number representation introduces a minimal amount of noise into the system. When a system with slow parameter variation enters an unstable quasi-static domain, this infinitesimally small noise acts as an equivalent noise intensity that is exponentially magnified within a finite time. This mechanism leads to a premature loss of finite-time stability (FTS), generating vibrations that prevent the attainment of the apparently stable deterministic state. We demonstrate that if the exponential magnification of noise leads to system failure even in state-of-the-art numerical simulations, then under the unavoidable environmental noise of real physical measurements (e.g., 0.1% noise in cutting forces), deterministic stability is an absolute illusion. To address this, we investigate the temporal evolution of the second moment (covariance) using matrix-free direct simulations, providing a framework for handling the hidden instability and stochastic resonance of slow time-delayed systems.

Keywords: Sustainable and circular manufacturing systems, Sustainable and circular supply chain and production
Abstract: Life Cycle Assessment (LCA) is essential for environmental analysis, but its application is hampered by inconsistency, subjectivity, and poor traceability. Current LCA tools lack measurable links between configuration choices (approaches, methods, databases, and tools) and performance outcomes (accuracy, reliability). This study introduces an attribute-based modeling framework to enhance LCA outcomes. Drawing on seismic attribute analysis, the framework decomposes LCA elements into a component–attribute–implication hierarchy. It systematically investigates the causes of inaccuracy in LCA results, linking the selection of specific LCA elements to challenges in reliability. The framework defines measurable attributes for core LCA elements and maps their causal relationships to quantifiable improvement implications. A conceptual and measurable attribute–implication matrix connects configuration decisions with expected performance outcomes. This approach establishes a reasoning basis for LCA configuration, improving traceability and reducing subjectivity. It lays the groundwork for integrating LCA with decision-support and knowledge-based systems, thus contributing to the systematic advancement of LCA practice.

Keywords: Sustainable and circular manufacturing systems, Sustainable and circular supply chain and production
Abstract: Life Cycle Assessment (LCA) and Product Carbon Footprint (PCF) provide essential insights for environmental decision-making, yet their reliability and accuracy remain challenged by inconsistencies in methods, data sources, and digital integration. This paper presents a decision-support conceptual framework designed to enhance the accuracy and reliability of LCA/PCF studies. The framework is derived from a systematic literature review and industrial watch, identifying critical bottlenecks across methods, tools, and data management practices. It integrates three layers as business applications, decision support, and knowledge base within three core modules— (1) a smart connector for data acquisition and integration (interoperability module), (2) an inference engine, (3) AI-based learning for predictive reasoning (2 and 3 form decision-aid module) for structured information management. Together, it creates an interoperable environment bridging data-driven and knowledge-driven approaches. The framework serves as a foundation for the next generation of digital and intelligent LCA systems, promoting traceability, consistency, and informed sustainability decisions.

Keywords: Sustainable and circular manufacturing systems, Data-driven and AI-based modelling of production and logistics
Abstract: The growing volume of electronic waste has emphasized the urgency of developing design strategies that enable efficient end-of-life (EoL) recovery. Printed Circuit Boards (PCBs), as a central part of most electronic devices, present a particular challenge due to their structural complexity, heterogeneity, and reliance on permanent joining methods, which complicate disassembly. To address this challenge, this paper starts by presenting how to close the loop between PCB design and disassembly and presents an ontology structure aimed at supporting PCB design for disassembly, by semantically linking design decisions with disassembly requirements. The ontology formalizes knowledge around three core dimensions: i) product structure, including PCB components, materials, and connection types; ii) disassembly process knowledge, such as required tools and disassembly metrics; and iii) design implications to provide eco-design feedback to support circular design strategies. By providing a structured, machine-interpretable representation of disassembly knowledge, the proposed ontology aims to close the loop between PCB design and EoL strategies and demonstrates the potential of semantic technologies to enhance circularity in electronics.

Keywords: Sustainable and circular manufacturing systems
Abstract: Manufacturers are increasingly required to adopt circular strategies (CS) to reduce resource consumption and environmental impacts. Effective decision-making at a product’s End-of-Usage/End-of-Life, however, depends on the availability of reliable lifecycle information and adequate decision-support systems, both of which are often missing. To address these challenges, the European Union is introducing the Digital Product Passport (DPP) as part of the Ecodesign for Sustainable Products Regulation, aiming to provide comprehensive product information across value chains, like its origin, materials, environmental impact, and disposal recommendations. Recent studies highlight the need for dynamic DPPs capable of integrating evolving product data along the lifecycle. However, current DPP frameworks seem mainly static, remain mostly conceptual and offer limited guidance for supporting circularity decisions. In parallel, Digital Twins (DTs) are increasingly recognized as promising enablers for enriching DPPs with accurate, real-time lifecycle data; however, the combined use of DT and DPP remains an emerging and insufficiently explored research domain. This paper investigates the state of the art on the possible integrated use of DTs and DPPs to support dynamic, data-driven decision-making towards circularity objectives. A research agenda is proposed to guide future research and developments in this field.

Keywords: Cyber-physical production systems, Intelligent manufacturing systems, Advanced manufacturing and remanufacturing technologies
Abstract: Manufacturing systems are increasingly forced to reconcile productivity, resilience, and sustainability in the face of growing environmental and social constraints. Circular manufacturing has therefore emerged as a strategic paradigm for the transition from linear production models to closed-loop lifecycle management. In this context, digital technologies associated with Industry 4.0 play a central role, enabling transparency, optimisation, and informed decision-making throughout the lifecycle of industrial equipment. This paper presents a defined framework to support the assessment of sustainable circular manufacturing, validated through four industrial pilot projects in the metal, stone, plastic, and food sectors by the AIDEAS project. The proposed approach is based on the ESIA (Environmental and Social Impact Assessment) framework, designed to identify, predict, and mitigate the negative effects of a project on the environment and local communities. This paper presents the ESTEIA framework, which defines how to conduct a structured impact assessment based on key performance indicators, analysing environmental, economic, social, and technological effects. Finally, a case study from the company D2Tech is presented to demonstrate how ESTEIA was applied in the factory and to assess its impact.

Keywords: Complex dynamic systems
Abstract: This paper investigates distributed adaptive fault-tolerant containment control for multi--quadrotor unmanned aerial vehicle (QUAV) formations subject to unknown nonlinearities, partial actuator loss, and input bias. To address these challenges, a neural network (NN)--based control scheme is proposed. NNs are employed to approximate the unknown dynamics online, while adaptive update laws are designed to compensate for the actuator faults. A distributed control protocol is formulated by integrating the NN approximation with relative state information from neighboring vehicles. The proposed controller guarantees that follower QUAVs converge to the convex hull spanned by the leaders, ensuring all signals in the closed loop system remain bounded. Numerical simulations validate the effectiveness of the proposed strategy.

Keywords: Complex dynamic systems
Abstract: 本文在混合索引模型框架下研究了冲动逻辑动力系统（ILDS）的输出跟踪问题。首先，揭示当跳跃转移矩阵为幂零矩阵时，ILDS是前向完备的。其次，通过构建一组可达集合，建立了ILDS输出追踪的必要且充分条件。最后，通过采用完整的可达集族，提出了一种合适的方法来设计相应的状态反馈控制，使ILDS能够实现输出跟踪。

Keywords: Complex dynamic systems
Abstract: This paper investigates whether similarity-based interactions are necessary for coordinators to guide opinion consensus in the Deffuant-Weisbuch (DW) model. We introduce two complementary types of coordinators: an inclusive coordinator (IC), which interacts with agents based on opinion similarity, and a strategic coordinator (SC), which interacts based on opinion dissimilarity. For both coordinators, we derive a sufficient condition under which their introduction guarantees that consensus becomes the unique equilibrium of the DW model. Extensive simulations reveal distinct operational regimes for the two types of coordinators. When the DW model rarely reaches consensus by itself, the SC is more effective in accelerating convergence, indicating that similarity-based interactions are not necessary. In contrast, in high-consensus regimes, the IC achieves faster consensus.

Keywords: Complex dynamic systems, Decentralized and distributed control for large-scale systems
Abstract: This paper investigates the distributed cooperative moving path-following (MPF) problem for underactuated autonomous surface vehicles (ASVs) with uncertainties. The proposed cooperative guidance strategy comprises guidance signals, path variable updating laws, and virtual control inputs. Specifically, the guidance signals and path variable updating laws are devised by leveraging moving guidance vector fields to achieve cooperative path-following performance. The virtual control inputs are designed for the accurate tracking of guidance signals in the presence of uncertainties, which include internal modeling errors and external disturbances. Rigorous analysis guarantees the convergence of the cooperative MPF errors for the multi-ASV system from any initial states. Finally, numerical simulations are presented to validate the effectiveness of the proposed controller.

Keywords: Complex dynamic systems, Decentralized and distributed control for large-scale systems, Interconnected dynamical systems
Abstract: This paper studies resilient consensus of multi-agent systems in the presence of dishonest agents. A trust-based distributed weight-adjustment mechanism is proposed, in which each normal agent evaluates its in-neighbors and updates the weights associated with them, using a penalty parameter and a regulation factor. Under mild connectivity conditions, convergence of the normal agents to a consensus value is established. Unlike state-of-the-art approaches, the proposed approach imposes less stringent requirements on network connectivity and relies solely on one-hop neighbor information. Simulation results demonstrate the advantages of the proposed method over MSR-type algorithms under certain attack scenarios.

Keywords: Complex dynamic systems, Decentralized and distributed control for large-scale systems, Interconnected dynamical systems
Abstract: This paper investigates the problem of cooperative predefined-time target surrounding for multiple autonomous surface vehicles (ASVs). A predefined-time control framework is developed to ensure that all the error dynamics of ASVs converge to zero within a rigorously specified time, independent of the initial conditions. The control design follows a multi-layer structure, where predefined-time kinematic controllers are first constructed, and then combined with a predefined-time kinetic controller to guarantee closed-loop convergence performance. The proposed scheme enables the ASVs to establish a coordinated and stable surrounding formation around a moving target. Simulation results validate the effectiveness of the proposed control scheme.

Keywords: Monitoring, performance assessment, and fault detection in chemical process control, Process performance monitoring/statistical process control, Machine learning and artificial intelligence in chemical process control
Abstract: Operating data frequently exhibit significant mode drift due to condition switching and load disturbances. As a result, conventional monitoring models built on static assumptions are prone to catastrophic forgetting and limited knowledge transfer. To overcome these challenges, we propose a row-weighted replay lifelong dictionary learning method (RrLDL) for multimodal process monitoring. RrLDL incorporates a variable contribution driven selective memory mechanism to mitigate feature-level steady-state forgetting, where a row-weighted constraint reinforces the discriminative power of individual process variables during knowledge retention and incremental learning. Furthermore, a compact replay strategy is adopted, prioritizing samples dominated by high-contribution variables to retain critical mode features. Numerical simulation and Tennessee Eastman benchmark experiments demonstrate that RrLDL achieves superior multimodal anomaly monitoring accuracy and enhanced cross-mode generalization compared with representative baseline methods.

Keywords: AI methods for FDI/FTC, Monitoring, performance assessment, and fault detection in chemical process control, Process performance monitoring/statistical process control
Abstract: Industrial alarm systems often suffer from excessive alarms, which overwhelm operators and hinder timely response. Existing studies mainly focus on alarm system design and alarm event analysis, but rarely leverage auxiliary knowledge such as operational manuals to support operators’ decision-making during alarm handling. Motivated by such an issue, this paper proposes a dual-source hybrid Retrieval-Augmented Generation (RAG) approach for industrial alarm knowledge parsing and operational decision support. The contributions are twofold: First, a dual-source heterogeneous encoding method is proposed to convert Alarm & Event (A&E) data into sparse vectors and operational manuals into dense embeddings. Second, a hybrid retrieval strategy is designed by integrating sparse lexical matching with dense semantic retrieval. The effectiveness of the proposed method is demonstrated through validation on alarm data and related operational manuals from a public simulation model.

Keywords: Fault detection and isolation methods, Monitoring, performance assessment, and fault detection in chemical process control, Process performance monitoring/statistical process control
Abstract: Cross-condition bearing fault diagnosis remains challenging because vibration patterns vary substantially under different operating conditions, and most existing methods depend mainly on data-driven features with limited physical interpretability. This paper proposes a physics-informed semantic domain adaptation (PISDA) method for cross-condition bearing fault diagnosis. Specifically, bearing semantics are extended with an adjustable bandwidth to form a continuous semantic field, based on which an interpretable physics-informed convolution layer is designed to provide physically meaningful decision outputs. In addition, a lightweight dual-branch architecture is devised, where the semantic branch and a general branch operate in parallel to jointly ensure physical consistency and discriminative capability. The effectiveness of the proposed PISDA is demonstrated by a case study with a public dataset. The results show that the proposed model achieves superior performance under various operating conditions while maintaining physical interpretability.

Keywords: Monitoring, performance assessment, and fault detection in chemical process control, Machine learning and artificial intelligence in chemical process control, Health/condition monitoring in processes
Abstract: The performance of industrial process fault diagnosis is often degraded by noise of varying intensities during data acquisition and by ambiguous fault class boundaries. To address these challenges, we propose a spatio-temporal fusion variational graph convolutional shrinking network (STVGCSN). In the spatial domain, a gating-based adaptive soft-thresholding function is integrated into the graph convolutional network to suppress noise of varying intensities, while in the temporal domain, a one-dimensional convolutional neural network is used to extract temporal dependencies. Moreover, a variational autoencoder architecture is employed to fully exploit both supervised and unsupervised information, thereby mitigating challenges caused by ambiguous fault boundaries. Finally, an attention-based strategy is further introduced to dynamically weight multiple loss terms, enabling balanced optimization across multiple objectives. Experimental results on two representative chemical process datasets demonstrate that the proposed model STVGCSN delivers superior performance in fault diagnosis tasks.

Keywords: Monitoring, performance assessment, and fault detection in chemical process control, AI methods for FDI/FTC, Machine learning and artificial intelligence in chemical process control
Abstract: Reliable fault diagnosis in industrial processes requires not only accurate classification but also trustworthy and actionable explanations for decision support. This paper addresses the need for validating the interpretability of machine learning models used in alarm-based monitoring. We use fault-specific permutation importance and introduce the Root Cause Alignment Score (text{RCAS}), a new metric designed to quantify the correspondence between a classifier's feature importance and known physical root causes. By applying the proposed interpretability assessment framework to a set of classification models on alarm data generated from the Tennessee Eastman Process (TEP), the trade-off between predictive performance and diagnostic interpretability was evaluated. The results show that complex ensemble models, despite higher predictive performance, can exhibit lower RCAS values, indicating that the explanations from these models may be less aligned with known physical root causes. In contrast, models of lower complexity, such as support vector machines, achieved the highest diagnostic validity, though often at the cost of reduced classification performance. This study highlights that, in addition to conventional performance metrics, the RCAS can guide the selection among well-performing classifiers to achieve interpretability that directly supports operator decision-making for effective fault mitigation.

Keywords: AI tools in automation engineering and operation, Digital twins for cyber physical systems, Model driven engineering of control systems
Abstract: Engineering projects depend on the processing of unstructured, multimodal design documents such as P&IDs, Control Narratives, and IO lists. Converting these into structured, machine-readable representations remains a major challenge. In a previous work, we presented the Engineering Data Funnel (EDF) system, which already demonstrated significant progress in automating multimodal engineering data processing. However, its single-pass extraction approach and limited reasoning capabilities left gaps in information completeness and consistency. To address these limitations, building on EDF, we here introduce the Extended Engineering Data Funnel (xEDF), which wraps EDF into a Neuro-Symbolic Agentic Loop. xEDF iteratively applies symbolic reasoning, gap analysis, and targeted re-extraction to maximize information completeness and semantic consistency, optionally involving human experts. Evaluation of xEDF shows that EDF got significantly improved, and it demonstrates how combining neural flexibility with symbolic precision enables an effective and more reliable and comprehensive engineering data processing for instantiation of industrial plant digital twins.

Keywords: Electrical distribution systems, Distributed optimization for smart grids, Solar energy
Abstract: Reinforcement learning (RL) is increasingly explored for power system control, yet deploying purely learned policies remains challenging due to partial observability and data inefficiency in complex distribution grids. We apply the Contextualized Hybrid Ensemble Q-learning (CHEQ) framework to centralized voltage control in low-voltage distribution grids, combining a Sequential Droop Controller (SDC) prior with a learned RL policy. An uncertainty-driven mixing mechanism, estimated from a critic ensemble, adaptively adjusts the relative contribution of each: granting the RL agent full authority in well-explored operating conditions, and deferring to the reliable SDC prior when uncertainty is high. Experiments on a realistic low-voltage benchmark grid demonstrate that the proposed approach eliminates voltage violations more effectively than the SDC and pure RL baseline while substantially reducing active power curtailment, confirming that reactive power is prioritized before resorting to generation curtailment.

Keywords: Electrical transmission systems, Energy management systems
Abstract: This paper introduces a novel evolutionary algorithm which simulates how the achievements of generations are inherited and optimized (AGIO) during the evolution of human’s family trees. The AGIO supposes a number of progressed family trees which have one offspring at each generation (single family trees). Each individual in a family tree with good intention tends to improve the inherited achievement by his ancestors. The achievement of each family tree is maximized based on the capability of its individuals. The capability of each individual modulated in accordance with creativity, personality, random/normal conditions of the generation and the inherited achievements. The novel AGIO algorithm can solve a number of engineering problems, where the state variables are considered as capabilities of individuals and the objective values as achievements of family trees. In this paper, the results of reactive power and voltage controls obtained by AGIO are compared with those given by the very popular PSO and the evolutionary algorithm of Grey-Wolf Optimizer demonstrating the superiority of the first.

Keywords: Health aware control in processes, Energy storage systems, Real time simulators for energy systems
Abstract: Battery EIS characteristics vary substantially with SOC, causing impedance-based features selected at a single SOC to lose reliability under different operating conditions. This study identifies SOC-invariant frequency features in the mid to high-frequency region through correlation analysis using cells aged at two temperature conditions. The four selected features were validated through 24-fold cross-validation and five randomized seeds, achieving an RMSE of 0.505% and a standard deviation of 0.021%, compared with the baseline (0.823%, 0.050%). These results demonstrate significant improvements in accuracy, robustness to SOC variation, and reproducibility, confirming the proposed features’ potential for reliable SOH estimation in practical applications.

Keywords: Distributed optimization for smart grids, Power systems stability, Electrical distribution systems
Abstract: This paper introduces a graph attention network-proximal policy optimization (GAT-PPO) framework for optimal placement of hybrid measurement devices in power systems. A stochastic power system observability model is proposed, which unifies heterogeneous measurements and uncertain constraints to ensure robust system observability amidst uncertainties caused by renewable generation. The challenge is first formulated as a stochastic optimization problem, which is then framed as a Markov decision process (MDP) to find a robust placement policy. A robust training approach simulates the stochastic failure of zero-injection buses (ZIBs), enabling the development of resilient, cost-effective measurement strategies that enhance state estimation accuracy in renewable-integrated power systems.

Keywords: Energy market, Distributed optimization for smart grids, Control and management of energy systems
Abstract: This paper studies strategic bidding in network-constrained electricity markets from an online learning perspective. In repeated DC optimal power flow auctions on the IEEE 14-bus system, generators choose bid multipliers and adapt them over 200 rounds using full-information no-regret algorithms. We implement Follow-the-Leader (FTL), Follow-the-Perturbed-Leader (FTPL), Multiplicative Weight Update (MWU), Regret Matching Plus (RM+), Discounted Regret Matching (DRM), and Optimistic RM+ (Opt-RM+), and compare them with truthful bidding and a welfare-maximizing correlated-equilibrium benchmark. Mean welfare remains close to the truthful and benchmark value of about 5720/h, with less than 1% loss even for MWU. MWU obtains the highest mean profit, about 106/h compared with about 81/h under truthful bidding, but has the largest average external regret, about 2.3/h, and persistent volatility. FTPL and Opt-RM+ provide the strongest welfare–regret trade-off by increasing profit while keeping regret near 0.1–0.2/h. Overall, carefully selected no-regret bidding rules preserve near-benchmark welfare while improving generator profitability.

Keywords: Power electronics, Real time simulators for energy systems, Electric vehicles and charging stations
Abstract: This paper explores a direct data-driven control (DDDC) strategy for a dualactive bridge (DAB) DC/DC converter employing single-phase-shift (SPS) modulation. In this proposed approach, the nonlinear term of the DAB DC/DC converter is defined as the control input, establishing a one-to-one relationship between the control input and the phaseshift ratio. This formulation enables the application of the DDDC method. The fundamentals of the DDDC approach are first presented, focusing on determining a stabilizing matrix for the closed-loop system. However, since the obtained solution may not be unique, an optimal DDDC approach is subsequently proposed to optimize the controller gain and introduce a tuning factor. In particular, a closed-loop data-driven controller with an integral tracking error term is developed, and its gains are analytically derived using Lyapunov theory to guarantee system stability. The proposed control approach is validated through real-time simulations on a Typhoon HIL 404 platform interfaced with a TMS320F28379D DSP microcontroller, demonstrating its effectiveness.

Keywords: Demand response, Energy market, Energy management systems
Abstract: Recently, pricing for electricity contracts has been becoming increasingly difficult due to changes in electricity demand due to the large-scale introduction of solar power generation and increased volatility in international energy commodity prices. Under these circumstances, demand response (DR) can be expected to become a highly flexible and desirable contract for buyers and sellers in the current situation of increasing uncertainty. In this study, the Least Squares Monte-Carlo Simulation method (LSM) is applied to value the DR concluded between an electricity retailer and a consumer who has batteries. Numerical experiments confirm that increasing the number of trials in the Monte Carlo simulation also makes the DR value evaluation more accurate. For the range of parameters given in this study, the relative standard deviation of the DR value was found to be less than 1 percent when the number of trials was 100,000.

Keywords: Electrical distribution systems
Abstract: In recent years, an increasing number of renewable energy sources, such as photovoltaic power generation (PV), have been interconnected to power systems. The output fluctuations of these sources have made it difficult to maintain appropriate voltage in power systems. The authors have previously investigated ways to maintain appropriate voltage by using the reactive power control of a PV inverter (smart inverter SI). This paper describes an optimal reactive power control method for a power system consisting of multiple distribution lines, taking into account the reduction of tap operations of Load Ratio control Transformers (LRTs) and Step Voltage Regulators (SVRs), the reduction of distribution line losses, and the fairness of the lifetime among multiple SIs. The authors also compare different impedance of distribution lines and consider the optimal reactive power output setting conditions.

Keywords: Electrical distribution systems, Solar energy
Abstract: This paper proposes a data-driven distribution network reconfiguration optimization method designed to address renewable energy source (RES) uncertainty. Unlike overly conservative robust optimization in our previous works, the proposed method integrates data-driven distributionally robust optimization into a constrained multiobjective evolutionary algorithm (CMOEA) for solving the reconfiguration problem. By leveraging historical PV data and employing an approximated chance-constrained approach to model uncertainty, the proposed method achieves scalable and stable search performance. The efficiency of the proposed method is validated through computational experiments on a large-scale 118-bus distribution network model subject to severe constraints arising from RES uncertainties.

Keywords: Hydrogen systems for energy generation and storage, Electric vehicles integration in energy networks, Distributed optimization and control for smart cities
Abstract: Active Distribution Networks (ADNs) with high penetration of renewable generation, energy storage, and electric vehicles (EVs) require advanced Energy Management Systems (EMSs) capable of handling operational uncertainty while maintaining power quality. This paper presents a unified EMS framework for ADNs that integrates neural-network-based forecasting, hybrid battery-hydrogen energy storage scheduling, stochastic EV uncertainty modeling, and decentralized voltage control. The proposed EMS comprises coordinated functional modules for forecasting, supply-demand management, handling EV uncertainty, and multi-agent voltage regulation. EV arrival/departure times, travel distances, and initial state-of-charge are modeled probabilistically, and representative scenarios are generated using Monte Carlo simulation and reduced using a Wasserstein-distance method. A two-stage stochastic optimization framework is then used for day-ahead scheduling and real-time operational adjustment through model predictive control (MPC). Voltage regulation is achieved using sensitivity-based coordination of on-load tap changers, step voltage regulators, and inverter reactive power support. Although these functions operate independently in the present study, the framework demonstrates a pathway toward an integrated EMS capable of jointly managing distributed energy resources, voltage regulation, and EV uncertainty in ADNs. Simulation results demonstrate that the proposed EMS reduces operating cost and imbalance energy while maintaining acceptable voltage profiles under uncertain EV behavior and renewable generation variability. The framework provides a scalable and practical solution for future ADNs with high penetration of inverter-based resources and EVs.

Keywords: Cybersecurity in smart grids, Power systems stability, Wind power
Abstract: This paper addresses the design of a decentralized resilient controller for frequency regulation in Multi-Area Power Systems integrated with wind power generation. Diverging from standard load frequency control design methods, we first propose a refined equivalent model that explicitly incorporates the equality constraint arising from the lossless power conservation across inter-area tie-lines. Building upon this formulation, the system, operating under Denial-of-Service attacks resulting from network openness, is modeled as a Markov Jump System, specifically considering the non-ideal scenario where the attack transition probabilities are partially unknown. Subsequently, by employing the Cone Complementarity Linearization approach, we present a systematic design scheme for a controller that ensures effective frequency regulation despite cyber-attacks and unknown attack evolution patterns. Finally, the efficacy of the proposed methodology is validated through simulations on a three-area power system.

Keywords: Demand response, Energy management systems, Distributed optimization for smart grids
Abstract: With the electrification of our uses and the massive integration of intermittent renewable production sources, demand-side flexibility is a way to maintain balance between production and consumption. This paper presents an individual flexibility capacity approximation that ensure confidentiality and low communication effort. This method is derived from the homothetic approximation technique. Our method is compared to other methods through a microgrid control problem. The results show that it provides a good tradeoff between fidelity and computation time, which makes it efficient for a massive group of buildings.

Keywords: Soft sensors in MMM systems, Machine learning and artificial intelligence in MMM process control, Machine learning and artificial intelligence in chemical process control
Abstract: Robust probabilistic latent variable models (RPLVMs), with excellent tolerance of outlying data, are widely used in industrial processes for soft sensing. However, the existing RPLVMs neglect the process causalities among process variables, degrading the generalization performance and model interpretability. To this end, a causal robust probabilistic principal component regression (Cau-RPPCR) is proposed. In the Cau-RPPCR, the causal relationships among process variables are first analyzed and a novel RPLVM structure incorporating the physical causal priors is thereupon designed. Then, an effcient semisupervised training algorithm, based on expectation–maximization, is exploited for the Cau-RPPCR. The performance of the Cau-RPPCR is evaluated on a numerical example and an actual industrial process.

Keywords: Process modeling, identification, and estimation techniques, Machine learning and artificial intelligence in chemical process control
Abstract: Soft sensing techniques have been extensively utilized in industrial applications for estimating key quality variables, thereby providing strong support for process control and optimization. However, the typical features of data generated in industrial processes such as the dynamics and nonlinearity pose significant challenges to traditional soft sensing methods, limiting their ability to extract these features accurately. Consequently, the prediction of key quality variables becomes unreliable. In this work, a hybrid model of LSTM-MHDA-iTransformer is proposed, where these correlations and significant nonlinearity of sequential process data are considered. First, an LSTM network is employed to extract local information from the process data. A modified iTransformer module is then applied to further capture global information. By integrating these components through a feature fusion mechanism, the proposed model achieves collaborative multi-scale modeling, which effectively combines micro-scale variations with macro-scale trends. Finally, compared with five soft sensor models in a real Chinese hydrocracking unit, the developed model achieves greater predictive accuracy and demonstrates superior overall forecasting capability.

Keywords: Machine learning and artificial intelligence in chemical process control, Machine learning and artificial intelligence in MMM process control, Process modeling, identification, and estimation techniques
Abstract: Modelling of spinning coagulation represents a crucial task in the wet spinning process of carbon fiber precursor filaments. This paper proposes an ARES-PINN (Adaptive Reweighting and Restart Strategy for PINNs) method based on a dynamic mechanism model for the wet spinning coagulation process. This approach addresses the current model's requirements for subsequent determination of optimal grid accuracy and computational complexity issues. This framework employs an adaptive loss weighting strategy and a restart approach for parameter initialization. By embedding physical laws (in the form of partial differential equations and boundary conditions) within the neural network's loss function, the model and method's efficacy are validated through a case study of a specific ternary polymer system. Dynamic simulation results demonstrate the method's performance under perturbations.

Keywords: Monitoring, performance assessment, and fault detection in chemical process control, Model-predictive and optimization-based control in chemical processes, Process modeling, identification, and estimation techniques
Abstract: Long-Term Series Forecasting (LTSF) is of great significance for energy efficiency analysis and long-term optimization control in industrial production. However, the complex non-stationarity and dynamic time-varying characteristics of industrial process data pose substantial challenges to industrial LTSF tasks. Therefore, this paper proposes a Temporal-Frequency Decomposition Network (TFD-Net) for industrial process LTSF tasks. The TFD-Net innovatively adopts a parallel dual-stream architecture to separately forecast the non-stationary and stationary information in industrial process data. Specifically, the Feature Decoupling Module (FDM) uses Fourier transform to decouple the time series, obtaining stationary and non-stationary components. A simple multi-layer perceptron (MLP) is utilized to predict non-stationary features. Meanwhile, the feature extraction module based on the seasonal-trend decomposition is utilized to capture stationary components. And the TFD-Net effectively captures non-stationary features of data while retaining detailed stationary feature, significantly improving prediction accuracy. Finally, the TFD-Net is validated on five benchmarks and an actual industrial production dataset. The experimental results show that compared with the current baselines, the TFD-Net achieves state-of-the-art results, with a reduction of at least 11.7% in the mean square error and 11.0% in the mean absolute error, which can effectively guide industrial production.

Keywords: Machine learning and artificial intelligence in MMM process control, Industrial applications of process control, MMM process modeling, identification, and estimation techniques
Abstract: Laser-based directed energy deposition is an additive manufacturing process that builds 3D parts by melting metal wire or powder with a concentrated laser. This work presents a data-driven reduced-order model (ROM) to predict the temperature field during this process. The developed ROM combines a dimensionality reduction step using proper orthogonal decomposition (POD) and a long short-term memory (LSTM) network to predict the temporal evolution in the reduced space. We demonstrate that the ROM achieves superior prediction accuracy for temperature profiles and melt pool geometry using 64 POD modes compared to a dynamic mode decomposition with control (DMDc) model. Furthermore, we integrate the developed ROM into a nonliner model predictice control (MPC) framework to achieve reference tracking of the melt pool area by controlling the laser power, a task which would be computationally intractable using a high-fidelity PDE model.

Keywords: Production and operations management, Supply chain management in manufacturing, Simulation and optimization in production, operations and services
Abstract: This paper presents a novel stochastic optimization model for long-term production planning in hybrid supply chains that combine manufacturing, remanufacturing, and recycling under uncertainty. The model introduces a unified chance-constrained framework that captures dynamic interactions among three inventory types and four production processes, extending conventional deterministic or single-stream approaches. It is reformulated as a deterministic linear–quadratic equivalent, enhancing tractability while preserving uncertainty in demand and return flows. This structure supports efficient solution methods within standard optimization tools and enables sensitivity analyses of recovery parameters. A numerical case study illustrates how varying recycling ratios affect optimal production, inventory, and cost over the long term. The results quantify the trade-off between manufacturing and remanufacturing, offering insights for robust and sustainable production planning. Overall, the framework integrates stochastic modeling and hybrid supply chain control, advancing decision-making for closed-loop and circular production systems.

Keywords: Farmland irrigation and drainage control, Water resource system modeling and control, Modeling and estimation in agriculture
Abstract: Hydrological models used by model-based control for drained field water management operate with weather forecasts, yet few studies have compared drained field models considering forecasts and limited calibration data. A linear model, a non-linear conceptual model, and a distributed physics-based model were calibrated and evaluated using field data from four subsurface-drained blocks. Short-term daily simulations driven by weather forecasts showed that all models outperformed a baseline assumption of constant groundwater depth. However, forecast uncertainty constrained the performance of all models. These findings suggest that dynamic modeling improves groundwater depth predictions and that conceptual models, more practical than physics-based models for real-world agricultural control or monitoring applications, can perform as well as physics-based models in forecast-driven short-term predictions.

Keywords: Computer vision in agriculture, Sensing and perception in agriculture
Abstract: This work presents a camera-based sensor for mapping plant growth over an agricultural field. The sensor can be used on ground-based vehicles like a tractor and relies on RTK-GNSS to correctly merge many multispectral images onto one map. NDVI is used as an index to estimate plant growth, but the approach can be used with other indices as well depending on the camera. The images from the camera are projected onto the estimated ground plane using perspective projection. This computationally simple approach allows for real-time processing of all images on the field even with low-end hardware. The results are compared to a commercially established vehicle-mounted sensor. Absolute values are hard to compare, but both sensors show similar trends. The camera-based approach also allows for filtering of ground and non-ground areas, potentially reducing the impact of crop density on the average measured NDVI.

Keywords: Control in precision agriculture, Modeling and estimation in agriculture, Sensing and perception in agriculture
Abstract: Autonomy is essential for addressing many of the future challenges in agriculture. In autonomous soil cultivation, the process of preparing the field before sowing, knowledge of the spatial soil conditions is crucial for achieving stable and efficient operation. This paper formulates a new control problem for a general single-tool cultivation machine, where the goal is to estimate the spatially varying soil conditions online while maintaining high control performance. To handle the partially unknown system dynamics, a consistent estimation method based on the Instrumental Variables (IV) framework is proposed. A simulation study demonstrates that the method accurately identifies the soil-condition function and clearly outperforms the classical Least Squares (LS) approach.

Keywords: Modeling and estimation in agriculture, Computer vision in agriculture, Sensing and perception in agriculture
Abstract: Against the backdrop of the increasing need for techniques to monitor and determine the canopy structure in viticulture, this work deals with the search for a method for analysing canopies in vineyards, considering LiDAR and RGB data, and investigating the development of leaf areas across several stages of vine development. To this end, LiDAR and RGB data were combined in a first step to generate artificial RGB-D data. These were used to train two U-Net models, one using RGB data (U-NET_RGB) the other using RGB-D data (U-NET_RGB-D). Since accuracy and Dice Score values for U-NET_RGB were higher, it was used to estimate the foreground canopy from Eden Viewer RGB frames, of different growing stages of the vine and different defoliation levels, and calculate the corresponding leaf wall area. The probability density distribution was plausible and in line with other related studies.

Keywords: Robotic manipulation of agricultural materials, Agricultural robotics, Sensing and perception in agriculture
Abstract: This paper presents a tactile predictive pushing framework for gentle cluster-envelope clearing of unstable hanging fruit in dense agricultural canopies. A secondary arm equipped with a laminate capacitive tactile array uses sparse intensity signals to estimate a local contact-center surrogate, a directional pseudo-force, and a rate-weighted moment surrogate. These cues bias a Cartesian TCP controller for short-range lateral displacement of a grape-cluster envelope while limiting contact-surrogate peaks. Experiments over 20 repeated trials on the same grape cluster show smoother trajectories and shorter high-contact duration than a visual-feedback teleoperation baseline.

Keywords: Modeling and estimation in agriculture, Automation for post harvest technology, Control in precision agriculture
Abstract: This paper proposes an observer-based control engineering approach for in-silo drying and moisture content regulation of cereal grains. By using a suitable linearized finite-dimensional approximation of drying process and the two-time-scale nature of in-silo drying phenomena which dynamically decouples intergranular air humidity and grain moisture content evolutions, a two-loops cascaded control structure is built. A linear observer is used to complete previous control studies by estimating the non-measurable moisture content values in the grain mass. The grain drying process is driven by controlling intergranular air humidity within an inner loop built around a PI controller. The averaged moisture content is controlled within an outer loop containing a state feedback and an integral driving action, its feedback being based on grains moisture estimate. Performance of the proposed control approach has been validated via numerical simulation. The results enable control implementation on a real-world grain silo.

Keywords: Biomedical system modeling, identification, and simulation, Digital twins in healthcare, model-based therapeutics, Pharmacokinetics, tracer kinetic modelling and drug delivery
Abstract: Maintaining hemodynamic stability during general anesthesia is essential for ensuring patient safety. However, accurately predicting blood pressure responses to anesthetic drugs remains a major challenge due to strong inter- and intra-patient variability and the presence of unforeseeable surgical events. As a result, existing models often show substantial mismatches when tested on real-world surgical data, limiting their suitability for incorporation into multivariable, model-based control systems. In this study, we introduce a framework for intraoperative hemodynamic prediction based on neural controlled differential equations. Leveraging their ability to learn non-autonomous dynamics from irregularly sampled clinical time series, and through a straightforward architectural adjustment, the model can be deployed online to update its latent state in real time as new patient observations become available. By jointly considering drug concentration trajectories and blood pressure measurements, the model continuously realigns its predictions with the patient’s changing physiological condition. This approach lays the groundwork for future automated, multivariable anesthesia control strategies implemented within a digital twin setting.

Keywords: Biomedical and medical imaging, image processing, visualization, Biomedical signal measurement and processing
Abstract: This paper presents a novel methodology for classifying symptomatic and asymptomatic carotid artery plaques from CT images using quantum-inspired features. The proposed approach applies two-dimensional semi-classical signal analysis (2D-SCSA) to extract spectral features from each slice and subsequently constructs spatial sequences that track the evolution of these features across the entire volume. Statistical and frequency-domain descriptors computed from these spatial sequences capture three-dimensional morphological and textural characteristics of the plaques. Using stratified group K-fold cross-validation with hyperparameter tuning, multiple machine-learning models are trained on the extracted features. Experimental results on real clinical data confirm the effectiveness of this hierarchical feature-extraction strategy, demonstrating its strong potential to improve clinical diagnosis and risk assessment of carotid artery disease.

Keywords: Biomedical signal measurement and processing
Abstract: Heart rate variability (HRV) is a key marker of autonomic nervous system function, but its reliability depends critically on accurate detection of R–R intervals from ECG signals. Conventional peak detection algorithms are highly susceptible to noise, motion artefact, and missed or spurious beats, often requiring manual correction and limiting suitability for realtime clinical use. While Bayesian filtering has previously been applied to heart rate estimation, it has not been used for full HRV extraction or to fuse multiple noisy detectors. We present a discrete-time Bayesian histogram filtering framework for robust HRV analysis that integrates peak candidates from multiple sensors or detection algorithms. The method models detector outputs as Gaussian observations and applies a physiologically informed transition model based on adaptive R–R interval predictions. At each step, the filter produces a posterior distribution over likely beat locations, enabling automated peak selection, rejection of outliers, and recovery from missed detections. We evaluate the approach on synthetic and real ECG datasets with annotated ground truth. Results demonstrate improved accuracy and substantially greater robustness to severe noise and detection failures compared with standard algorithms, particularly in worst-case epochs. The method provides a principled foundation for reliable real-time HRV processing in clinical and wearable settings.

Keywords: Biomedical system modeling, identification, and simulation, Digital twins in healthcare, model-based therapeutics, Biomedical and medical imaging, image processing, visualization
Abstract: Arterial narrowing and stiffening are root causes of poor heart health, myocardial infarction and peripheral vessel diseases. An in-vivo sensor-actuator system embedded on inflatable elastomer balloon is under development at Furtwangen University. The signals from the sensing segments on balloon are intended to provide a tactile assessment of tissue stiffness and shape, which can be further used to generate decision aids for a Vascular surgeon. This 2D study focused on implementing a Finite Element model to evaluate differential strains on lumen surface of 4 partly calcified arteries captured using an inflatable balloon setup embedded with 128 strain sensing elements. The deformation analysis of the balloon shows few elements initially conforming to stiffer unhealthy section of vessel lumen shift over to the section with normal stiffness as balloon pressure was increased. An analytical method was developed that tracks the maximum/minimum strain on elements and the corresponding sensing element number. A simplified calculation further allows estimation of relative local strains of the tissue region via balloon elements and therefore relative stiffness. The unhealthy region underwent relative strains ranging from 0.3 to 0.47 compared to the healthy region as observed from the Finite Element Analysis model whereas the analytical method computed the same in range of 0.4 to 0.5. It could be shown from simulation that an in-plane assessment involving inter-regional stiffness behaviour could be obtained. This sensor system could be used to estimate local tissue health and generate better informed decision aids.

Keywords: Biomedical system modeling, identification, and simulation, Biomedical signal measurement and processing, Decision support and control in medicine
Abstract: Cardiovascular management in the intensive care unit (ICU) is difficult and often sub-optimal. Model-based methods offer the opportunity to turn a relatively low number of clinically available measurements into a clearer picture to guide care. However, to use models effectively, the aortic pressure and conditions around the heart, which are not directly measured, must be known. This paper presents a method of identifying model parameters of two validated single-channel arterial transfer function models in parallel, to estimate central aortic pressure (Pao,est) from two peripheral arterial measurements. The method reduces the need for invasive pressures to be measured for use in cardiovascular models, particularly the three-chamber model, which requires the mean, range, and contractility of the aortic pressure as an output. The method uses MATLAB’s built-in genetic algorithm to minimize the difference between the output waves of two arterial transfer functions, using the pressure measurements and foot-to-foot time difference between the pressure waves. The method was implemented on 100 virtual patients from the HeaMod online database with known ground truth for quantitative validation, followed by two patients of clinical data from Christchurch Hospital ICU for qualitative validation. The range, mean and contractility error was calculated between the known and estimated virtual patients to assess method accuracy. For the clinical data, visual inspection of the waveforms and objective function evaluation were used. The method performed well in in-silico data, but precision in real clinical measures and their relative pulse transit time difference were a limitation in evaluating clinical patient-specific aortic pressures.

Keywords: Digital twins in healthcare, model-based therapeutics, Healthcare management, disease control, critical care, Biomedical signal measurement and processing
Abstract: Lumped-parameter cardiovascular models combine routine patient data with mathematical representations of circulatory physiology to estimate otherwise unmeasurable hemodynamic parameters. When individualised using patient-specific measurements, these models can yield clinically actionable insight beyond raw bedside signals. The three-chamber model (TCM) offers such potential, but its standard formulation requires direct aortic and ventricular measurements absent from routine ICU practice. This study introduces a clinically feasible TCM implementation (TCM_CF) that estimates the missing model inputs from available ICU data, and evaluates its performance against a full-measurement version (TCM_FM) using invasive reference measurements.