Keywords: Adaptive control design
Abstract: Control systems are traditionally designed using continuous-time or time-triggered sampled signals, where information is encoded in amplitudes. Recently, event-triggered control has shown that control actions can be generated asynchronously and only when needed, thereby improving efficiency in networked and embedded implementations. Neuromorphic control, inspired by the functioning of biological neurons, pushes this paradigm to its extreme: control signals consist of sequences of fixed-amplitude spikes, so information is encoded entirely in the timing of spiking events.

In this talk, we address the fundamental question of how feedback controllers can be designed when the only control freedom lies in spike timing. We present a framework in which neuromorphic controllers are modeled as hybrid systems, combining continuous evolution with discrete state jumps induced by spikes. Based on this viewpoint, we discuss complementary design approaches, including Lyapunov-based triggering strategies and emulation-based constructions that approximate continuous feedback laws. Together, these results provide building blocks toward a design theory for spiking control systems.

We illustrate these design ideas through a nuclear fusion application, where plasma fueling through pellet injection yields inherently impulsive actuation, making neuromorphic feedback strategies highly relevant. The talk concludes with an outlook on this exciting research area.

Keywords: Advanced process control
Abstract: The world is on the cusp of a new era of electrification across different sectors of industry and economy. A key technology driving this transformation is lithium-ion batteries. As the best available power source, lithium-ion batteries provide high energy/power density and long cycle life. As they find every-growing use in electric vehicles, electric aircraft, grid storage and autonomous platforms, demands for their performance and safety have been rising. Systems and control theory can play a key role in meeting the needs to advance the application of battery systems, resulting in provable progresses.

This tutorial-style workshop is designed to provide a deep, structured introduction to the state of the art and new frontiers in battery modeling, monitoring and control, with a focus on integrating physical insights and control-theoretic rigor with practical implementation. The workshop will be particularly relevant to researchers and practitioners within the systems and control community looking to expand their research towards developing and applying control-theoretic methods for lithium-ion batteries and emerging battery-powered systems.

Keywords: Adaptive control design
Abstract: Generating high-resolution stability charts typically requires exhaustive point-by-point grid sweeps across a multi-dimensional parameter space, leading to exponential computational costs. This paper introduces the Multi-Dimensional Bisection Method (MDBM) as a robust, highly efficient alternative to overcome this curse of dimensionality. Implemented as a well-known, highly optimised, and mature open-source package in Julia and in Matlab, MDBM is designed to implicitly find and trace complex multi-dimensional boundaries, such as stability borders and envelopes of parametric families. By evaluating constraints on hypercube vertices and utilising localised iterative refinement, MDBM restricts computational effort solely to the vicinity of the actual boundaries, reducing processing times from hours to seconds. In this session, we will go beyond the core mechanics to explore the finer settings, extra features, and advanced properties of this well-established method, demonstrating how to extract maximum performance when charting stability boundaries in delayed dynamical systems.

Keywords: Adaptive control design
Abstract: While identifying the marginal stability boundary is critical, practical engineering applications require optimizing performance and ensuring robustness against parameter uncertainties. This section extends the framework to performance optimisation by analysing swept regions and envelopes of parametric families. Instead of using brute-force discretisation to map out performance drop-offs, we propose a unified boundary handling method based on non-linear parameter reparameterization and global directional constraints. This enables us to formulate the tracking of the "fastest decay" rate (such as the spectral abscissa) and the calculation of explicit robustness margins as a single, constrained system. By utilising a component-wise mapping that vanishes at interval endpoints, we eliminate the combinatorial overhead of analysing lower-dimensional boundary strata separately. When integrated with advanced boundary-tracking algorithms like the Multi-Dimensional Bisection Method (MDBM) , this approach bypasses the tracking of irrelevant interior curves and isolates only the active, exposed boundaries of the robust parameter domains.

Keywords: Adaptive control design
Abstract: For systems with time-periodic parameters—such as machining chatter with spindle speed variation or delayed networked controls—stability is governed by Floquet theory. Traditional time-domain techniques, like the standard Semi-Discretization method, approximate the infinite-dimensional Monodromy operator but suffer from high computational demands (typically cubic or quadratic complexity) due to repeated matrix multiplications over discrete time intervals. This paper presents the Multiplication-Free Semi-Discretization (MFSD) method, a novel numerical approach that compiles the discretized governing equations into a single, highly sparse generalized eigenvalue problem. Inspired by collocation techniques, MFSD completely circumvents successive matrix multiplications, reducing the computational complexity to linear time. The paper will demonstrate how MFSD breaks traditional computational bottlenecks, enabling near-real-time spectral analysis and stability prediction for complex time-periodic delayed systems. Furthermore, we will highlight how this architectural breakthrough extends to stochastic delayed systems; by reformulating the second-moment dynamics, the computational complexity of the stochastic semi-discretization method has been successfully reduced from the traditional quartic complexity to just quadratic, opening new frontiers for robust control design under random perturbations.

Keywords: Control using FAS approach
Abstract: This study investigates the tracking control for discrete-time strict-feedback systems with the mismatched disturbances by means of a fully actuated system (FAS) method. Firstly, an equivalent transformation is constructed to convert the discrete-time strict-feedback systems with the mismatched disturbances into a class of input-delay FASs with the lumped disturbance. Then, a high-order disturbance observer (HODO) is designed by using a difference operator and its high-order form to achieve the accurate estimation of the lumped disturbance via a less conservatism assumption. With the help of the FAS method, a predictive proportional-integral (PI) control with the disturbance compensation is designed to implement the desired tracking control with eliminating the open-loop nonlinearities and compensating for the input delays. A sufficient criterion is presented to analyze the bounded stability and asymptotic tracking of the closed-loop FASs. Finally, a simulation of the Chua's circuit is shown to verify the effectiveness.

Keywords: Control using FAS approach, Fully-actuated systems in industry, Global fully actuated systems
Abstract: A model-free controller that employing only the link angle measurement is developed for flexible joint robots, accompanied by a rigorous stability analysis. Notably, the developed controller guarantees stabilization from arbitrary practical initial conditions.

Keywords: Control using FAS approach, Global fully actuated systems, Fully-actuated systems in industry
Abstract: Most existing control barrier function (CBF) strategies for robotic manipulators either suffer from complex optimization arising from inertia matrix coupling or demand precise model knowledge unavailable in practice. To address these limitations, we propose a robust safety-critical control framework for uncertain manipulators with strict multiple constraints, built upon the fully actuated system (FAS) approach. A robust integral of the sign of the error (RISE) disturbance observer is incorporated to achieve exponential estimation of lumped uncertainties, yielding a computable time-varying error bound that directly reduces conservatism in safety certification. By exploiting the FAS transformation, the nonlinear manipulator dynamics are converted into a perturbed double-integrator structure with respect to a virtual control input, whereupon all safety constraints reduce to decoupled, component-wise affine inequalities that eliminate inertia-matrix coupling from the optimization. The resulting strictly convex quadratic program (QP) guarantees high-precision trajectory tracking under simultaneous position and velocity constraints. Simulation results on a two-link planar manipulator validate the effectiveness of the proposed framework.

Keywords: Global fully actuated systems, Control using FAS approach, Predictive control of fully-actuated systems
Abstract: This paper presents a fully actuated system (FAS) predictive control method for stabilizing discrete-time nonlinear systems with time-varying input delays. The design integrates FAS approach with an online predictor that actively compensates for the variable delay by forecasting system states based on current delay measurements. The control law is synthesized by imposing a desired linear dynamics on the predicted future state, thereby aligning the delayed actuation with the intended control objective. Stability is rigorously guaranteed under Lipschitz nonlinearity and bounded delay variation. Simulations show the proposed method achieves faster convergence, lower steady-state error, and more efficient control compared to conventional delay-ignoring or fixed-delay compensation strategies, offering a systematic solution for time-delay discrete-time high-order FASs (DT-HOFASs).

Keywords: Sub-fully actuated systems, Global fully actuated systems, Control using FAS approach
Abstract: Traditional control barrier function (CBF)-based controllers may become infeasible when the CBF control direction vanishes, preventing the enforcement of safety constraints. In this work, the problem is resolved by transforming the wheeled mobile robot (WMR) into a fully actuated system (FAS). Nevertheless, when the FAS form are used to achieve tracking, the position loss of WMR is inevitable. To avoid this issue, the nominal controller and the associated control Lyapunov function (CLF) are constructed using only a subset of the FAS structure inherent to the WMR. A disturbance-observer-enhanced CLF-CBF quadratic program ensures feasibility under disturbances while guaranteeing both tracking performance and safety. Simulation results validate the effectiveness of the proposed method.

Keywords: Data-efficient control via foundation models
Abstract: Designing controllers that simultaneously achieve strong performance and provable closed-loop stability remains a central challenge in control engineering. This work introduces a diffusion-based generative framework for linear controller synthesis grounded in the Youla–Kučera parameterization, enabling the construction of stabilizing controllers by design. The diffusion model learns a conditional mapping from plant dynamics and desired performance metrics to feasible Youla parameters, guaranteeing internal stability while flexibly accommodating user-specified targets. Trained on synthetically generated stable SISO plants with fixed-order Youla parameters, the proposed approach reliably synthesizes controllers that meet prescribed sensitivity and settling-time specifications on previously unseen systems. To the best of our knowledge, this work provides the first demonstration that diffusion models can generate stabilizing controllers, combining rigorous control-theoretic guarantees with the versatility of modern generative modeling.

Keywords: LLMs for modeling and control
Abstract: Industrial automation increasingly requires flexible control strategies that can adapt to changing tasks and environments. Agents based on Large Language Models (LLMs) offer potential for such adaptive planning and execution but lack standardized benchmarks for systematic comparison. We introduce a benchmark with an executable simulation environment representing the Blocksworld problem providing five complexity categories. By integrating the Model Context Protocol (MCP) as a standardized tool interface, diverse agent architectures can be connected to and evaluated against the benchmark without implementation-specific modifications. A single-agent implementation demonstrates the benchmark's applicability, establishing quantitative metrics for comparison of LLM-based planning and execution approaches.

Keywords: LLMs for modeling and control
Abstract: This paper proposes an activation control method for Large Language Model (LLM) safety, especially for Mamba which is the representative model constructed with the State Space Models (SSMs). Unlike the Reinforcement Learning from Human Feedback (RLHF) or prompt engineering, our proposed method directly manipulates the internal activations of the LLM, and it needs no fine tuning. The existing activation engineering methods are the static approach, such as simple addition of the fixed concept vector to the activation. The proposed method overcomes this indiscriminate steering and gives flexibility. At first, the Control Barrier Function (CBF) is defined with the determination of the unsafe set in the activation space of SSM layers. Then, the expert channels are also determined by the contrastive prompt's activations. Finally, the activations are controlled when they enter the predetermined unsafe set. The control inputs are calculated based on the current states so as to drive the activations out of the unsafe set. The experiments with mamba-2.8b model and Real Toxicity Prompts dataset show that the proposed method enhances the safety of LLMs.

Keywords: LLMs for modeling and control, Data-efficient control via foundation models
Abstract: Accurate speed estimation in sensorless brushless DC motors is essential for high-performance control and monitoring, yet conventional model-based approaches struggle with system nonlinearities and parameter uncertainties. In this work, we describe a transformer-based in-context learning framework to perform zero-shot speed estimation using only electrical measurements. By training the filter offline on simulated motor trajectories, we enable real-time inference on unseen real motors without retraining, eliminating the need for explicit system identification. Experimental results demonstrate that our method outperforms traditional Kalman filter-based estimators, especially in low-speed regimes that are crucial during motor startup.

Keywords: LLMs for modeling and control, Data-efficient control via foundation models
Abstract: We address the problem of high-performance speed control for BLDC motors under unknown and varying dynamics, where conventional robust or adaptive designs often require conservative tuning or extensive experimentation. We propose a Sim-to-Real In-Context Learning framework in which a Transformer policy, trained exclusively in simulation via Domain Randomization, acts as a zero-shot meta-controller that adapts directly from recent input--output data, without parameter identification or online adaptation laws. Experiments on a real BLDC platform with large inertial variations show that the learned controller generalizes reliably across operating conditions. Comparative results further demonstrate that it matches the performance of state-of-the-art black-box optimization while requiring no real-world training iterations, thus shifting the burden of adaptation from online experimentation to offline computation.

Keywords: LLMs for modeling and control, Development of assistant systems for manufacturing systems, Explainability and safety of LLM-based controllers
Abstract: We present A-SID (Agent-based System Identification), a proof-of-concept framework that combines a large language model with a Python-based execution environment for automated data preprocessing and linear model identification. Python routines first compute statistical diagnostics describing missing data, trends, noise levels, and outliers. These diagnostics are then passed to the LLM (OpenAI GPT4), which generates a structured preprocessing plan and executable code for subsequent Auto-Regressive with eXogenous input (ARX) model estimation and validation. The framework is evaluated on four synthetic input–output datasets covering ideal linear data, corrupted linear data, severe data anomalies, and nonlinear dynamics. The results show that A-SID can execute the complete identification workflow without manual intervention and that LLM-guided preprocessing improves ARX model accuracy in the linear cases. However, the nonlinear case also shows an important limitation: inappropriate preprocessing can remove genuine system dynamics while still improving numerical fit metrics. The study therefore demonstrates the feasibility of LLM-orchestrated system-identification workflows while highlighting the need for stronger validation of agent decisions.

Keywords: Aerial, field, and marine robotics, Autonomous navigation
Abstract: This paper presents a safe backup control method for multirotor systems, which generates a time-optimal trajectory to a predefined safe set and integrates a safety filter to ensure constraint satisfaction in real time. The approach enables rapid recovery from potentailly unsafe states while maintaining dynamic feasibility, adaptively adjusting the recovery horizon based on the current state to maximize the operational envelope. Simulation results demonstrate that the proposed controller guarantees safety under input and state constraints while effectively keeping the system within the safe operating envelope.

Keywords: Aerial, field, and marine robotics, Autonomous navigation
Abstract: This paper presents a geometric attitude tracking controller on SO(3) based on the control contraction metric (CCM) framework. Unlike conventional Lyapunov-based methods that guarantee stability around a fixed equilibrium, the proposed approach ensures exponential convergence of differential dynamics along trajectories. A Riemannian metric is constructed for rigid-body rotational dynamics, and a feedback law satisfying the contraction condition is derived. Numerical results validate the contraction property through eigenvalue analysis.

Keywords: Aerial, field, and marine robotics, High-performance motion control systems, Task and motion planning
Abstract: Accurate and robust control of multirotor systems requires consideration of actuator dynamics, particularly during aggressive maneuvers and rapid command changes. Most existing control strategies assume instantaneous actuator response and neglect rotor and, if applicable, servo dynamics in the stability analysis. This assumption can degrade performance and lead to instability under fast transients. This paper proposes a control framework that explicitly incorporates actuator dynamics into the closed-loop design. A backstepping-based approach is employed to address the cascaded structure between rigid-body and actuator dynamics. This formulation is particularly advantageous for variable-tilt platforms, which experience heterogeneous actuation delays due to the distinct responses of rotors and servos. Supported by a rigorous Lyapunov-based stability analysis and robustness guarantees against actuator time constant uncertainties, the effectiveness of the proposed method is demonstrated through numerical simulations. The results demonstrate improved tracking performance compared to baseline method that neglect actuator dynamics.

Keywords: Agricultural robotics, Computer vision in agriculture, Sensing and perception in agriculture
Abstract: This study proposes a hybrid perception framework that integrates occlusion-aware apple detection with Vision Language-Model reasoning for human-in-the-loop supervision during robotic apple harvesting. A transformer-based RF-DETR model performs occlusion-based apple detection, while a probabilistic calibration scheme identifies predictions with low reliability. These ambiguous cases are provided to VLM for semantic occlusion reasoning and selectively escalated to human supervision when uncertainty or disagreement between RF-DETR and VLM decision persists. Results demonstrated occlusion-aware detection mAP@50 of 86.2%, with calibration allowing 47% of detections to be directly accepted. VLM showed 74.2% classification accuracy on uncertain cases, highlighting promise for selective human-in-the-loop correction.

Keywords: AI-powered robotics, Human machine cooperation & integration, Robotic learning and adaptation
Abstract: Cooperative multi-agent reinforcement learning often relies on intrinsic rewards to discover coordination, yet naive reward mixing can distort the learning signal and degrade team performance. We propose a constrained exploration framework that maximizes exploration subject to a conservative non-degradation constraint on extrinsic task return, solved via an epigraph reformulation with adaptive exploration budgets. This approach separates intrinsic rewards from task objectives and regulates exploration through task feasibility. Our approach consistently outperforms strong baselines and induces novel cooperative strategies in multi-agent tasks.

Keywords: Soft robotics, Robot perception and sensing, Autonomous navigation
Abstract: This paper presents a low-cost, bio-inspired soft robotic antenna designed for close-proximity depth sensing, modeled after the tactile functionality of insect antennae. Fabricated from a flexible silicone rubber substrate with embedded microchannels containing conductive carbon paste, the sensor utilizes resistance-based feedback to estimate distance. The sensor's performance is characterized through time-domain analysis and modeled using a fifth-order recursive least squares estimation (RLSE) algorithm. The effectiveness of the soft antenna is demonstrated through an autonomous wheeled robot executing Bug1, Bug2, and TangentBug algorithms for wall-following and obstacle avoidance. Results show that the soft antenna provides reliable distance sensing and accurate obstacle detection, offering a resilient and scalable solution for bio-inspired perception in autonomous systems.

Keywords: Aerial, field, and marine robotics, Task and motion planning, High-performance motion control systems
Abstract: Aerial manipulation combines the maneuverability of multirotors with the dexterity of robotic arms to perform complex tasks in cluttered spaces. Yet planning safe, dynamically feasible trajectories remains difficult due to whole-body collision avoidance and the conservativeness of common geometric abstractions such as bounding boxes or ellipsoids. We present a whole-body motion planning and safety-critical control framework for aerial manipulators built on superquadrics (SQs). Using an SQ-plus-proxy representation, we model both the vehicle and obstacles with differentiable, geometry-accurate surfaces. Leveraging this representation, we introduce a maximum-clearance planner that fuses Voronoi diagrams with an equilibrium-manifold formulation to generate smooth, collision-aware trajectories. We further design a safety-critical controller that jointly enforces thrust limits and collision avoidance via high-order control barrier functions. In simulation, our approach outperforms sampling-based planners in cluttered environments, producing faster, safer, and smoother trajectories and exceeding ellipsoid-based baselines in geometric fidelity. Actual experiments on a physical aerial-manipulation platform confirm feasibility and robustness, demonstrating consistent performance across simulation and hardware settings. The video can be found at https://youtu.be/hQYKwrWf1Ak.

Keywords: Adaptive observer design, Time/parameter varying system identification, Discrete event modeling and simulation
Abstract: This paper addresses adaptive observer design for Linear Parameter Varying (LPV) systems with sampled outputs and time-varying unknown parameters. The proposed observer combines a robust Multiple Proportional Integral (MPI) structure with a closed-loop output predictor to jointly estimate the states and unknown parameters while compensating for sampled measurements. The design is formulated through delay-dependent Linear Matrix Inequalities (LMIs), ensuring convergence of the estimation errors to zero and providing an upper bound on the Maximum Allowable Sampling Period (MASP). Nonlinearities are characterized by One-Sided Lipschitz (OSL) and Quadratic Inner-Boundedness (QIB) conditions, yielding less conservative results than classical Lipschitz-based designs. A numerical example illustrates the effectiveness of the proposed approach.

Keywords: Adaptive observer design, Hybrid and switched systems modeling
Abstract: Observation and control of switched systems have been widely studied in control theory, most of the work focuses on switched linear systems where each active mode is an LTI plant. However, this assumption can be conservative for many practical applications. This paper extends the analysis to switched bilinear systems, which more accurately capture the interaction between inputs and states in processes such as chemical reactors, flexible mechanical structures, and electrical machines. In this work, we propose numerically tractable, sufficient conditions for gain synthesis of Luenberger-Like Observers, Unknown Input Observers and unknown input reconstruction, using both common and non-common Lyapunov function approaches. The analysis considers both dwell-time constraints and arbitrary switching between active modes. A case study on a photovoltaic–thermal system illustrates the effectiveness and practical relevance of the proposed methodology.

Keywords: Adaptive observer design, Multi-agent systems, Distributed control and estimation
Abstract: This paper proposes a constructive distributed adaptive observer for discrete-time jointly observable uncertain linear systems over directed networks. A discrete-time system decomposition method is developed to mitigate the effects of unknown parameters. The proposed distributed adaptive observer consists of a parameterized observer for the observable-subsystem state and time-varying unobservable-subsystem dynamics, together with two nonlinear mappings that link the unknown parameters and the system states to the states of these dynamics. By establishing a parametric representation of the measurement output, the parameter estimation problem is converted into a parameter identification problem and is then solved via a gradientdescent method. By Lyapunov analysis, we show the asymptotic stability of the estimation error system, ensuring distributed state estimation despite system uncertainties and the joint observability condition.

Keywords: Adaptive observer design
Abstract: This paper investigates the problem of estimating the orientation of mobile robots described by a unicycle dynamics and using position measurements. By applying transformations to both coordinates and time, we transform the dynamics into a chained form of nonholonomic systems, which is weighted homogeneous. We then propose a homogeneous observer that converges in finite time under mild restrictions on robot velocities, and we analyze its properties in the original time domain. The efficiency of the new observer is illustrated by comparative simulations with popular solutions.

Keywords: Model reference adaptive control, Linear system identification
Abstract: A model-reference adaptive control (MRAC) scheme with fast output tracking is combined with an external parameter estimator for multi-input-multi-output (MIMO) systems. The controller design is based on the SDU approach. Fast adaptive tracking is achieved by reducing the error equation relative degree to zero and employing a standard gradient-type adaptive law driven by the tracking error. The external parameter estimator uses a least-squares algorithm driven by prediction error. The key challenge in combining controller and estimator parameters is their dimensional incompatibility. The novelty of this work lies in the mechanism developed to integrate these two estimators into a stable adaptive control algorithm. Simulation results for a 2-input 2-output system with 14 parameters demonstrate remarkable convergence properties.

Keywords: Nonlinear adaptive control, Model reference adaptive control, Filtering and smoothing
Abstract: Accurate online identification of aerodynamic coefficients is essential for stable tracking performance and safe operation across varying UAV operation, yet classical adaptive estimators require persistent excitation (PE), a condition rarely met during typical flight manoeuvres. Recent advances have shown that incorporating memory filters can relax PE to an interval-excitation requirement. Based on this idea, this paper introduces an extended Dynamic Regressor Extension and Mixing (DREM) scheme for the nonlinear longitudinal dynamics of a fixed-wing UAV. The method transforms the aircraft equations into filtered regression models and employs an enhanced mixing strategy to relax the PE requirement to an interval/initial excitation (IE) condition. This results in fully decoupled scalar regressions, parameter convergence, and stable performance with low-excitation manoeuvres. Simulation studies demonstrate that the proposed estimator retrieves lift, drag, and pitching-moment coefficients with more accurate convergence than classical DREM under relaxed excitation, enabling more accurate and adaptive downstream control.

Keywords: Distributed control and estimation, Quantized systems, Linear system identification
Abstract: In this paper, we consider distributed parameter estimation with binary observations under measurement-side tampering, where each node observes a thresholded output whose label may be flipped and exchanges information over a communication graph. We develop a distributed recursive projection algorithm based on the diffusion strategy. Without imposing independence, stationarity, or Gaussian assumptions, we establish an almost sure logarithmic bound for a Lyapunov-type estimation error. Under a mild cooperative excitation condition, the estimates of all nodes are strongly consistent, even when each individual node is non-exciting. Simulations on a jointly exciting network corroborate the theory and show that the proposed algorithm converges, whereas non-cooperative and tampering-unaware baselines do not.

Keywords: Distributed control and estimation, Cyber security networked control, Quantized systems Profile: Invited Session
Abstract: In bearings-only target tracking, the nonlinearity of the measurement model amplifies the detrimental impact of measurement outliers. Consequently, distributed sensor networks become highly sensitive to these outliers, which can severely degrade both local estimation and network‑wide fusion accuracy. To address this, this work proposes a robust distributed pseudolinear Kalman filter (RD‑PLKF) for planar tracking that combines a bias‑corrected weighted least‑squares estimator with an adaptive outlier suppression mechanism. Rigorous stability analysis proves uniform boundedness of the mean‑square error under mild network connectivity and outlier assumptions. Numerical simulations demonstrate that the proposed RD‑PLKF maintains high tracking accuracy, strong robustness, and low computational cost across a range of outlier intensities, outperforming conventional pseudolinear Kalman filter and other robust filtering schemes.

Keywords: Distributed control and estimation, Cyber security networked control, Quantized systems
Abstract: This paper mainly investigates the distributed state estimation problem for multi-sensor networks under communication constraints. Considering the limited bandwidth and communication resources, we propose a binary-valued data transmission scheme and the corresponding distributed filter, which overcomes the drawback of current methods that require nodes to exchange the high-dimensional state information. Moreover, the filter requires only a single data exchange among nodes per iteration to synchronize observation updates and achieve consensus of state estimates. Finally, we provide proofs for the stability of the filtering algorithm, which have been validated through simulation examples.

Keywords: Linear system identification, Kalman filtering
Abstract: Reliable state estimation depends on accurately modeled noise covariances, which are difficult to determine in practice. This paper formulates the noise covariance estimation as a bilevel optimization problem that factorizes the joint likelihood of primary and supervisory measurements to reconcile information exploitation with computational tractability. The factorization converts the nested Bayesian dependency into a Markov-chain structure, allowing efficient computation. At the lower level, a Kalman filter with state augmentation performs such computation. Meanwhile, closed-form forward and reverse differentiation provide efficient gradients for the upper-level updates, and we compare the two modes’ space and time complexities to inform their practical selection. The upper level subsequently refines the noise covariances to guide the lower-level estimation. Taken together, the proposed algorithms offer a systematic and computationally efficient approach to noise covariance estimation in linear Gaussian systems.

Keywords: Nonlinear system identification, Quantized systems
Abstract: This paper studies the joint optimization problem of bit allocation and parameter estimation in distributed sensor networks, where sensors transmit quantized measurements to a fusion center under a total bit budget imposed on backhaul links. By leveraging the Bussgang decomposition, we formulate a covariance-based mean-square-error minimization problem that jointly optimizes the linear estimator and the bit allocation among sensors. Since the resulting problem is a non-convex mixed-integer program, direct global optimization is computationally prohibitive. To address this difficulty, an alternating minimization algorithm is developed for the continuous relaxation problem, followed by local search to obtain a feasible integer bit allocation. Simulation results show that the proposed method outperforms equal bit allocation and achieves accuracy close to exhaustive enumeration with much lower computational cost.

Keywords: Model reference adaptive control, Linear system identification
Abstract: The paper addresses the problem of parametric convergence enhancement up to finite time convergence (FTC) in schemes of direct adaptation that stem from the dynamic error model with measurable state. FTC is achieved under persistent or interval excitation conditions. In contrast to the majority of adaptation schemes with FTC, the main advantages of the proposed FTC mechanism consist in the following: 1) the proposed mechanism is compatible with standard algorithms of adaptation designed for a dynamic error model; 2) zeroing of the control error is always achieved for any bounded regressor without any additional conditions (like persistent or interval excitation, or like not-in-L_1 condition) what is important for direct adaptive control; 3) the FTC alertness is preserved under persistent excitation condition.

Keywords: Adaptive observer design, Estimation and filtering
Abstract: This paper presents a design of a heavy–ball algorithm–based adaptive observer for the simultaneous estimation of states and constant parameters in a class of uncertain nonlinear systems subject to external disturbances. The proposed estimator consists of a Luenberger–like observer for state estimation and a heavy–ball–based algorithm for identifying unknown constant parameters. For the ideal case, the adaptive observer estimates the actual values of both state and parameter vectors with an exponential convergence rate, possessesing the input–to–state stability property with respect to bounded external disturbances. The closed–loop stability analysis can be performed using a Lyapunov function approach under conventional conditions on the system’s persistence of excitation. The effectiveness of the proposed estimation algorithm is demonstrated through simulation results, which highlight an improvement in the accuracy compared to a conventional adaptive observer.

Keywords: Nonlinear adaptive control, Neural and fuzzy adaptive control, Event-based control
Abstract: This paper presents an adaptive prescribed performance control scheme for uncertain systems. The certainty equivalence prescribed performance controller is formulated by leveraging the error transformation method. The radial basis function neural networks are employed to approximate the unstructured uncertainties. A batch identifier is established, which makes full use of historical excitation information to update the estimate of optimal weight vector when the regulation-triggered condition is met. The extended state observers are employed to compensate for the approximation errors and environmental noises. It is demonstrated that the prescribed performance is achieved and the stiff differential equation problem is mitigated.

Keywords: Nonlinear adaptive control
Abstract: Parameter convergence is crucial for improving the stability and robustness of adaptive robot control, and exploiting online data memory is a natural approach to enhancing parameter estimation. This paper proposes a directional forgetting-based composite learning robot control (DF-CLRC) method to achieve parameter convergence under a condition of interval excitation that is strictly weaker than persistent excitation. In the DF-CLRC, the forgetting rate is adjusted using a directional forgetting rate matrix, and the excitation time is updated based on a torque prediction error, thereby exploiting online data memory more effectively to achieve time-varying parameter estimation. Simulations on a seven-degrees-of-freedom robot have verified the performance of the proposed method.

Keywords: Teleoperation, Adaptive and adaptable automation, High-performance motion control systems
Abstract: Synchronization tracking control of teleoperation systems is crucial for ensuring reliable remote operation in complex environments. However, in practice, such systems often suffer from limited communication bandwidth and complex output constraints, while actuator faults may further degrade the system’s control performance and even lead to instability. To this end, we develop a dynamic memory event-based adaptive practical predefined-time fault-tolerant control scheme. First, to relax the strict assumption on initial conditions in existing constraint methods while simultaneously accommodating various types of output constraints, a virtual control term integrating a shifting function and unified transformation functions is constructed. Second, two novel adaptive update laws with predefined-time convergence and low complexity are developed to compensate the system's lumped uncertainties and to solve the unknown control gain issue induced by actuator faults, respectively. Unlike finite- or fixed-time control schemes, the convergence-time upper bound of the developed approach is independent of the system's initial conditions and can be tuned via a single parameter. Simulation results show that, for any initial conditions, the developed method can constrain the system output within the boundary constraints after a predefined time, while further reducing communication resources, even in the presence of actuator faults.

Keywords: Physics informed and grey box model identification, Active learning and experiment design, Diffusion process
Abstract: This article presents an inverse source problem for the 1D advection-diffusion equation. From noisy data collected by a network of fixed sensors, we compute both the initial condition and the parameters of a moving Gaussian source. The problem is formulated as a variational optimization framework, resulting in a state equation, an adjoint equation, and gradient expressions calculated via a discrete adjoint of an implicit upwind scheme. A BFGS algorithm is used to update all unknowns. Numerical tests, including a fixed source case with optimal sensor placement, demonstrate the reconstruction accuracy and the essential role of sensor location.

Keywords: Physics informed and grey box model identification, Diffusion process, Machine and deep learning for system identification
Abstract: The purpose of this paper is to improve physics-informed neural networks (PINNs) algorithm for solving forward and inverse problems of nonlinear fractional-order diffusion partial differential equations (PDEs). Toward this aim, we first utilize the finite difference L_1 method on non-uniform meshes to embed fractional-order derivative into PINNs for achieving a higher accuracy, while making up the drawback that automatic differentiation is invalid to fractional-order derivatives. Solving algorithms via the improved PINNs for forward and inverse problems of nonlinear fractional-order diffusion PDEs are then presented. Subsequently, we take fractional-order Fisher equation as an examples to verify the effectiveness and robustness of our obtained results. These numerical results allow us to see that our improved PINNs can efficiently deal with the forward and inverse problems of nonlinear fractional-order diffusion PDEs.

Keywords: Physics informed and grey box model identification, Filtering and smoothing, Learning methods for control
Abstract: We revisit the problem of physics-informed regression, and propose a method that directly computes the state at the prediction point, simultaneously with the derivative and curvature information of the existing samples. We frame each prediction as a constrained optimisation problem, leveraging multivariate Taylor series expansions and explicitly enforcing physical laws. Such an approach makes the role of physical assumptions transparent: the governing equations enter the optimisation problem through equality constraints on the relevant differential quantities, rather than absorbed into physics-loss minimisation terms. Our comparative benchmarks on a reaction–diffusion system demonstrate predictive accuracy that is on par with a neural network–based approach, while exchanging the requirement for a single training phase for the execution of separate prediction computations.

Keywords: Physics informed and grey box model identification, Iterative and repetitive learning control, Distributed control and estimation
Abstract: Modeling distributed dynamical systems governed by hyperbolic partial differential equations (PDEs) remains challenging due to discontinuities and shocks that hinder convergence of traditional physics-informed neural networks (PINNs). The recently proposed vanishing stacked residual PINN (VSR-PINN) embeds a vanishing-viscosity mechanism within stacked residual refinements, enabling a smooth transition from the parabolic to hyperbolic regime. This paper integrates three curriculum-learning methods into VSR-PINN: primal-dual (PD) optimization, causality progression, and adaptive sampling. The PD strategy balances physics and data losses, the causality scheme unlocks deeper stacks by respecting temporal and gradient evolution, and adaptive sampling targets high residuals. Numerical experiments on traffic reconstruction confirm that enforcing causality systematically reduces the median point-wise MSE and its variability across runs, yielding improvements of nearly one order of magnitude over non-causal training in both the baseline and PD variants.

Keywords: Physics informed and grey box model identification, Linear system identification
Abstract: Many models of physical systems, such as mechanical and electrical networks, exhibit algebraic constraints that arise from subsystem interconnections and underlying physical laws. Such systems are commonly formulated as differential-algebraic equations (DAEs), which describe both the dynamic evolution of system states and the algebraic relations that must hold among them. Within this class, port-Hamiltonian differential-algebraic equations (pH-DAEs) offer a structured, energy-based representation that preserves interconnection and passivity properties. This work introduces a data-driven identification method that combines port-Hamiltonian neural networks (pHNNs) with a differential-algebraic solver to model such constrained systems directly from noisy input–output data. The approach preserves the passivity and interconnection structure of port-Hamiltonian systems while employing a backward Euler discretization with Newton’s method to solve the coupled differential and algebraic equations consistently. The performance of the proposed approach is demonstrated on a DC power network, where the identified model accurately captures system behaviour and maintains errors proportional to the noise amplitude, while providing reliable parameter estimates.

Keywords: Physics informed and grey box model identification, Machine and deep learning for system identification, Nonlinear system identification
Abstract: State-space models are traditionally based on physical knowledge, but multi-step predictions from these physical models can be poor due to model inaccuracy. Black-box deep learning has shown promise as an alternative. However, these methods rely on the availability of large datasets and potentially available physical knowledge is neglected. We propose the PG-RSSNN, a physics-guided recurrent state-space neural network that incorporates recurrent structures to enable the use of non-saturating activation functions in multi-step prediction. It mitigates the vanishing gradients and eliminates the risk of numerical divergence in training seen in existing structures that feed back state estimates. Results across multiple systems with various physical model imperfections, from linear state-space models with Gaussian noise to a robotic arm and a cascaded water tank system, show that the proposed PG-RSSNN maintains stable training behavior, and improves multi-step predictions, as compared with black-box neural networks and physics-only models, even with limited training data and when physical models are only partially known.

Keywords: Iterative and repetitive learning control
Abstract: An inherent assumption of perfect tracking in iterative learning control (ILC) is that there exists an ILC input such that the generated output can track the desired trajectory reference. This assumption may fail in practice, which gives rise to desired but untrackable tasks. This paper gives an end-to-end ILC design for repetitive untrackable tasks in closed-loop systems. The reference input is trial-to-trial updated together with the ILC feedforward input based on the measurement data. This two-player behavior of the closed-loop ILC system is investigated from a cooperative game perspective. A sufficient condition for the two-player end-to-end ILC to have a lower cost than the one-player norm optimal ILC (NOILC) is discovered. Finally, a numerical example is given to verify the effectiveness of the developed method.

Keywords: Iterative and repetitive learning control, Learning methods for control
Abstract: In many common industrial applications, Iterative Learning Control is a suitable technique to achieve accurate tracking of a reference trajectory. Using the principle of repeated attempts at a task, the previous behaviour can be used to generate an improved control sequence that achieves increasing tracking accuracy over these multiple trials. In our previous work Hobson et al. (2025) we take inspiration from the structure and performance of biological motion control systems (sensorimotor systems), leading to a design that learns to generate control signals using a linear combination of parametric ‘modules’. In Hobson et al. (2025) we assume that the primitive functions (forming the ‘modules’) are known in advance. However, the biological systems that provide the inspiration appear to be more flexible, with the basis functions themselves changing during learning and also being learnt. In this flexible setting, not all of these modules are necessarily known a priori and it is not possible to apply existing methods because they require that these modules be pre-defined and fixed. In this paper, we propose a flexible algorithm that permits modification of these modules between trials and provide convergence guarantees. We then demonstrate improved learning performance using a numerical example where a low-fidelity model is learnt quickly and then refined to increase accuracy while maintaining superior convergence speed.

Keywords: Iterative and repetitive learning control
Abstract: Norm Optimal Iterative Learning Control is an established optimization-based, iterative methodology for constructing inputs for a linear system that generates, exactly, a desired reference signal. It is designed for applications that use a combination of offline, model-based computation with plant operation to generate tracking data to assess the progress of the iterations. It has well defined convergence and robustness properties to linear modelling errors. This paper provides an analysis of the robustness of the algorithm to nonlinear input characteristics, producing a simple algebraic test relating robustness to parameters that characterize the nonlinearity, the plant model and the optimization process. This is a problem with significant practical importance but there is little understanding in the literature. The analysis uses an operator theoretical methodology in Hilbert space. This gives the results great generality. In particular, they cover continuous and discrete state space tracking, end-point and intermediate point problems.

Keywords: Iterative and repetitive learning control
Abstract: A coarse-fine system, consisting of coarse and fine systems, is widely employed in industries that require long stroke, yet fast and precise motion. In this paper, we propose a framework that jointly learns the feedforward signal and the reallocation of reference signals with explicit consideration of output constraints via Iterative Learning Control. The proposed method minimizes the overall tracking error by cooperatively compensating for tracking errors in the fine and coarse systems through coordinated task allocation between the two systems. The effectiveness of the proposed method is demonstrated through simulations on an output-constrained coarse-fine system, with comparison against existing approaches.

Keywords: Iterative and repetitive learning control
Abstract: The paper develops an iterative learning control law for uncertain linear differential systems with actuator saturation. A new 2D systems-based design is developed by applying gradient optimization and vector Lyapunov functions from the stability theory of repetitive processes. The resulting control law provides improved performance compared with existing ones. An example demonstrates the effectiveness of the new design.

Keywords: Iterative and repetitive learning control, Filtering and smoothing, Learning methods for control
Abstract: Tollmien-Schlichting waves are flowfield instabilities that can eventually evolve into turbulence, which can degrade aircraft performance. This paper presents a repetitive control approach for the attenuation of Tollmien–Schlichting (TS) waves through active flow control. Since the TS-wave behavior depends on flow conditions that may change over time, we pair the repetitive controller with an on-line period detection algorithm. This results in a repetitive controller where the internal model is time-varying. Therefore, we derive sufficient conditions under which the controller remains stable, for time-varying or incorrectly detected periods. For validation, high fidelity CFD simulations are conducted wherein TS waves are artificially generated by an upstream actuator and canceled downstream through the control scheme. The disturbance attenuation performance of the proposed solution is benchmarked against a state-of-the-art strategy, namely a Filtered-x Least Mean Squares (FxLMS) controller. The repetitive controller is shown to deliver comparable performance to FxLMS, while eliminating the need for a second upstream reference sensor (which is necessary for the FxLMS). At least 92% amplitude attenuation of both single-frequency and periodic TS waves is demonstrated in the high-fidelity simulation environment.

Keywords: Model predictive control, Control of hybrid systems, Applications of optimal control
Abstract: We employ model predictive control to reduce the energy cost of a building by optimizing electrical and thermal energy flows jointly over a horizon. The discrete-controllable heat pump possesses nonlinear behavior that is modeled by a first-order approximation, resulting in overall bilinear system dynamics. The resulting optimal control problem is then represented as an equivalent mixed-integer linear program. Additionally, a linear-continuous relaxation is incorporated to enable long-horizon optimizations. A simulative case study demonstrates the high potential of the proposed approach, yielding performance comparable to the theoretical optimum while being mindful of the computational effort.

Keywords: Model predictive control, Applications of optimal control, Learning methods for optimal control
Abstract: Optimization-based methods are essential for enforcing state constraints and ensuring flight safety in unmanned aerial vehicles (UAVs). However, predictive optimization techniques that explicitly handle multivariable constraints often suffer from high computational cost due to the nonlinear and strongly coupled UAV dynamics. This paper proposes an event-triggered reference governor (ET-RG) for UAV control that enforces constraints within an optimal control framework. To balance performance and efficiency, a reinforcement learning (RL)-based triggering mechanism is introduced to activate the RG adaptively only when necessary. Simulations on an autonomous quadrotor hovering task demonstrate that the proposed ET-RG achieves constraint-compliant tracking while significantly reducing computational burden compared to conventional reference governors.

Keywords: Applications of optimal control, Differential or dynamic games, Adaptive control design
Abstract: An insider is defined as a team member who covertly deviates from the team’s optimal collaborative control strategy in pursuit of a private objective, while maintaining an outward appearance of cooperation. Such insider threats can severely undermine cooperative systems: subtle deviations may degrade collective performance, jeopardize mission success, and compromise operational safety. This paper presents a comprehensive framework for identifying and mitigating insider threats in cooperative control settings. We introduce an insider-aware, game-theoretic formulation in which the insider’s hidden intention is parameterized, allowing the threat identification task to be reformulated as a parameter estimation problem. To address this challenge, we employ an online indirect dual adaptive control approach that simultaneously infers the insider’s control strategy and counteracts its negative influence. By injecting properly designed probing signals, the resulting mitigation policy asymptotically recovers the nominal optimal control law -- one that would be achieved under full knowledge of the insider’s objective. Simulation results validate the effectiveness of the proposed identification–mitigation framework and illustrate its capability to preserve team performance even in the presence of covert adversarial behavior.

Keywords: Applications of optimal control, Differential or dynamic games
Abstract: We study orbital pursuit and evasion games in which each spacecraft operates with a misspecified model of relative motion and updates it from noisy measurements. We introduce a Berk Nash framework in which each player designs a linear quadratic Gaussian controller for a subjective model and adjusts its parameters by minimizing an innovation based Kullback–Leibler divergence. We establish existence of equilibrium and implement a two sided learning scheme. A three dimensional Clohessy–Wiltshire case study demonstrates convergence toward pseudo true parameters and closed loop behavior that closely matches the classical zero sum benchmark, offering a principled representation of learning and bounded rationality in space engagements.

Keywords: Applications of optimal control, Control barrier functions and state space constraints
Abstract: A new methodology for solving the Dynamic Economic Dispatch problem, based on the use of slack variables, is proposed. The approach provides a systematic procedure for transforming the Dynamic Economic Dispatch problem into an extended unconstrained optimal control problem, the solutions of which directly yield solutions to the original problem. A key advantage of the proposed method is its ability to handle an arbitrary number of state constraints. Two simple examples illustrate the theory.

Keywords: Model predictive control, Real-time optimal control, Applications of optimal control
Abstract: Model predictive control (MPC) has attracted considerable attention in robotic systems due to its ability to explicitly handle state and input constraints. However, MPC performance depends on model accuracy, and the need to repeatedly solve an optimal control problem (OCP) often prevents fast robotic systems from obtaining the optimal control sequence in time. To address this issue, this paper proposes an online learning-based predictive-triggered MPC approach. This approach first integrates the predictions obtained from online learning into the MPC cost function, thereby improving the control performance of the robotic system. Then, a parallel computation mechanism is employed to advance the OCP solution ahead of the control-update instant, thereby relaxing the implicit assumption that the OCP can be solved instantaneously in standard MPC. Importantly, sufficient conditions are derived to guarantee algorithmic feasibility and closed-loop stability of the robotic system. Finally, simulation studies demonstrate the effectiveness of the proposed approach.

Keywords: Soft computing and robust intelligent control
Abstract: This paper proposes a methodological framework for tuning multivariable controllers and globally evaluating their performance using a multiobjective optimization approach. The methodology was applied to a multivariable system with two inputs and two outputs, in which Proportional-Integral (PI) and Dynamic Matrix Controllers (DMC) were tuned. Two analysis scenarios were established, each revealing relevant information to a designer regarding the choice of controller type and/or control loops appropriate for stabilizing the system. The DMC controller exhibits superior performance compared to the diagonal and off-diagonal PI controllers evaluated in this paper and is considered a reference controller. This comparison enables a designer to analyze the trade-off between the simplicity of implementation and performance in detail, highlighting scenarios in which a PI design concept may be more suitable than a more complex predictive solution proposed by a DMC design concept.

Keywords: Soft computing and robust intelligent control
Abstract: This work investigates two alternative models—a detailed and a simplified one—for a hybrid solar-battery inverter system using a multi-objective optimization approach. Both models are identified and validated with real measurement data. To mitigate outlier effects, the objective function minimizes the mean error after removing extreme values. Results show both models accurately capture system behavior, with mean errors below 20W for battery power and 2% for SOC. Slightly higher SOC errors appear in the simplified model when prioritizing battery performance. The results highlight the effectiveness of multi-objective optimization for comparing model structures and assessing trade-offs.

Keywords: Bio-inspired algorithms and optimization-based control, Fuzzy and neural systems in control
Abstract: This work presents a comparative analysis between a Takagi–Sugeno type-1 fuzzy controller in a MIMO configuration and the classical PI controller tuned using the BLT methodology. The study focuses on three well-known benchmark plants: Wood–Berry, Tyreus Stabilizer, and Wardle & Wood, all characterized by strong loop interactions. The tuning of the fuzzy controller is carried out through a multiobjective optimization framework, which simultaneously minimizes the Integral of Absolute Error (IAE) and the Total Variation (TV). Eleven independent optimization runs are performed per plant, and the best-performing solution is selected according to the median hypervolume metric. The results demonstrate favorable tracking performance and reduced control effort when compared with the BLT-based PI controllers. Pareto fronts, trajectory-tracking responses, and control-signal profiles are presented for the median-hypervolume solutions, highlighting the improved trade-off achieved by the proposed fuzzy strategy.

Keywords: Bio-inspired algorithms and optimization-based control, Soft computing and robust intelligent control
Abstract: The Proportional–Integral–Derivative (PID) controller remains widely adopted in industry due to its simplicity and robustness, yet its performance may degrade in high-order or strongly coupled two-input two-output (TITO) processes. To address these limitations, the Proportional–Integral–Derivative Acceleration (PIDA) controller extends the PID structure by adding an acceleration term. This work proposes a sequential multi-objective optimization framework for tuning the PIDA controller. By progressively refining the Pareto front, the method enhances convergence and improves decision making. Results demonstrate that the optimized PIDA configuration achieves superior trade-offs between tracking performance and control effort when compared with alternative controller designs.

Keywords: Bio-inspired algorithms and optimization-based control
Abstract: This paper investigates the role of multimodal optimization in multivariable PI controller tuning and proposes a pipeline that integrates multi-objective and multimodal search strategies. In multivariable control, the simultaneous optimization of several performance and control-action objectives often leads to many-objective formulations, which complicate convergence and diversity preservation in evolutionary algorithms. To address this issue, we employ an aggregation strategy to reduce the dimensionality of the objective space while preserving interpretability for decision-making. Experimental results demonstrate that the reduced two-objective formulation yields comparable hypervolume performance to the full six-objective problem. Additionally, the analysis reveals that the aggregated Pareto front exhibits multimodality: several distinct decision vectors produce equivalent performance levels. By applying an emph{a posteriori} multimodal search around preferred Pareto-optimal solutions, we show that these alternatives can be systematically identified and visualized. This broadens the multi-criteria decision-making process, allowing the engineer to explore structurally different controller configurations that deliver similar trade-offs between performance and control effort.

Keywords: Model driven engineering of control systems
Abstract: Unstable systems have long attracted significant interest within the automatic control community. Their particular characteristics often demand greater attention during tuning compared to stable systems. The methodology followed in this work consisted of deriving several PI controllers for an unstable system and evaluating them in terms of stability, robustness, and performance before applying the multi-objective Promethee framework. Each tuning criterion can yield its own best solution, yet these results may still be enhanced through very specific parameter variations.

Keywords: Control of distributed parameter systems, Boundary control of distributed parameter systems, Linear systems
Abstract: In this paper, we consider dynamic feedback stabilization of combustion oscillations in a Rijke tube. The combustion oscillations in the Rijke tube are modeled by the linearized Euler equations of gas dynamics, with the heat release rate described as a pointwise source term governed by an ordinary differential equation. Based on the extended state observer (ESO) in active disturbance rejection control (ADRC), finite-dimensional dynamic boundary and pointwise controllers are designed to suppress combustion oscillations. The stability analysis of the closed-loop system consists of two parts: first, the Nyquist criterion is applied to confirm that the real parts of all closed-loop eigenvalues are negative under specific feedback gains; second, operator semigroup theory combined with the Riesz basis approach demonstrates the validity of the spectrum-determined growth condition, thereby establishing exponential stability of the closed-loop system. The Nyquist plot is also utilized to develop a methodology for computing the stabilizing parameter region of the controller. Numerical simulations are conducted to validate the effectiveness of the proposed control design.

Keywords: Control of distributed parameter systems, Model reduction of distributed parameter systems, System identification and adaptive control of distributed parameter systems
Abstract: As global battery demand increases, real-time process control becomes increasingly important for battery electrode manufacturing, yet slot-die lines are still mostly manually operated in open loop. This paper develops a physics-based modelling-and-control pipeline for film-thickness regulation. Computational fluid dynamics (CFD) simulations provide the data from which a low-order cross-directional model is identified and calibrated. Numerical simulations demonstrate close agreement between the CFD and the cross-directional model, which is used to design a controller that can be used in both real-time, automated feedback operation and manual feedforward operation during line commissioning.

Keywords: Integration of ML/AI for control of DPS, Systems theoretic properties of distributed parameter systems, Optimization-based estimation and control
Abstract: This manuscript proposes a novel transfer learning-based moving horizon estimation method for state/output estimation of an infinite-dimensional system (i.e., target system) with limited output measurements, by leveraging the output measurements from another infinite-dimensional system (i.e., source system). This study is motivated by the fact that practical applications often suffer from data unavailability issues due to sensor hardware failure and/or harsh operating conditions. Practical applications such as cooperative sensing in parallel pipelines/reactors and autonomous vehicles indeed provide natural settings in which data from a twin system can compensate for insufficient monitoring data in another system. This manuscript leverages the idea of transfer learning to borrow the monitoring data from the source system for the state estimation of the target system. In particular, a transfer moving horizon estimation algorithm is proposed for output estimation in the presence of process and measurement disturbances as well as inequality hard constraints, and its stability analysis is further provided. The proposed method is demonstrated on a Schrödinger equation example through numerical simulation.

Keywords: Infinite-dimensional multi-agent systems and networks, Boundary control of distributed parameter systems, Observer design
Abstract: This paper investigates the consensus problem for multi-agent systems modeled by reaction-diffusion partial differential equations. Considering the practical constraints that complete state information is often unavailable and that control inputs can only be applied at the boundary of the spatial domain, a boundary control strategy based on partial spatial domain measurements is proposed. First, to address the issue of unmeasurable states, an observer is designed using measurement information from piecewise spatial subdomains or discrete points. Subsequently, a distributed boundary control protocol is developed based on the estimated states to ensure consensus. By employing the Lyapunov direct method and Poincaré-Wirtinger inequalities, sufficient conditions for the asymptotic stability of the closed-loop error system are derived and formulated as linear matrix inequalities. Finally, the effectiveness of the proposed method is validated through numerical simulations.

Keywords: Distributed parameters port Hamiltonian systems, Observer design, Passivity-based control
Abstract: This paper applies an early lumped observer‐based state‐feedback (OBSF) control design methodology, originally developed for one‐dimensional (1D) boundary‐controlled port‐Hamiltonian systems, to a two‐dimensional (2D) boundary‐controlled Mindlin plate. To this end, the 2D port‐Hamiltonian Mindlin plate model is first introduced and then discretized using a structure‐preserving finite‐difference method on staggered grids. A controllability decomposition is subsequently applied to identify the controllable modes of the discretized model. Furthermore, the state-feedback and observer gains are designed so that the OBSF controller is strictly positive real. This guarantees the stability of the closed-loop system when the finite-dimensional OBSF controller is interconnected with the 2D boundary-controlled Mindlin plate. Numerical simulations are finally presented to illustrate the effectiveness of the proposed method.

Keywords: Observer design, Control of distributed parameter systems, Analytic design
Abstract: The aim of this paper is to deal with the state estimation problem of a class of two-dimensional (2-D) semilinear distributed parameter systems governed by parabolic partial differential equations (PDEs) with space-dependent diffusivity. For this, the boundary collocated outputs, which only measure the boundary linear information of the studied PDEs, are considered. We then propose a Luenberger-type PDE observer and investigate exponential stability of the resulting observer error system by utilizing Lyapunov stability analysis theory and the linear matrix inequalities (LMIs). Simulation results are finally presented to verify the effectiveness of our methods.

Keywords: Data-driven and AI-based modelling of production and logistics, Simulation and optimization in production, operations and services, Production and operations management
Abstract: The Sales and Operations Planning process faces high uncertainties and complex supply–demand interactions, complicating decision-making. We propose a Reinforcement Learning-based multi-agent simulation framework, validated by domain experts, to model and simulate realistic market behaviors, supply dynamics, and production processes. A Soft Actor-Critic agent learns adaptive policies, optimizing production and supply decisions under stochastic conditions. Experiments show that the Reinforcement Learning agent effectively responds to fluctuations in both demand and supply. This framework supports robust scenario planning and enables automated policy generation, demonstrating the potential of Reinforcement Learning to enhance decision-making in complex Sales and Operations Planning environments.

Keywords: Data-driven and AI-based modelling of production and logistics, Simulation and optimization in production, operations and services, Production and operations management
Abstract: Developing discrete-event simulation (DES) models is resource-intensive and requires significant domain expertise. However, instantiating model structure from existing 2D layouts remains largely manual and non-value-added. We propose a hybrid multi-agent system leveraging semantic parsing to automatically digitize static layouts. The architecture integrates computer vision for component identification and Large Language Models (LLMs) for semantic context extraction, including material flow pathways and operational parameters. Deterministic agents structure this information into Core Manufacturing Simulation Data (CMSD) XML schema, eliminating manual layout replication. Such automated generation capabilities are particularly valuable for B2B digital industrial platforms serving small and medium enterprises (SMEs), enabling rapid supply chain scenario evaluation and reconfiguration during demand fluctuations and disruptions without requiring deep simulation expertise.

Keywords: Manufacturing plant simulation, control and optimization, Cyber-physical production systems
Abstract: The increasing diversity of autonomous mobile robots (AMRs) in industrial environments creates significant challenges regarding interoperability. While the VDA 5050 standard addresses operational interoperability by defining a common communication interface between AMRs and fleet management systems (FMS), its application in the planning and commissioning phases remains underexplored. This paper proposes a virtual commissioning (VCOM) architecture designed to streamline the integration of VDA 5050-compliant AMRs using material flow simulation. It presents a generalized system architecture that interfaces simulation software with an FMS, alongside a practical middleware-based solution for simulation environments lacking native MQTT support. The approach is validated through a demonstrator integrating Bosch Rexroth’s Active Shuttle Management System with Visual Components software via Node-RED. Results indicate that this architecture facilitates the early verification of control logic, the identification of deadlock scenarios, and the acceleration of the validation process through simulation running faster than real time.

Keywords: Simulation and optimization in production, operations and services, Smart production and logistics in manufacturing, Industry X.0 for production and logistics
Abstract: Digital twins are gaining increasing importance in production and logistics, particularly when simulation-based models are employed for decision support. Despite a wide range of existing approaches, a systematic categorisation of the possible forms of synchronisation between the physical system and its virtual counterpart – an essential characteristic of digital twins – has so far been lacking. This paper addresses this gap by developing a classification of relevant aspects of synchronisation and illustrating it through exemplary application cases. The paper demonstrates how this structuring supports the conceptual design of appropriate synchronisation mechanisms and highlights the design elements that must be considered when implementing simulation-based digital twins in industrial environments. The proposed morphology is intended as a starting point for further scientific discussion

Keywords: Supply chain and logistics engineering, simulation and optimization, Simulation and optimization in production, operations and services, Logistics and warehouse management
Abstract: Reliable transportation planning is a cornerstone of industrial logistics, particularly for Just-In-Time (JIT) operations and strict service level agreements. However, simulation-based planning tools often suffer from a ``realism gap'' due to poorly calibrated background traffic, leading to overly optimistic schedules and operational failures. This paper bridges this gap by introducing the Score-based Adaptive Recharge Algorithm (SARA), a data-driven method that calibrates microscopic traffic simulations against real-world sensor data. Applied to a microscopic simulation model of Nantes, France, SARA reveals severe congestion dynamics, such as localized bottlenecks and time-dependent variability that naive models fail to capture. By grounding the simulation in real-world sensor data provided by Nantes Métropole Open-Data API, it is demonstrated that uncalibrated models underestimate delivery times by up to SI{40}{percent} during peak hours, rendering them unsuitable for robust industrial planning. In summary, this comprehensive end-to-end framework provides logistics planners with the realistic ground truth required to optimize Time-Dependent Vehicle Routing Problems and establish operational supply chain reliability.

Keywords: Supply chain and logistics engineering, simulation and optimization, Supply chain management in manufacturing, Production and operations management
Abstract: Supply chains have faced significant disruptions due to unforeseen events like the COVID-19 pandemic, semiconductor shortages, and natural disasters, highlighting the need for enhanced supply chain resilience. This study investigates the role of additive manufacturing (AM) in improving the resilience of spare parts supply chains (SPSC), especially during low-frequency, high-impact events. While existing research has explored AM's impact on specific supply chain (e.g. medical), there is a lack of quantitative studies focusing on enhancing SPSC resilience. To address this gap, we conducted a simulation study to evaluate the benefits of insourced AM for spare part production in terms of supply chain resilience under different disruption scenarios. The results indicate that incorporating AM into SPSC can significantly enhance their resilience by ensuring quick recovery and sustained high machinery availability while reducing management costs of up to 90.26%. This study adds simulation-based quantitative evidence on the resilience benefits of insourced AM in a modeled SPSC under low-frequency, high-impact (LFHI) supplier disruptions.

Keywords: Sustainable and circular supply chain and production, Sustainable and circular manufacturing systems, Data-driven and AI-based modelling of production and logistics
Abstract: The relationship between the implementation of Industry 4.0 and 5.0 (I4.0/I5.0) technologies in manufacturing enterprises and the enhancement of their sustainability levels presents a significant challenge today. The primary issue is understanding how the effects of production automation contribute to the achievement of the Sustainable Development Goals (SDGs) within manufacturing contexts and support the realization of their core principles. This study, based on empirical survey data collected from 200 European automotive manufacturing companies from western Poland, analyzes the impact of adopting I4.0/I5.0 technologies to automate manufacturing processes and increase the sustainability level of production. Specifically, our research aims to assess the effects of enhanced automation on changes in Key Performance Indicators (KPIs) related to the implementation of selected SDGs in production. The focus is on ensuring: sustainable water management (SDG 6), efficient energy use (SDG 7), sustainable consumption and production patterns (SDG 12), and actions to mitigate climate change (SDG 13). The study highlights that the implementation of Digital Twins and collaborative robots for process automation in the automotive industry contributes to achieving SDG12 and SDG13, particularly by reducing generated waste and decreasing gas emissions. Furthermore, the research results provide recommendations for managers regarding the expected impact of integrating automation into sustainable production (SP).

Keywords: Intelligent manufacturing systems, Industrial artificial intelligence, Data-driven and AI-based modelling of production and logistics
Abstract: This study investigates the use of supervised machine learning models to predict grinding‑induced quality defects in polymer flowmeter body manufacturing. The empirical data set comprises approximately 2,000 parts produced on an industrial grinding line and described by nine process parameters capturing the thermal state of the workpiece, the kinematics of the operation and the condition of the grinding tool (X1–X9). Four defect categories are considered—indentations and scratches (D1), cracks (D2), surface irregularities and waviness (D3) and dimensional deviations of critical features (D4)—which form the target variable in a multiclass classification task. Four classifiers are analysed: a bilayer neural network (BNN), a decision tree (DT), a medium neural network (MNN) and a fine Gaussian support vector machine (SVM). The models are compared using accuracy, error rate, macro‑averaged precision, recall and F1‑score, as well as confusion matrices and ROC curves. The SVM achieves the best overall performance (99.6% accuracy, 0.4% error rate), closely followed by the MNN (99.4% accuracy). To interpret the SVM decisions, a Shapley‑value analysis is performed. It reveals that surface temperature (X1) and grinding wheel speed (X8) are the most influential parameters, followed by cutting speed (X6), feed rate (X4) and grinding time (X9), while coolant type (X2) and cooling rate (X3) play a minor role. The results demonstrate that data‑driven models can provide reliable defect predictions and actionable insight into the key drivers of grinding quality.

Keywords: Sustainable and circular manufacturing systems, Smart production and logistics in manufacturing, Human-technology integration in manufacturing
Abstract: Improving energy efficiency in production is currently one of the key challenges in sustainable production (SP) from environmental, economic, and social perspectives. Both the parameters of production processes and the buildings in which production takes place impact energy efficiency, especially in the context of achieving the 7 Sustainable Development Goal (SDG7) aimed at ensuring access to sustainable and modern energy. The appropriate selection and integration of technologies for buildings and mechanical systems to save energy in production remain a gap in the field of SP. Therefore, this study aims to develop an approach consisting of: (1) parameters describing the selected production processes and environmental conditions, (2) indicators determining energy efficiency, (3) technologies and data acquisition, and (4) a rule-based approach. An innovative rule-based and data driven approach has been developed to simulate changes in energy management in production, based on the integration of alternative energy sources, variability in production processes, and infrastructure constraints. As a case study, a small and medium-sized enterprise (SME) from the metal industry has been selected. The research results enable the identification of scenarios for the necessary actions to reduce energy consumption in production. Furthermore, it demonstrates that the proposed approach (scenarios) is a useful tool for supporting decision-making by production managers in the context of introducing changes in work organization and energy resource management, with the aim of improving energy management efficiency in production and enhancing the level of SP.

Keywords: Logistics and warehouse management, Supply chain and logistics engineering, simulation and optimization, Simulation and optimization in production, operations and services
Abstract: Cold storage environments pose challenges for both workers and UAVs, as freezing temperatures accelerate battery depletion and reduce efficiency. This study introduces a three-dimensional Energy-Aware Drone Routing Problem that integrates UAVs with UGVs, using the latter as a mobile base for stocktaking. We compare Ant Colony Optimization (ACO) with a Traveling Salesperson Problem (TSP) approach implemented via Google OR-Tools. Results show TSP offers faster computation, while ACO reduces energy consumption by approximately 12%, making it the preferred method for cold storage operations. Future work will scale the model to larger warehouses and refine energy estimates using real-world data.

Keywords: Simulation and optimization in production, operations and services, Digital supply chain and production, Supply network dynamics and control
Abstract: This paper addresses the problem of fast feasibility assessment for service delivery missions under travel time uncertainty. A preliminary feasibility assessment of the service mission plan, which synchronizes transport operations and service activities within pre-determined time periods, taking into account travel time uncertainties caused by traffic jams, accidents or other disruptions, allows for the elimination of variants that require time-consuming calculations but do not guarantee implementation. To enable a formal feasibility evaluation, the paper introduces a graph-based declarative model of time-window structure, where each feasible mission corresponds to a graph coloring with a number of colors not exceeding the available vehicle fleet size. The analysis is extended to uncertain travel times modeled as intervals, leading to a set of graphs representing possible mission realizations. The proposed approach provides a necessary feasibility condition that can be verified efficiently before the full search of a service plan. Experimental results confirm the applicability of the method for preliminary feasibility assessment in dynamic service delivery environments subject to travel time uncertainty.

Keywords: Human-centered production and logistics, Viable and resilient supply chain and production
Abstract: Work-related musculoskeletal disorders in manual material handling (MMH) persist due to static-threshold safety systems that ignore individual variability. Industry 5.0 calls for human-centric, adaptive solutions. This paper presents a retrospective synthesis of empirically investigated Human Digital Twin (HDT) components developed across multiple studies. The Micro-Twin enables marker-less human action recognition; the Meso-Twin models individualized fatigue baselines; the Holographic Twin fuses physiological, biomechanical, and cognitive signals for real-time feedback; the Synthetic Bridge provides computational kinematic simulation; and the Distal Shadow enforces edge-based safety interlocks. Together, these modules advance ergonomic monitoring toward adaptive, integrated systems supporting workforce resilience and inclusive manufacturing.

Keywords: Decentralized and distributed control for large-scale systems, Large-scale complex systems, Interconnected dynamical systems
Abstract: This paper provides a scalable consensus condition for large-scale linear multi-agent systems subject to information-processing delays. The agents are homogeneous Single-Input–Single-Output systems and interact over a directed communication graph that contains a spanning tree. The consensus condition is given through a stability region in the complex plane of the Laplacian matrix eigenvalues. The existence of such a stable region is given in terms of a delay threshold, which is not a critical delay in the classical sense. Different from similar approaches in the literature, in which the critical delay has to be computed for every disagreement mode of the system, the delay threshold in this work is linked to the agents' dynamics and not to the network topology. As the number of agents does not affect the treatment, the consensus condition is scalable. Several numerical examples illustrate the result.

Keywords: Hierarchical control, Composite systems, Systems-of-systems
Abstract: We present a games-in-games architecture for cyber-physical systems in which a physical plant is regulated in continuous time, while a cyber defender and attacker steer the generator of a Markov jump mode process. The construction produces a finite-horizon piecewise deterministic Markov process (PDMP) whose inner layer is a continuous-time Isaacs game and whose outer layer is a stochastic game over mode distributions. We derive the associated coupled Hamilton-Jacobi-Isaacs (HJI) equations via a Dynkin argument, prove well-posedness under mild assumptions, and specialize the theory to linear-quadratic data, yielding mode-coupled Riccati equations and a monotonicity property with respect to cyber hardening. A voltage regulation study showcases that the game-enabled generator shaping reduces the compromised dwelling time without performance downgrading.

Keywords: Interconnected dynamical systems, Hierarchical control, Complex dynamic systems
Abstract: This paper introduces a novel solution concept termed the Hierarchical Group-Optimal Equilibrium (HGOE) to characterize the intergroup competition and intragroup cooperation in group-based congestion games. Specifically, we establish the existence of HGOEs by proving that such games inherently possess a potential structure. However, this potential structure may be disrupted when resource failures are introduced, making the search for HGOEs challenging. This motivates us to analyze how failure probabilities affect these equilibria and derive sufficient conditions under which the original HGOEs and weak HGOEs from the failure-free game remain preserved. Furthermore, an event-triggered control scheme is developed via an algebraic state-space approach, which guarantees global stabilization to a set of epsilon-HGOEs with time-optimal performance. The results have been applied to a resource allocation problem in network function virtualization.

Keywords: Large-scale complex systems
Abstract: The Euclidean Steiner tree problem, normally posed in two dimensions, seeks to connect a set of prescribed terminal nodes by placing additional nodes, known as Steiner points, with edges connecting such nodes either to another Steiner point or a terminal node, and with the placements minimising the sum of all the edge lengths of the associated tree. We consider a problem in which we start with a known solution to a Steiner tree problem, and the terminal positions are then perturbed. A first-order approximation theorem is established for efficiently updating the Steiner point positions to recover a Steiner tree solution after the perturbations to terminal nodes. A numerical example that uses a stepwise application for a large perturbation illustrates the effectiveness of our approach.

Keywords: Large-scale complex systems
Abstract: This article develops the logical matrix factorization technique to explore the robust stabilization of Boolean control networks (BCNs) subject to function perturbation. Firstly, the index set of factorized structure matrix is obtained, based on which, a size-reduced system is constructed which remains part of the transition information for the original BCN. Secondly, the equivalence of state-feedback stabilization (SFS) between the size-reduced system and the original BCN is derived. Then, the impact of function perturbation on the SFS of original BCN is converted into that of size-reduced system. Thirdly, a perturbed position index matrix is constructed for the size-reduced system, and some criteria are proposed for the robust SFS of BCNs subject to function perturbation. Finally, the validity of obtained results is supported by the model of lac operon in the Escherichia coli.

Keywords: Large-scale complex systems
Abstract: This paper investigates the global matrix-weighted finite-time bipartite consensus (FTBC) problem of multi-agent systems (MASs) subject to asymmetric saturation constraints, where zero saturation bounds are considered. Distributed control laws using matrix weights are proposed, and sufficient conditions to guarantee the global FTBC are derived. The theoretical analysis are validated by numerical examples.

Keywords: Cyber-physical production systems, Intelligent manufacturing systems, Industrial artificial intelligence
Abstract: In 2005, the world’s first smart factory was founded and then built in Kaiserslautern, Germany, as a PPP (public-private partnership). In close cooperation between industry, academia, and politics, a test bed for smart factory technologies was set up, which then became a blueprint for similar activities worldwide. This activity was presented in a plenary paper at the IFAC World Congress 2008 in Seoul Zuehlke (2008, 2010) and thus became internationally known. The preliminary work carried out here and then disseminated by IFAC ultimately led to the term Industry 4.0 in 2011. This marked the beginning of the triumphant advance of Industry 4.0 in all industrialized countries around the world. Twenty years have passed since then, so it is appropriate to take stock. The papers in this session provide an overview of the activities of smart factory technologies in different countries and in different fields of application.

Keywords: Intelligent manufacturing systems, Cyber-physical production systems, Production and operations management
Abstract: The SmartFactory Reference Architecture proposes software agents on top of interoperable, semantic descriptions of resources and products based on Asset Administration Shells and OPC UA. Using these standardized interfaces, agents can adapt to various production environments to plan, control, and monitor flexible processes. By integrating language models, an agent becomes a so-called AI agent, enabling natural language-based interaction with human operators. Within this work, a comprehensive Multi-Agent System (MAS) for flexible Industry 4.0 environments is presented. The MAS defines autonomous agents in a distributed data space for services, products, and resources, allowing these agents to coordinate tasks and allocate resources in a dynamic production environment.

Keywords: Intelligent manufacturing systems, Human-technology integration in manufacturing, Model-driven enterprise-system engineering
Abstract: Industry 4.0 systems increasingly rely on heterogeneous Digital Twin representations, creating challenges for synchronization, semantic interoperability, and human supervision across engineering and operational models. This paper proposes a conceptual architecture for Agentic Industry 4.0 systems in which specialized AI agents coordinate model synchronization across heterogeneous representations such as AutomationML, Asset Administration Shells, and simulation models. The architecture combines semantic resources, standardized interaction mechanisms, and human-in-the-loop supervision to support adaptable and traceable model updates. Large Language Models (LLMs) are positioned mainly as supporting components for natural interaction and context-sensitive mapping assistance, rather than as fully autonomous decision makers. A proof-of-concept implementation based on an industrial motor use case illustrates event-driven coordination, model access through MCP-based interfaces, and user-approved synchronization updates. The implementation demonstrates the feasibility of the architecture while also revealing current limitations regarding semantic complexity, robustness evaluation, and scalability.

Keywords: Data-driven and AI-based modelling of production and logistics
Abstract: Digital Supply Networks (DSNs) are rapidly evolving, driven by a variety of Industry 4.0 technologies, generating vast amounts of heterogeneous, distributed, and often sensitive data across suppliers, manufacturers, logistics providers, users, and many other participants. Yet, traditional, centralized Machine Learning (ML) approaches struggle to harness the opportunity embedded in this diverse data due to, e.g., privacy concerns, competitive barriers, and regulatory constraints. Federated Learning (FL) emerges as a promising solution to this obstacle, enabling collaborative model training without raw data sharing, thereby preserving privacy while enhancing predictive capabilities across decentralized supply networks. This paper provides a synthesis of recent advancements in FL for DSNs, examining how this technology tackles challenges such as data heterogeneity, information asymmetry, cross-border regulatory constraints to enable privacy-preserving collaboration, and data-driven decision making across decentralized supply networks. Furthermore, we discuss how FL can transform DSNs by fostering secure collaboration, enhancing scalability, and building resilience across geographically dispersed and operationally diverse partners. Finally, challenges of introducing FL in DSNs, including incentive alignment, personalization, model convergence under non-independent and identically distributed (non-IID) data, and future research directions toward adaptive, trustworthy, and sustainable DSNs are identified to guide future research and exploration in this domain.

Keywords: Manufacturing plant simulation, control and optimization, Manufacturing engineering and management, Large-scale complex systems
Abstract: Manufacturing systems face increasing challenges from uncertain regulatory and geopolitical landscapes, pressuring factories to continuously adapt operations and internal design to remain“future-proof” and satisfy demand. Workstation configuration can describe both design and operational aspects, indicating what can be produced and associated costs and time. In this paper, we present a system-wide configuration representation, formalised as a metric space. We propose a method to construct such a System Configuration Space (SCS) from data available to most legacy manufacturing systems. An illustrative example shows how the SCS can help decision-makers better understand a system’s configurations and how this information can be used by optimization and simulation methods to solve decision problems, such as capacity allocation and investment.

Keywords: Structural analysis/quantitative methods for FDI/FTC, Wind power, Data-driven methods for FDI/FTC
Abstract: This paper presents a hybrid fault diagnosis framework for a utility-scale 5 MW wind turbine (FAST-based model) that combines physics-guided structural analysis with data-driven interval estimation. Analytical Redundancy Relations (ARRs) are derived to generate residuals from measured/estimated variables. To provide robust, tight uncertainty bounds, an Optimized-Rectangular Gaussian Process Regression (OR-GPR) is introduced that learns asymmetric, input-dependent prediction intervals with target empirical coverage while minimizing mean width. Residuals are computed by comparing sensor measurements with OR-GPR predictions, and faults are detected using a combination of interval-based thresholds and Cumulative Sum (CUSUM) control charts. Experiments on representative scenarios show earlier detection and lower false alarms compared with conventional GPR intervals and fixed-threshold method, demonstrating the effectiveness of ARR-guided residuals with OR-GPR for wind turbine condition monitoring.

Keywords: Wind power, Control and management of energy systems
Abstract: Several works have proposed Nonlinear Model Predictive Control (NMPC) schemes using quasi-Linear Parameter Varying (qLPV) embeddings. Despite recent advances, many results simply rely on gain-scheduling (i.e. frozen predictions) or using the practical iterative implementation from (Cisneros et al., 2016). However, if the involved prediction uncertainties are not considered, the control becomes neither optimal nor robust. In this work, we exploit the recent idea introduced by K ̈ohler and Zeilinger (2025) as a pragmatic way to ensure input- to-state stability (ISS) and recursive feasibility in qLPV NMPC. In particular, we let the state prediction assume an artificial initial value, and then penalise the corresponding deviation to the real sampled state in the optimisation cost. We debate how this simple relaxation indeed enables ISS, despite the inherent prediction mismatch due to the unavailable scheduling trajectories. We also discuss direct extensions for when load disturbances are present and for the reference tracking case. The method is experimentally validated using a scaled wind turbine benchamrk.

Keywords: Wind power, Control and management of energy systems
Abstract: Wind turbulence is the dominant disturbance in wind turbine operation. Wind gusts can drive large fluctuations in generator speed and structural loads. Generator overspeed events can trigger wind turbine shutdowns, reducing energy production and increasing the cost of wind energy. Specific wind sequences, such as a lull followed by a sharp positive ramp, are particularly prone to causing overspeed peaks, especially in near- and above-rated operation. Gust-aware control strategies have been explored to detect such events and adjust the turbine operation accordingly. This paper presents an experimental validation of gust-aware control strategies in a wind tunnel using a scaled wind turbine and a state-of-the-art actively controlled wind grid. The baseline wind turbine controller is augmented with gust-measure-enabled rating rules that can (i) pre-emptively derate the turbine when the gust measures indicate an incoming gust, thus limiting generator overspeed, and (ii) boost the power set point when the gust measures indicate calmer, steadier conditions that allow the turbine to safely increase energy capture. Two gust measures are computed online from local inflow signals. Performance is evaluated under prescribed (a) repeating gust and (b) turbulent wind patterns. The tested controllers are blind to the inflow test conditions in all cases. The gust-aware controllers consistently execute anticipatory derating prior to gust arrivals, reducing generator speed peaks, while enabling systematic power boosts during steady wind intervals. Across both inflow families, the rating rules achieve the intended condition-dependent modulation of power, experimentally demonstrating the feasibility of gust-aware power management. Blade-load responses, however, are sensitive to decision thresholds and timing, indicating that careful tuning is needed to fully realize overspeed mitigation and energy gains while managing structural loads. The results support gust-measure-driven control as a potential pathway to improved operational resilience and opportunistic energy capture that warrants further exploration.

Keywords: Wind power, Control and management of energy systems
Abstract: This paper proposes a multi-objective control framework for wake-affected wind farms to manage the trade-off between power maximization and fatigue load minimization. The conflicting objectives are formulated using Nash Bargaining theory, providing a fair, Pareto-efficient solution without heuristic weight tuning. A Warm-started Proximal Alternating Direction Method of Multipliers (W-PADMM) algorithm is proposed to efficiently solve the bargaining problem, which embeds a learning-aided mechanism using a Long Short-Term Memory (LSTM) network to proactively guide the optimization. Case studies on a 9-turbine wind farm using historical operational data validate that the algorithm achieves a superior power-fatigue balance with significant gains in computational efficiency.

Keywords: Wind power, Control and management of energy systems, Process modeling, identification, and estimation techniques
Abstract: Modern wind turbines need high situational awareness for decision making and control scheduling. It allows for a trade-off between greedy power maximisation and further control objectives such as load alleviation or grid compliance. The paper formulates a nested Extended Kalman Filter that is able to precisely reconstruct and distinguish the inflow wind and load-relevant structural dynamics of a wind turbine. The formulation is directly motivated by the demands of field applicability, thus robust, low in computational costs and flexible with respect to sensor availability or calibration drifts. The estimator is field-tested on a utility-scale commercial turbine. It shows good agreement with independent reference measurements, namely a hub-mounted scanning lidar for the spatially resolved inflow wind field and camera-based digital image correlation for the structural movements.

Keywords: Wind power, Forecasting of power supply and demand
Abstract: To address the problem of complex spatiotemporal correlations between the target wind farm and neighboring wind farms, as well as the difficulty of fully characterizing spatial dependencies with a single graph structure in regional multi-wind-farm forecasting, this paper proposes a Chebyshev Graph Convolutional Network with Node Adaptation Long Short-Term Memory (ChebGCN-NA-LSTM) method for wind power forecasting. The proposed method first employs multi static graphs to characterize spatial relationships among wind farms. Based on these graph structures, ChebGCN is used to extract high-order spatial features, whereas a node-adaptive mechanism complements the static graphs by capturing latent spatial dependencies. Furthermore, LSTM is incorporated to model temporal dependencies. Finally, experimental results show that the proposed method can effectively improve the accuracy and robustness of target wind farm power forecasting.

Keywords: Modeling and control in mechanical ventilation, Digital twins in healthcare, model-based therapeutics
Abstract: Spontaneous efforts during mechanical ventilation significantly affect airway pressure and complicate the accurate assessment of respiratory mechanics. This study presents a novel non-invasive method based on parametric modelling and optimization that can directly reconstruct both baseline pressure and spontaneous effort from the airway pressure. The method was validated using real clinical data. Results shown robust performance under diverse conditions, and substantial reduction in variability of estimated resistance and elastance (R: 42% → 14%; E: 45% → 16%). This work establishes a foundation for intelligent, effort-aware mechanical ventilation monitoring.

Keywords: Biomedical system modeling, identification, and simulation, Medical devices, systems and solutions, Digital twins in healthcare, model-based therapeutics
Abstract: Chronic respiratory disease poses significant clinical and quality of life burden on a global scale. The ability to track and self-manage respiratory disease is limited by the pulmonary function testing tools available to provide a comprehensive, repeatable, and reliable respiratory assessment, outside of a specialist clinical setting. Outlined, is a proposed framework for monitoring lung health in chronic respiratory disease outside of clinical settings. This new comprehensive pulmonary function test, designed to elucidate simulated changes in key respiratory parameters, is investigated in this study using a mechanical test lung. The results of this paper demonstrate an ability to describe set changes in model terms, as well as informing continued development of hardware and planned clinical trial protocols. This paper provides a foundation for patient-specific automated respiratory disease tracking. Therefore, has the potential to improve patient care, adherence to therapy, and precision / patient-specificity of care.

Keywords: Medical devices, systems and solutions, Biomedical signal measurement and processing, Modeling and control in mechanical ventilation
Abstract: Access to continuous airway pressure (P) and flow (𝑉̇) waveforms is important for assessing respiratory mechanics during mechanical ventilation (MV) treatment. Yet, such data are often difficult to obtain due to ventilator-specific interfaces and closed communication protocols. This study proposes a compact, ventilator-independent pressure–flow sensor module (P𝑉̇ sensor module) integrated with a portable data acquisition system for capturing high-fidelity mechanical ventilation waveforms. The system was tested against a validated reference device using a mechanical test lung across multiple respiratory system elastance and resistance configurations in both volume and pressure control ventilation modes. Flow measurements from the sensor closely matched ventilator output with an absolute percentage error at <10%, while expected pressure differences were observed due to proximal placement. Respiratory mechanics estimated using the single-compartment model showed variable agreement with the reference system across conditions, with discrepancies increasing under more extreme mechanical loads. Overall, the proposed P𝑉̇ sensor module demonstrated reliable waveform capture and the ability to track trends in respiratory mechanics, supporting its potential use in continuous monitoring, research applications, and the development of advanced analytics such as digital-twin modelling and clinical decision support systems.

Keywords: Medical devices, systems and solutions, Biomedical signal measurement and processing, Clinical trial, clinical validation
Abstract: Venous blood oxygen saturation (SvO2) measurements are critical for monitoring oxygen use. Currently, accurate SvO2 measurement requires invasive blood sampling. Conventional peripheral pulse oximetry techniques cannot measure SvO2 because peripheral veins are not pulsatile. This study explores the potential for non-invasive SvO2 estimation using internal jugular vein (IJV) pulsations detected by a flexible sensor array, to thus exploit standard pulse oximetry. A prototype sensor was developed to detect pulsatile signals from the carotid artery and IJV on the right-hand side of the neck. The sensor's performance was evaluated in 6 healthy adult subjects with arterial oxygen saturation (SaO2) measured using a commercial pulse oximeter and SvO2 estimated using the detected IJV pulse and breathing modulations. These estimates were combined to determine the subjects' cerebral oxygen extraction ratio (O₂ER). The sensor reliably detected pulsations in both upright and supine positions, with an average SvO2 of 65.70-80.91% with standard deviations of 1.67% and 4.26% for the IJV pulse and breathing modulations respectively, corresponding to an O2ER of 0.18-0.33 (±0.03). Differences in IJV pulse and breathing modulation estimations, as well as between upright and supine positions, are likely due to the influence of crosstalk between the carotid artery and IJV. The proof-of-concept sensor demonstrated the ability to continuously and non-invasively monitor SvO2 and O₂ER, producing estimates within expected literature ranges. Further validation with a larger cohort and comparison to blood gas analysis is necessary to confirm the sensor's accuracy, along with methods to mitigate crosstalk between arteries and veins.

Keywords: Biomedical system modeling, identification, and simulation, Biomedical signal measurement and processing
Abstract: Current models for volumetric capnography (VCap) curve analysis fail to capture nonlinear phase III dynamics associated with dysfunctional breathing. Hysteresis loop analysis (HLA) addresses this error by decomposing a VCap curve into a minimal number of linear segments, capturing nonlinearities as they appear. This paper compares HLA against VCap analysis via Functional Approximation based on the Levenberg-Marquardt algorithm (FA-LMA) to identify airway dead space (VDAW) and the slope of phase III (SIII) using publicly available clinical data.

Both methods perform equally well on this clinical data. However, FA-LMA consistently overestimates SIII compared to HLA, suggesting different performance against non-linear phase IIIs in clinical breaths. Overall, HLA accurately identifies clinically relevant parameters from VCap curves in clinical data, and is well posed to adapt to atypical VCap curves associated with dysfunctional breathing. Full testing on a dedicated respiratory cohort is justified by these results.

Keywords: Machine learning and artificial intelligence in chemical process control, Model-predictive and optimization-based control in chemical processes, Advanced process control
Abstract: Capturing long-horizon economic effects in EMPC can require extended prediction horizons or detailed nonlinear models, increasing the online problem size and modeling effort. This paper proposes a reinforcement learning (RL)-assisted terminal cost approximation for QP-formulated EMPC, in which a low-rank quadratic terminal cost embeds long-horizon economic information while retaining a short prediction horizon and the QP structure. For a continuous stirred-tank reactor (CSTR) with economically optimal periodic operation, the proposed method recovers cyclic behavior, improves the discounted return by 21% over short-horizon EMPC, and reduces computation time by about 90% relative to a nonlinear EMPC benchmark.

Keywords: Model-predictive and optimization-based control in chemical processes, Process modeling, identification, and estimation techniques, Machine learning and artificial intelligence in chemical process control
Abstract: A prediction model that achieves the high performance required for model predictive control (MPC) can be obtained by system identification using high-quality data. Such high-quality data can be acquired from design of experiments (DoE), which reduces the confidence region of the model parameters by minimizing a geometric quantity of the region. However, such an index of the confidence region does not directly quantify the performance of MPC. This paper proposes MPC-oriented DoE that maximizes the expected setpoint-tracking performance of MPC with the prediction model estimated by the prediction error method. The proposed DoE index is calculated by Monte Carlo simulation of the closed-loop system, where a process with parameters sampled from the confidence region is controlled by the MPC. The proposed DoE index is minimized using Bayesian optimization. In the case study, the prediction model obtained using MPC-oriented DoE provided the control performance comparable to the best control performance of the prediction models for the existing three DoE methods, i.e., D-optimal, E-optimal, and A-optimal DoE.

Keywords: Machine learning and artificial intelligence in chemical process control, Advanced process control, Batch and semi-batch process control
Abstract: This work presents a surrogate-assisted reinforcement learning framework for optimal control of N-methyl-2-pyrrolidone (NMP) recovery by batch distillation in lithium-ion battery manufacturing. A first-principles dynamic model of a pilot-scale column is calibrated with plant data and approximated by a physics-constrained Neural ordinary differential equation surrogate. A Double Deep Q-Network agent uses this surrogate to learn reflux-ratio trajectories and batch termination decisions under a 99.99 wt% NMP purity constraint. Two reward formulations are considered: minimization of production cost and minimization of CO2-equivalent emissions, both evaluated at an industrial scale via techno-economic and life-cycle calculations. Compared with fixed reflux operation, the learned policies reduce specific production cost by up to 3.8% and specific CO2-equivalent emissions by up to 57%, demonstrating that reinforcement-learning-based control can simultaneously improve economic and environmental performance of solvent recycling processes.

Keywords: Machine learning and artificial intelligence in chemical process control, Advanced process control
Abstract: Electrified chemical processes are incentivized by exposure to time-varying electricity markets to operate flexibly, but participating in demand response schemes can require satisfying terminal constraints over long horizons. Specifically, terminal constraints may be required when computing optimal schedules in order to preserve dynamic stability. Model-based optimization methods are computationally costly, and data-driven scheduling via reinforcement learning (RL) faces severe credit-assignment challenges. We integrate Goal-Space Planning (GSP) with Deep Deterministic Policy Gradient (DDPG), using learned temporally abstract models over discrete subgoals to propagate value across extended horizons. Using a simulated air separation benchmark, we demonstrate the proposed approach improves sample efficiency over standard DDPG while satisfying terminal storage constraints, mitigating myopic control behavior.

Keywords: Energy management systems
Abstract: This study evaluates model predictive control (MPC), equivalent emissions minimization strategy (EEMS), and deep reinforcement learning (DRL) for energy management in an LNG hybrid-powered vessel using real sailing data. A modified EEMS with a hyperbolic tangent penalty is proposed to improve battery state of charge (SOC) regulation. A constrained-loss TD3 method is introduced, which enforces operational limits in the actor loss and requires no SOC model or power demand forecasting. Results show that the proposed EEMS and TD3 approaches achieve an emissions reduction performance comparable to MPC while maintaining healthy SOC levels. These findings highlight practical and scalable energy management systems for low-emission hybrid ship operation.

Keywords: Real-time optimization and control in chemical processes, Advanced process control, Industrial applications of process control
Abstract: Temperature control in large-sized monolithic silicon epitaxy equipment reactor is a significant challenge due to strong nonlinearities, time delays, and cross-coupling among multiple heating zones. Currently, ADRC-based multi-zone temperature control systems suffer from repeative parameter tuning and fixed controller parameters across zones, which prevents the controllers from responding promptly to setpoint changes or variations in nonlinear dynamics. This study proposes a radial basis function (RBF) neural network–based self-tuning Linear Active Disturbance Rejection Control (LADRC) multi-zone temperature control system. During processes, the RBF neural network performs real-time system identification, and a gradient descent algorithm is employed to adaptively adjust the LADRC parameters of all four zones. Simulation results demonstrate that the proposed self-tuning LADRC temperature control system achieves faster settling and rise times, as well as reduced overshoot. Finally, experiments confirm that the proposed system can deliver high-performance temperature control without manual parameter tuning.

Keywords: Monitoring, performance assessment, and fault detection in chemical process control
Abstract: Process monitoring of complex, multivariate systems presents a significant challenge when the process is operated at multiple operating conditions. In such cases, distinguishing a valid change in operating conditions from abnormal process deviations may be treated as a multimode process fault detection problem. Some methods to address this problem model the process using a multimode probability distribution. In practical applications, the true distribution, or indeed a good estimate of the same, may not be readily available to the user. Such inexact information on the probability distribution lead to poor fault detection performance that may further lead to process operations degradation. A distributionally robust design of fault detection systems is preferable in the face of ambiguous uncertainty. In this work, we propose a Bayesian fusion approach using Gaussian mixture models (GMM) and introduce a distributional uncertainty model based on optimal transport distance between GMMs. We formulate a worst-case performance evaluation problem and propose a detection threshold optimization algorithm. The effectiveness of the proposed method is demonstrated through a simulated example.

Keywords: Monitoring, performance assessment, and fault detection in chemical process control, Fault detection and isolation methods
Abstract: The traditional Mamba model has demonstrated its success in the field of fault detection. However, it focuses on temporal characteristics but omits spatial dependencies among multivariate variables. To fully utilize spatial information and enhance anomaly detection performance, this paper proposes an improved Mamba model, called the mutual information-guided spatio-temporal Mamba (MiST-Mamba). First, the method constructs a multi-scale mutual information-based convolutional encoder, which measures interdependencies among input variables through multiple mutual information matrices and further applies convolution operations to capture spatial features. The obtained spatial features are then fed into a spatio-temporal Mamba module, which employs a selective state-update mechanism to represent the long-term temporal dynamics of spatial features. Finally, a convolutional decoder is used to construct the reconstruction residuals, which serve as anomaly scores for anomaly detection. Experiments on the Tennessee Eastman benchmark demonstrate that the proposed method outperforms the traditional Mamba model in terms of fault detection performance.

Keywords: Monitoring, performance assessment, and fault detection in chemical process control, Process modeling, identification, and estimation techniques, Industrial applications of chemical process control
Abstract: In non-stationary industrial processes, fault variables often exhibit complex causal interactions across multiple time scales, posing significant challenges for accurate root cause diagnosis (RCD). To address this problem, this study proposes an RCD algorithm based on full time scales Granger causality analysis, enabling accurate root cause identification and comprehensive characterization of fault propagation mechanisms. Specifically, the original time series is decomposed into multi-scale intrinsic mode functions using multivariate variational mode decomposition, minimizing both frequency-domain bandwidth and reconstruction error. An error correction term is then incorporated into a structural equation model parameterized by multilayer perceptrons to mitigate the adverse effects of non-stationarity on causality identification. Instantaneous causal relationships inferred at the original time scale through variational inference are integrated with those inferred across multiple scales to construct a comprehensive causal diagram among fault variables. Experimental results from a real-world injection molding process demonstrate that the proposed algorithm accurately identifies root cause variables.

Keywords: Monitoring, performance assessment, and fault detection in chemical process control, Predictive maintenance and equipment condition monitoring, Machine learning and artificial intelligence in chemical process control
Abstract: Predictive maintenance has demonstrated significant efficacy in minimizing resource losses and maintaining production continuity in advanced manufacturing environments. However, traditional approaches face substantial challenges when applied to industrial IoT systems characterized by non-stationary time series with high-dimensional, multi-source, noisy, and spatiotemporally correlated sensor data. Key limitations include the inability to capture dynamic interdependencies due to independent channel assumptions, insufficient integration of domain expertise, and absence of uncertainty quantification, which collectively elevate operational risks. To address these challenges, this paper introduces a text-prompt-guided temporal diffusion transformer model incorporating static-dynamic graph learning for predictive maintenance in non-stationary industrial processes. The model’s primary contribution is its adaptive modeling of multivariate coupling relationships through a spatiotemporal interaction network. Through customized text prompts and modal alignment constraints, the conditional guidance mechanism coordinates extraction of long-term and short-term dependencies with uncertainty-aware prediction during the diffusion denoising process. This approach substantially enhances the reliability of long-horizon predictions for industrial applications.

Keywords: Process modeling, identification, and estimation techniques, Industrial applications of process control
Abstract: In situ monitoring of fluid viscosity is of high interest for efficient process control, but measuring the viscosity of fluids is challenging and expensive. Progressing cavity pumps serve as valuable data source, because they induce a mild pressure pulsation that contains information on the fluid viscosity. We present a model-based approach to estimating fluid viscosity using only the easy-to-measure pressure pulsation at the progressing cavity pump. We derive a lumped-parameter model that relates the pump dynamics to the fluid viscosity, and introduce a phase-locked-loop-enhanced extended Kalman filter to estimate the viscosity online. We validate our approach using measurement data from a laboratory test stand and demonstrate its effectiveness in monitoring fluid viscosity across a range of operating conditions.

Keywords: Biomedical and medical imaging, image processing, visualization, Biomedical signal measurement and processing, Medical devices, systems and solutions
Abstract: Electrical Impedance Tomography (EIT) seeks to recover internal conductivity distributions from boundary voltage measurements, but the inverse problem is severely ill-posed and highly sensitive to noise. This work investigates sinusoidal and hyperbolic–sinusoidal neural networks for domain-dependent 3D EIT reconstruction. The proposed formulation provides a representation capable of capturing both smooth and high-frequency spatial features. Results using a synthetic dataset containing 3D phantoms show that the hyperbolic–sinusoidal model offers sharper boundary recovery, more accurate conductivity contrast, and higher resilience to measurement noise compared to the standard sinusoidal model. These findings indicate that periodic-activation neural networks form a promising framework for high-resolution 3D EIT reconstruction.

Keywords: Digital twins in healthcare, model-based therapeutics
Abstract: Accurate three-dimensional (3D) modelling of the breast surface is one basis of emerging breast cancer diagnosis and risk assessment systems. Typical explicit geometric approaches, such as structured-light scanning and stereo vision, require complex multi-camera hardware and calibration, which limits easy, robust integration into routine clinical workflows. This study investigates an implicit differentiable rendering (IDR) framework for breast surface reconstruction and introduces a normal-based Laplacian regularization term tailored to the smooth geometry of soft tissue. The proposed method reconstructs a signed distance field of the breast from multi-view RGB images acquired with a single camera and optimises the surface using only image-level supervision, without ground-truth 3D meshes or precise calibration. Experiments on a breast phantom with three random initialisations show the Laplacian-regularized model (L-IDR) consistently outperforms the original IDR baseline, achieving lower Chamfer distance and equal or higher PSNR across all seeds. Error histograms, cumulative error distributions and boxplots further show L-IDR shifts the point-wise error distribution towards smaller values, shortens high-error tails, yielding smoother reconstructed surfaces. Ablation studies indicate a moderate Laplacian weight and Fourier feature resolution provide a favourable balance between geometric fidelity and regularisation. These results suggest Laplacian-regularized implicit differentiable rendering offers a low-cost, calibration-free and flexible solution for breast surface reconstruction, and a promising step towards lightweight, image-based 3D breast modelling methods to complement a range of emerging breast cancer diagnostics.

Keywords: Biomedical and medical imaging, image processing, visualization, Digital twins in healthcare, model-based therapeutics, Biomedical signal measurement and processing
Abstract: Electrical Impedance Tomography (EIT) is an emerging medical imaging modality. It is non-invasive and operates by injecting small amounts of current into the body and measuring the induced boundary voltages. These voltages identify an inverse problem that allows the estimation of internal body impedances. Due to the large ill-posed nature of the inverse problem and the physics associated with the induced electric field, inclusions with different geometries may appear similar in reconstructed images. This lack of resolution limits the clinical applicability of EIT. Inclusions with different vertical geometries were reconstructed using an open-source EIT device. Each inclusion was moved through 25 different positions in a saline phantom. The reconstructed images were processed by considering a 1D slice of impedance values and approximating it with a Gaussian function. The parameters of the Gaussian function were used to find patterns in type and position of the inclusion. When the 1D slice was coincident with the target the impedances had one peak and small amounts of ringing making it a good approximation (R^2≥0.93). The Gaussian parameters showed repeatable patterns as the inclusion was moved around the phantom. The amplitude was the only parameter that changed with inclusions. The mean changed based on the horizontal position of the inclusion, and the standard deviation was consistent between inclusions. This simplified method is unable to reliably distinguish between complex geometries due to an ambiguity between the impacts of volume and changes in vertical shape. However, the position and size of the inclusions can be recovered from the 1D slices. Thus, the same information about the inclusion can be determined from a 1D set of impedances as a 2D image.

Keywords: Biomedical and medical imaging, image processing, visualization, Medical devices, systems and solutions, Healthcare management, disease control, critical care
Abstract: Globally, 1 in 9 women experience breast cancer during their lifetime. Although mammography is the current large-scale screening modality, compliance is reduced due to radiation exposure, discomfort, and invasiveness. Digital Imaging Elasto-Tomography is an emerging radiation-free, non-invasive and portable screening technology designed to improve equity of access and outcomes for underserved populations. The surface motion response of the breast under a low-amplitude, mechanical vibration is measured with cameras and lasers. Based on the 400-1000% stiffness contrast between healthy and cancerous tissue, automated diagnostic methods analyse tissue properties and detect potential cancers. The capability of the detection algorithm is thus a key focus area to ensure high diagnostic accuracy.

Further development and validation of an existing stiffness-based diagnostic algorithm is presented for this technology. Hysteresis Loop Analysis (HLA) was used to estimate tissue stiffness. Improved phantom breasts that better mimic the boundary between healthy tissue and stiff inclusions, as well as clinical data, were used to test the optimised algorithm. The Empirical Cumulative Distribution Function (ECDF) of the HLA stiffness enabled quantitative observation of abnormal regions of tissue, through skew, asymmetry about the centre, and the proportion of motion points above an empirically determined threshold. Future research efforts will aim to improve motion reconstruction success to enable diagnostic testing of higher amplitude data, more extensive clinical data testing and algorithm optimisation. Conclusions from the ECDF stiffness analysis show promise for improving the diagnostic algorithm capability of this technology, ultimately enhancing potential to improve breast screening equity and outcomes for women.

Keywords: Biomedical and medical imaging, image processing, visualization, Medical devices, systems and solutions, Healthcare management, disease control, critical care
Abstract: For women worldwide, breast cancer has both the highest incidence and highest mortality of all cancers. Mammography is the current large scale screening modality. However, screening compliance is reduced by radiation exposure, discomfort and invasiveness. Digital Imaging Elasto-Tomography is an emerging radiation free, non-invasive and portable breast cancer screening technology aimed at improving access and outcomes for underserved populations. A low amplitude, steady-state, mechanical vibration is applied at the nipple to measure the surface motion response. With the 400-1000% contrast in stiffness between healthy and cancerous tissue, underlying tissue properties and potential cancers can be detected. The imaging system and calibration methods are thus a key stress area for ensuring accuracy and quality.

The full imaging system and methods of calibration are presented for this technology. Camera and laser calibrations were validated visually and through performance metrics such as RMS error and inter-beam angle. These calibrations were used to evaluate surface motion reconstruction success, through imaging of four silicone phantoms at various frequencies [25Hz, 30Hz, 35Hz, 40Hz, 45Hz] and amplitudes [1mm, 1.5mm, 2mm], producing a total of 60 data sets. Phantoms with stiff inclusions had reconstruction success rates of > 80%, compared to only 26.7% in the healthy phantom, likely due to the deteriorated surface affecting laser correspondence. Across the 19 failures, the most common was disagreement between the laser correspondence and surface model, driven by unpredictable surface and light interactions. Future research efforts will focus on improving laser calibration inaccuracies and analysing phantom geometry to better reflect human breast variability. A higher quality laser model and more representative phantom breasts should enhance motion reconstruction reliability, ultimately strengthening the potential of this technology to increase breast screening equity and outcomes for women.

Keywords: Medical devices, systems and solutions, Biomedical and medical imaging, image processing, visualization
Abstract: Electrical Impedance Tomography (EIT) is widely used for bedside lung monitoring, but clinical imaging is still dominated by differential reconstructions. This paper investigates an absolute EIT reconstruction method based on the adjoint state method and a quadratic Wasserstein data misfit. A Complete Electrode Model-based thorax phantom with lung-like conductivity inhomogeneities was used to generate simulated boundary voltages, and the proposed method was compared with a conventional EIDORS time-difference Gauss–Newton reconstruction. The Wasserstein–adjoint reconstruction recovered the global lung-like structure more faithfully than the differential baseline, achieving a best structural similarity index measure (SSIM) of 0.874, correlation of r = 0.876, and mean absolute error (MAE) of 0.075 when iterative Gaussian smoothing was included. The reconstruction metrics converged after approximately 1000 iterations, and the custom implementation required about one minute for 1500 iterations on the investigated hardware. A Gaussian-noise analysis simulating measurement noise further showed gradual metric degradation with increasing voltage perturbation, supporting robustness under moderate noise while highlighting the need for future validation with mesh, electrode model, and experimental mismatch.

Keywords: Artificial intelligence in transportation, Planning, management and security in transportation, Automatic control, optimization, real-time operations in transportation
Abstract: This paper proposes a large language model (LLM)-based approach for solving vehicle routing problems, with a single-origin, single-destination route network. In this LLM-based solution, different minimization targets, such as traveling distance, traveling time, and energy consumption, can be achieved. To derive the optimal route, the LLM abstracts and processes real-world historical traffic data, including GPS positions, real-time speeds, and energy consumption rates, from an Excel file. Then, prompt engineering is conducted using natural language to generate algorithms in Python. Simulation results show the effectiveness of the proposed approach.

Keywords: Trajectory and path planning for AVs, Nonlinear and optimal automotive control, Autonomous vehicles
Abstract: Data-Driven Predictive Control (DPC) has emerged as a promising paradigm for directly constructing controllers from measured data, bypassing explicit model identification. It has attracted increasing attention in recent years, with notable advances in both theory and practice. Rooted in behavioral systems theory and Willems’ fundamental lemma, most developments have focused on linear systems, yet a number of extensions have been proposed for nonlinear problems. This paper presents a comparative study of three representative nonlinear DPC strategies applied to autonomous vehicle trajectory tracking. The comparison is carried out along three dimensions: (i) sensitivity to data quality, (ii) tracking accuracy, and (iii) computational efficiency. The evaluation is performed using both a simplified vehicle dynamics model and a high-fidelity CarSim model. The results highlight the effectiveness and limitation of each strategy, confirming their real-time feasibility and offering practical guidance for method selection in trajectory tracking applications.

Keywords: Mission planning and decision making for AVs, Multi-vehicle systems, Trajectory and path planning for AVs
Abstract: Strategic reasoning in autonomous racing has emerged as a critical challenge for multi-agent systems, requiring vehicles to anticipate opponent behaviors while executing aggressive maneuvers. It has attracted increasing attention in recent years, with notable advances in game-theoretic planning and opponent modeling. Based on game-theoretic approaches, most research has assumed perfect knowledge of opponent rationality levels, yet this assumption is unrealistic as drivers exhibit time-varying strategic behaviors that must be inferred online. This paper presents a comparative study of adaptive opponent modeling strategies combining Level-k game theory with Hidden Markov Model (HMM)-based learning. The comparison is carried out along two dimensions: (i) strategic performance with fixed-level opponents and (ii) adaptability to dynamic strategy-switching opponents. The framework infers opponent rationality levels through real-time forward filtering and continuously adapts transition probabilities via Baum-Welch learning using forward-backward algorithm. Experimental validation includes competitive scenarios with dynamic level-switching opponents. Results show that online adaptation achieves superior performance compared to baseline fixed-level strategies across evaluation scenarios. The findings provide insights into the benefits of adaptive opponent modeling for multi-agent autonomous racing.

Keywords: Information processing and decision support in transportation, Kalman filtering techniques in automotive control
Abstract: As connected and autonomous driving technologies advance, vehicles increasingly rely on data from external sensors. Although this information can enhance state estimation, processing all available streams imposes significant communication and computational costs. To address this challenge, we introduce a Sensor Management Center (SMC) that selects a low-cost subset of external sensors in real time while satisfying chance-constrained error bounds derived from an Extended Kalman Filter (EKF) covariance. We formulate the selection problem as a multidimensional minimum knapsack problem and adopt a deficiency-weighted greedy algorithm as an approximate yet efficient solution. The proposed approach is validated through MATLAB simulations and experiments on a 1:15-scale cooperative driving testbed.

Keywords: Trajectory and path planning for AVs, Autonomous mobile robots
Abstract: Safe robotic control in contact-rich environments requires navigating highly non-convex optimization landscapes while enforcing safety constraints and achieving task-level precision. Annealing-based sampling MPC methods provide fast exploration of complex solution spaces, but they lack principled mechanisms for constraint handling and struggle to reach the tight tolerances demanded by factory manipulation tasks. We propose Constrained Annealing-based Sampling MPC (CAS-MPC), a unified framework that combines a constrained annealing-based sampling control with gradient-based refinement. First, we introduce a primal–dual weight reshaping scheme that incorporates inequality constraints into annealing-based sampling MPC, ensuring collision avoidance while preserving sample diversity that trajectory-filtering approaches lose. Second, we propose a hybrid refinement strategy that transitions from fast sampling-based exploration to high-precision control: when far from the goal, CAS-MPC leverages its sampling capability for robust global exploration, and when near the goal it switches to Sequential Linear–Quadratic MPC for millimeter-level pose refinement. We validate our approach on a Unitree Go2 quadruped navigating around obstacles and on a Fetch mobile manipulator performing SE(3) reaching tasks, demonstrating both safe operation and precise goal achievement.

Keywords: Trajectory and path planning for AVs, Learning and adaptation in autonomous vehicles, Autonomous vehicles
Abstract: Most imitation learning planners for autonomous driving are still supervised using displacement-based criteria that emphasize average proximity to the expert trajectory, even when safer alternatives exist. CARE Planner addresses this limitation by extending CAR Planner with risk-aware multimodal supervision. The proposed model preserves constrained ego-state attention to maintain robustness against shortcut-learning-driven attention collapse. It also introduces a clearance-based tail-risk score that guides supervision mode selection and soft targets over trajectory modes. On the nuPlan benchmark, CARE Planner improves overall performance and safety-related metrics over strong baselines, indicating that risk-aware supervision improves the reliability of multimodal imitation planning in challenging scenarios.

Keywords: Multi-vehicle systems, Trajectory tracking and path following for AVs, Autonomous vehicles
Abstract: This paper presents a novel bearing-only formation tracking control law for Euler–Lagrange systems. Compared with single- or double-integrator dynamics, Euler–Lagrange systems provide a more realistic representation of physical agents. The proposed control law enables follower agents to achieve the desired formation and track moving leaders without requiring inter-agent communication or bearing rate measurements. Furthermore, it effectively rejects time-varying disturbances without prior knowledge of their upper bounds. A sufficient condition for collision avoidance among agents is also established. Lastly, a numerical simulation is provided to demonstrate the effectiveness of the proposed control law.

Keywords: Trajectory and path planning for AVs, Autonomous vehicles, Trajectory tracking and path following for AVs
Abstract: Autonomous driving technology continues to evolve, yet it faces severe challenges in highly dynamic and long-horizon scenarios. Existing rule-based methods offer high precision but lack adaptability to changing environments. Conversely, learning-based methods demonstrate strong robustness but often suffer from low terminal accuracy and convergence difficulties. Furthermore, while segmented approaches that separate the approach phase from the parking phase reduce task complexity, they frequently encounter instability during state transitions between phases. To address these limitations, we propose a hybrid end-to-end parking framework that combines Reinforcement Learning with Reeds-Shepp curve guidance. This framework utilizes Reeds-Shepp curves to provide geometric priors that accelerate agent convergence, while leveraging the high-level decision capabilities of Reinforcement Learning to handle dynamic interactions. By adaptively fusing the control outputs from both components, the long-horizon and dynamic parking tasks is performed within a unified framework. Experimental results demonstrate that the proposed method significantly outperforms traditional baselines. Specifically, the parking success rate reaches 98% in scenarios with 4 dynamic obstacles and remains at 89% in complex interactive scenarios involving 8 dynamic obstacles. This study validates the effectiveness of the hybrid driving strategy in solving long-horizon complex parking problems.

Keywords: Trajectory and path planning for AVs, Trajectory tracking and path following for AVs, Autonomous vehicles
Abstract: This paper presents a novel integrated path planning and motion control system for a micro-vehicle that employs an improved A* algorithm with the Dynamic Window Approach (DWA) for path and trajectory planning and Model Predictive Control (MPC) for tracking. The proposed A* algorithm improves the efficiency and safety of path planning through adaptive neighborhood search, dynamic weighting, and an obstacle-proximity penalty. The proposed DWA algorithm incorporates a soft-constraint subfunction of relative motion, besides steering angle constraints, to generate smooth and feasible trajectories. Finally, MPC is used to achieve accurate and stable tracking control. The simulation results show that the proposed framework can generate safe and smooth trajectories in complex obstacle environments while maintaining high tracking accuracy and real-time feasibility.

Keywords: Trajectory tracking and path following for AVs, Autonomous vehicles
Abstract: Autonomous street sweepers require high-precision trajectory tracking to ensure effective cleaning and curb-collision avoidance. Standard Model Predictive Control (MPC) frameworks using nominal kinematic models struggle with significant, non-stationary mis-matches caused by unmodeled dynamics from road-crown slopes, friction variations, and tireslip effects. Although Gaussian Processes (GP) can learn such dynamics, their static, offline-tuned hyperparameters cannot adapt to changing conditions, limiting safety guarantees. This paper proposes a unified control framework integrating an Adaptive Sparse Gaussian Process (ASGP) with application-specific probabilistic chance constraints. The ASGP employs a forgetting factor to track time-varying residuals, while the chance constraints stem from the sweeping mechanism’s geometry, ensuring true safety and task boundaries. The framework is validated in simulation and on a full-scale commercial sweeper. Simulation and real-world experiments demonstrate improved tracking accuracy and robust safety performance, meeting the application’s precision requirements.

Keywords: Trajectory tracking and path following for AVs, Autonomous vehicles, Motion control for AVs
Abstract: The dynamics of quad-rotors under complex maneuvers present significant challenges due to their strong coupling and underactuation. This paper introduces a dual-quaternion–based modeling framework that provides a compact representation of quad-rotor dynamics on SE(3). By integrating geometric tracking control with a virtual frame transformation, the proposed method effectively mitigates the inherent underactuation of quad-rotor systems. Building on this formulation, a two-layer control architecture is developed, consisting of: (1) an outer-loop position and attitude tracking controller, and (2) an inner-loop twist stabilization controller. Numerical simulations and comparative studies validate the effectiveness of the proposed approach, demonstrating stable trajectory tracking at speeds up to 14 m/s and significantly outperforming conventional geometric controllers in high-speed and complex maneuvering scenarios.

Keywords: Trajectory tracking and path following for AVs, Marine system guidance, navigation and control, Multi-vehicle systems
Abstract: This paper addresses the distributed formation control of multiple unmanned surface vehicles (USVs) subject to unknown ocean currents and communication bandwidth constraints. To reduce communication load, an asynchronous event-triggered mechanism is introduced at the communication layer, eliminating the need for continuous monitoring inherent in synchronous or continuous transmission schemes. For the kinematic subsystem, an Extended State Observer based guidance law is developed to compensate for current-induced deviations and ensure accurate trajectory tracking. For the dynamic subsystem, an adaptive control law based on a linear analytical quantization model is proposed, which allows controller design without prior knowledge of the input quantization parameters, which avoids the restrictive assumption of fixed quantizer settings. Simulation results verify the effectiveness of the proposed strategy.

Keywords: Biomedical and medical imaging, image processing, visualization, Biomedical signal measurement and processing, Biomedical system modeling, identification, and simulation
Abstract: Alzheimer’s disease (AD) is increasingly seen as both structural degeneration and disruption of brain network dynamics. Yet, quantitative neurovascular dynamics characterization is limited. We propose a control-theoretic framework to examine neurovascular complexity using resting-state fNIRS. Data were collected from 83 participants: healthy controls (HC, n = 27), mild cognitive impairment (MCI, n = 37), and AD (n = 19). Sliding-window nonlinear complexity measures—Higuchi’s fractal dimension, spectral entropy, and wavelet entropy—were computed for each channel to construct complexity-coupling networks based on inter-channel temporal coordination. Static topology, dynamic descriptors (mean, variability, range, temporal dependence), and state-based metrics from k-means clustering were analyzed. Task–rest reconfiguration and cognitive performance associations were evaluated. AD showed reduced network integration and selective loss of positive complexity coupling. Decreased temporal variability and state entropy, with fewer visited states, indicate diminished dynamic flexibility and state space contraction. Additionally, AD exhibited reduced task-induced network reconfiguration, reflecting impaired neurovascular adaptability. Among measures, spectral entropy–based networks, especially in deoxyhemoglobin signals, showed the strongest group discrimination and robust cognitive score associations. These findings show AD involves collapse of neurovascular dynamic complexity, with reduced spectral diversity, increased temporal rigidity, and constrained state-space dynamics. The framework offers a sensitive, mechanistically interpretable biomarker for tracking disease progression.

Keywords: Biomedical signal measurement and processing, Control of physiological and clinical variables, Medical devices, systems and solutions
Abstract: Continuous myoelectric control of prosthetic hands requires robust regression from surface electromyographic (sEMG) signals, which are affected by noise, cross-talk, and overlapping muscle activity. We propose a hybrid approach combining the Frisch scheme, an Errors-in-Variables method, with Non-Negative Matrix Factorization (NMF) for muscle synergy extraction. The Frisch stage improves the statistical consistency of sEMG data prior to factorization, enabling more reliable continuous control signal estimation. The method was evaluated against standard and sparse NMF variants on data from five subjects performing four gestures. Results show that the proposed Frisch+NMF approach achieves lower reconstruction error, demonstrating the benefit of noise-aware preprocessing for myoelectric control.

Keywords: Biomedical system modeling, identification, and simulation
Abstract: Stem cells play a crucial role in biomedical research, offering remarkable potential for regenerative medicine, disease modeling, and drug discovery. Their ability to self-renew and differentiate into specialized cell types makes them essential for tissue repair and regeneration. This note explores a basic model of the differentiation/proliferation mechanisms while accounting for the maximum population size the environment can sustainably support due to limiting resources - i.e., the carrying capacity. Regulatory mechanisms affecting the proliferation rate are investigated using both deterministic and stochastic approaches. The deterministic analysis identifies regions of the parameter space that ensure a stable balance between stem and differentiated cells, while the stochastic approach provides valuable insights suggesting that a positive feedback on the proliferation rate leads to lower fluctuations in the accumulation of differentiated cells.

Keywords: Biomedical system modeling, identification, and simulation
Abstract: The Warburg effect describes the preference of highly proliferating cells (like cancer cells) for aerobic glycolysis and lactate production despite oxygen availability. In a recent paper, Jaiswal and Singh (2024) proposed that this behavior arises from a negative feedback loop linking cytoplasmic NADH levels and cell proliferation. Their model integrates glycolysis, oxidative phosphorylation, and pyruvate-to-lactate conversion to explain how the NADH/NAD+ ratio governs proliferation and quiescence. Here, we propose the qualitative behavior analysis, showing how quiescent and non quiescent equilibria arise according to model parameters. The corresponding bifurcation diagrams provide new biological insights on cellular behavior and pave the way to further investigation on the cellular machinery leading to the Warburg effect.

Keywords: Biomedical system modeling, identification, and simulation, Artificial pancreas or organs
Abstract: Mathematical models of glucose–insulin dynamics are essential for managing type 1 diabetes. Their applications extend to closed-loop control, forecasting glucose trajectories, anticipating and detecting hypo- and hyperglycemia, and supporting real-time decision-making. In this work, we introduce the Compartmental Recurrent Neural Network (COMP-RNN), which advances the Biologically-Informed Recurrent Neural Network (BI-RNN) framework for glucose–insulin dynamics modeling. The COMP-RNN extends the data-driven strengths of the BI-RNN by embedding physiological structure directly into the model architecture. Specifically, it leverages structured recurrent networks aligned with canonical physiological compartments and incorporates prior physiological knowledge into training through an augmented cost function. The COMP-RNN is trained and validated on in silico cohorts. Compared to both BI-RNN and a benchmark linear model, the proposed approach achieves higher predictive accuracy and improved parameter efficiency, while better reflecting the underlying physiological system.

Keywords: Biomedical system modeling, identification, and simulation, Dynamics and control of gene expression and metabolic pathways
Abstract: Feedback is at the core of biological systems found in medicine and in biotechnology. While design metaphors from control engineering are widely used to understand negative feedback in such systems, they are relatively uncommon for positive feedback, especially for biomolecular circuits. Here, we extended a block diagram modelling framework for the design of positive feedback. We found a quantitative design guideline for the strength of the positive feedback, which when wrapped around a saturation function can give a threshold and a hysteretic response. The critical feedback strength was inversely proportional to the saturation value and directly proportional to the input scale where saturation starts. We found that this saturation-threshold-hysteresis hierarchy persisted in a realistic model of a positively autoregulated gene. We showed how this classical model fitted well in a block diagram framework with multiplicative feedback and derived expressions for the critical feedback strength in terms of the saturation parameters. The dependence of the critical feedback on the parameters matched with the obtained design guideline. To complete a rigorous workflow, we discussed how Groebner Bases computations and an Interval Newton algorithm can be used to provide validated numerical solutions in biological positive feedback systems. These results should be helpful in the analysis and design of biomolecular systems with applications to the control of biomedical systems and in biotechnology.

Keywords: Hybrid, electric and alternative drive vehicles, Nonlinear and optimal automotive control, Engine and powertrain modeling and control
Abstract: Advancing fuel cell systems for non-road applications requires addressing key challenges in durability and fuel efficiency. This paper presents a health-aware predictive energy management strategy for fuel cell non-road vehicles. This strategy explicitly integrates component degradation and fuel economy within a multi-objective energy management optimization. Its performance is evaluated using real-world wheel loader driving data across various operating scenarios. Full-lifetime investigations accounting for progressive component degradation demonstrate that the presented approach enables improved fuel cell and battery lifetime balancing. Compared to a purely fuel-minimizing strategy, it significantly extends the vehicle’s service life while maintaining high fuel efficiency.

Keywords: Machine learning and artificial intelligence in chemical process control, Monitoring, performance assessment, and fault detection in chemical process control, AI methods for FDI/FTC
Abstract: Robust causal inference under noise and nonlinear dynamics is essential for industrial fault detection and root-cause diagnosis. A unified framework is presented that integrates Granger causality for structural priors, LSTM-based temporal encoding, GATv2-driven graph refinement, and FGSM adversarial training. The approach concurrently performs forecasting, detection, and diagnosis within a structure-aware pipeline. Evaluation on synthetic datasets and the Tennessee Eastman Process reveals substantial performance gains over Granger-only baselines in both graph reconstruction and root-cause localization. Adversarial augmentation yields consistent improvements across all metrics, with particularly strong gains in precision and noise resilience, validating the framework's reliability for industrial diagnostic applications.

Keywords: Mechatronics for advanced manufacturing and energy systems, Mechatronic system estimation, identification, control, Application of mechatronic principles
Abstract: In inkjet printing, optimal paper moisture is critical for print quality and is achieved through hot‑air impingement in a fixation unit. This paper presents a modular digital twin of the fixation unit that models the thermo‑fluidic drying process and enables real‑time monitoring of its spatio‑temporal behavior. The digital twin is formulated as an infinite‑dimensional state estimator that infers unmeasured thermal states from limited sensor data while remaining robust to external disturbances. Modularity is realized through a graph‑theoretic model in which each subsystem is represented by PDEs, with evaporation modeled as a nonlinear boundary effect via a Linear Fractional Representation. Using the Partial Integral Equation (PIE) framework, a unified approach for simulation, analysis, and estimator synthesis is developed and validated with data from a commercial inkjet printer. An mathcal{H}_{infty}‑optimal Luenberger estimator is synthesized to estimate internal thermal states, together forming a digital twin that enables spatio‑temporal monitoring capabilities beyond those available in traditional printing systems.

Keywords: Queuing systems and performance model                                       , Stochastic control, Control over networks
Abstract: This paper presents a new formulation of the back-pressure control problem based on continuous-time Markov chains defined over a discrete state space, combined with a game-theoretic framework for urban traffic networks. Queue pressure differences in the back-pressure scheme are embedded in the jump intensities of the Markov chains, yielding an equivalent stochastic reformulation of the problem as a non-cooperative game. The resulting equilibrium corresponds to a control strategy that governs the evolution of the underlying continuous-time Markov chains. Simulation results illustrate the qualitative behavior of the proposed framework under time-varying traffic demand and validate the auxiliary Markov-chain reformulation of the back-pressure objective. Rather than focusing on numerical benchmarking, the contribution of this work lies in providing a new theoretical perspective that unifies stochastic modeling and game-theoretic decision-making in the context of urban traffic control.

Keywords: Water distribution systems, Fault detection and isolation methods
Abstract: We formulate the pipe clogging detection and localization problem as a hydraulic resistance parameter estimation task. We derive conditions on the demands, which the system operator may control through pumps, under which resistance estimation admits a unique solution. We perform resistance estimation in a simulated version of a well-field in Viborg, Denmark, under progressively increasing clogging, demonstrating that the theoretical conditions accurately predict the estimation performance.

Keywords: Kalman filtering techniques in automotive control, Electric and solar vehicles, Vehicle dynamic systems
Abstract: This study evaluates the effect of filtering algorithms on the traction performance of a single-motor electric tractor during plow tillage. A 19-kW electric tractor equipped with telemetry-based wheel torque meters and proximity sensors was used to measure axle torque and rotational speeds. Two filtering approaches—Kalman Filter Algorithm (KFA) and an Artificial Neural Network (ANN) filter—were applied to estimate axle torque, and their effectiveness was assessed against measured torque data. The estimated torque was then used to compute traction-related indicators, including axle power, net traction force, and tractive efficiency (TE). Results showed that KFA closely matched the measured axle torque across the operation, demonstrating stable dynamic response and minimal overshoot. ANN produced moderate accuracy but showed greater fluctuation during rapid load transitions. In contrast, the unfiltered dataset significantly overestimated torque, leading to unstable traction behavior. Average TE values were 30.42%, 29.81%, 28.90%, and 17.10% for measured, KFA, ANN, and unfiltered data, respectively, with statistical analysis confirming no significant difference between measured and KFA values. These findings demonstrate that KFA is an effective filtering method for improving real-time load estimation and traction performance during plow tillage. The results provide essential insights for developing more efficient motor control and energy management strategies for electric tractors.

Keywords: Energy storage systems, Real time simulators for energy systems
Abstract: Battery modeling plays an important role in the application of Lithium-ion batteries as it provides important information for system design and control purposes. Physics-based models, such as the Doyle-Fuller-Newman model, are useful for more advanced applications such as ageing-aware charging as they provide information on the internal states of the battery. In this paper, we present an open-source physics-based battery modeling toolbox in Matlab, called TOOFAB (TOOlbox for FAst Battery simulation). This paper includes an overview of the numerical implementation of the Doyle-Fuller-Newman battery model, a thermal model, an ageing model and a parameter estimation procedure. We provide a detailed description of the toolbox features and functions and we further provide example use cases, including one example of model parameter estimation and one example of simulation of ageing.

Keywords: Internet based control education, Control education laboratories, Control engineering curricula
Abstract: Practical laboratory work is essential in control education, but its logistics do not scale well with growing student numbers and limited, expensive hardware. Standard remote-desktop solutions offer remote access but lack isolation, scheduling, and administrative control of laboratory setups. This paper presents TU/e Remote Labs, a platform that provides controlled, web-based access to real laboratory setups through virtualized environments and automated experiment session management. We describe the system architecture, the interactive and queued experiment workflows, and the integration of the platform into several control courses, supported by usage data demonstrating its flexibility and scalability.

Keywords: Micro and nano mechatronic systems, Mechatronic system modeling, design, optimization, High-performance motion control systems
Abstract: Park Systems FX40 is the latest innovation in atomic force microscopy (AFM), designed for high-resolution imaging of small samples. Its low noise floor, minimal thermal drift, and enhanced mechanical stability enable highly precise and reliable measurements. Key features include automatic probe exchange, automatic laser beam alignment, and a sample-view camera. With a powerful controller featuring an 8-channel lock-in amplifier and 5 MHz bandwidth for advanced signal processing, the FX40 supports a broad set of cutting-edge AFM modes. This paper demonstrates the enhanced performance and improved user-friendliness of the FX40 system.

Keywords: Model predictive control, Applications of optimal control, Real-time optimal control
Abstract: This paper introduces the Algebraic Model Predictive Control (A-MPC) toolbox for Simulink, developed for Linear Time-Invariant (LTI) systems. Although conventional Model Predictive Control (MPC) offers significant advantages, its industrial adoption is limited due to the computational burden of online optimization algorithms. This limitation often results in reduced real-world performance or increased product costs. To address this issue, a new algebraic formulation has been developed and implemented in the toolbox. The method reformulates constraints into a continuous form using the hyperbolic tangent function, enabling the application of first-order necessary conditions to the optimal control problem. Furthermore, the method has been extended with additional control options to enhance overall performance depending on the application, including Disturbance Input, Integral Action, and Rate Limiter. Since the proposed approach eliminates online optimization and relies entirely on algebraic computation, the computation time is significantly reduced. Simulation results for several use cases are presented and evaluated in terms of control performance and computation time, demonstrating the improved efficiency and effectiveness of the toolbox. The Algebraic MPC Toolbox is available at https://github.com/TalhaUlukir/AMPC.

Keywords: Power plant control, Nuclear power, Digital twins for power and process systems
Abstract: This paper proposes a hierarchical ensemble control framework for safe autonomous load-following of Small Modular Reactors (SMRs), integrating reinforcement learning (RL) with classical robust control under formal safety certification. A Physics-Informed Neural Network (PINN) digital twin provides a differentiable surrogate model enabling model-based policy gradient optimization with ~ 10X sample efficiency improvement. Parallel H∞ robust and SAC-CMDP controllers are adaptively blended by a meta-controller with Lyapunov-guaranteed switching stability, while a Control Barrier Function (CBF) safety filter enforces nuclear constraints via real-time quadratic programming. Gain margin analysis confirms that fixed-gain PID—standard practice in current PWR/SMR operations—loses stability margins below 50% power, providing the key engineering motivation for the multi-controller architecture. Simulation results on a 24-state first-principles SMR digital twin under 10% model–plant mismatch demonstrate 75.6% IAE reduction over baseline PID with zero safety constraint violations and 5×faster settling.

Keywords: Mechatronic system estimation, identification, control, Mechatronic system modeling, design, optimization, Soft robotics
Abstract: Accurate modelling of flexible robotic structures is essential for achieving precise control in many engineering applications. Parameter estimation for nonlinear flexible systems becomes particularly challenging when measurements of dynamic deformation are limited and the system’s initial states are unknown. This paper introduces a continuous-time parameter and deformation trajectory estimation method for robotic payloads modelled as Cosserat rods, using temporal and spatial basis functions. The method enables identification of parameters that enter the dynamics nonlinearly from limited measurements. Simulation and experimental validation on a flexible manipulator payload equipped with a tip-mounted IMU demonstrate accurate reconstruction of both deformation and wrench responses, highlighting strong performance under limited sensing conditions.

Keywords: Mechatronic system estimation, identification, control, High-performance motion control systems, Mechatronics for robotic systems
Abstract: Shape Memory Alloy (SMA) actuators are widely used in mechatronics and robotics due to their high energy density, flexibility, and relatively large strain. However, their nonlinear and hysteretic behavior poses great challenges for reliable deployment in real-world scenarios. This work presents a model-based robust control approach for SMA actuators. The rate-dependent hysteretic response of the actuator is described via a Wiener model, combining a linear time-invariant dynamics with a static hysteresis operator. While this model captures well the actuator's response, it does not allow direct compensation of the hysteresis at the plant's input via an inverse model. To address this issue, we introduce a compensator-free robust motion control strategy that treats the hysteresis as a bounded time-varying uncertainty, and achieves set-point regulation with a prescribed decay rate in the full actuator range. The robust controller consists of a PI law, whose gains are systematically tuned via a bilinear matrix inequality optimization. Experimental validation confirms the effectiveness of the closed-loop scheme.

Keywords: Mechatronic system modeling, design, optimization, Mechatronic system estimation, identification, control, Mechatronics for robotic systems
Abstract: This paper presents the mechatronic design, modeling, and experimental validation of a three-degree-of-freedom (3-DOF) micro parallel robot with a prismatic–spherical (3PS) topology actuated by Hydraulically Amplified Self-Healing Electrostatic (HASEL) actuators. Each actuator provides prismatic motion, while a compliant interface forms spherical joints. A prototype was built and tested using laser tracking. A port-Hamiltonian (PH) model combined with forward kinematics (FKM) captures the nonlinear dynamics, and inverse kinematics (IKM) estimates actuator inputs. Parameters were identified using nonlinear grey-box (NLGB) estimation, yielding a compact model suitable for control design.

Keywords: Mechatronic system modeling, design, optimization, Medical and rehabilitation robotics, Mechatronics for robotic systems
Abstract: With the inherent compliance and having an internal channel for tool delivery, tendon-driven notched continuum robots (TDNCRs) are widely used in the medical and industrial domains. However, the notched design introduces bending stiffness anisotropy in TDNCRs, which has not been explicitly considered in existing models. Therefore, we propose a mechanics-based model that accounts for stiffness anisotropy. Starting from the computation framework of area moment of inertia (AMI), both planar and spatial deformations are analyzed, and the mechanics-based model is developed based on Euler–Bernoulli beam theory. A prototype of TDNCR was designed, and experiments on bending angle and bending plane angle were conducted to validate the model. The root-square-mean errors (RSMEs) for bending angle are within 2 degrees, and the deviation for bending plane angle is within 4 degrees, validating the proposed model.

Keywords: High-performance motion control systems, Micro and nano mechatronic systems, Mechatronics for robotic systems
Abstract: This paper presents a robust sampled-data model-free adaptive control (RSDMFAC) strategy specifically designed for precise position control of actuators in a robotic hand, using only position feedback. The proposed method is simple to implement, computationally efficient, and entirely independent of parameter identification. This makes it highly suitable for applications requiring rapid and accurate precise manipulations, where computational efficiency is crucial. The proposed approach ensures high performance in closed-loop systems, maintaining robustness even in the presence of external disturbances, sensor noise, and uncertainties in system dynamics. Experimental validation demonstrates the effectiveness of the control algorithm, with results highlighting its capability to achieve precise positioning and stability under challenging real-world conditions. This method offers a reliable and scalable solution for advanced robotic manipulation tasks in environments where high precision and adaptability are required.

Keywords: Mechatronic system estimation, identification, control, Soft robotics, Mechatronic system modeling, design, optimization
Abstract: Developing dynamic models for tendon-driven continuum robots is challenging due to their nonlinear, high-dimensional, and friction-dominated dynamics. This paper presents a comparative study of data-driven system identification methods—including N4SID, ARX, and SINDYc—for modeling a tendon-actuated continuum robot with rolling joints developed at CERN. Despite the high number of joints of the robot, experimental analysis reveals that a two-degree-of-freedom dynamic model can accurately capture the system dynamics, owing to strong kinematic dependencies between the joints. The models are validated against experimental data, and used in the design of a model predictive controller, demonstrating their feasibility for real-time control.

Keywords: Mechatronic system estimation, identification, control
Abstract: This paper is concerned by the port-Hamiltonian modeling and control of a dielectric elastomer actuator designed for use in a cardiac assist device. The proposed nonlinear model captures the actuator’s hyperelastic material behavior, viscoelastic damping, and electromechanical coupling, and remains valid for large deformations up to 40%. An original Interconnection and Damping Assignment Passivity-Based Control strategy is developed to achieve closed-loop stabilization at a desired position. The accuracy of the multiphysics model and the performances of the proposed controller are experimentally validated.

Keywords: Mechatronic system estimation, identification, control, Autonomous navigation
Abstract: This paper addresses the design and validation of robust estimation strategies for intelligent vehicle systems, with a particular emphasis on lateral velocity and tire-road interaction force estimation. A Takagi-Sugeno (TS) fuzzy model is used to represent the nonlinear vehicle dynamics, and a new proportional-integral (PI) observer is developed based on this representation. The observer synthesis is reformulated as a convex optimization problem under linear matrix inequality (LMI) constraints, and its performance is formally guaranteed using Lyapunov stability theory. The proposed observer is validated using real data collected from a Renault ZOE experimental platform. The results demonstrate high estimation accuracy and improved reliability compared with existing methods across a wide range of driving scenarios, highlighting the suitability of the approach for advanced driver assistance systems (ADAS) and autonomous driving applications.

Keywords: Mechatronic system estimation, identification, control, Autonomous navigation, Mechatronics for mobility systems
Abstract: The versatility of Wheeled Mobile Robots in industry and research has motivated the development of advanced control strategies. One of their main tasks is trajectory tracking, which aims to track a time-varying path to reach a desired pose. However, wheeled mobile robots are inherently subject to external and internal uncertainties, hence degrading their performance in real-world scenarios. To address this issue, this work proposes a prescribed-time controller combined with a twisting control to compensate for matched disturbances. Moreover, the proposed scheme employs the Euler-forward discretization method, which makes the tracking error converge asymptotically to zero despite the effects of kinematic disturbances, sampling time increments, and the zero-order-hold effect. Closed-loop stability is demonstrated via a discrete Lyapunov analysis. In addition, a further analysis is performed to determine the exact sampling instant at which the closed-loop system becomes unstable, and a reset is performed to avoid instability. Finally, extensive numerical simulations validate the proposed scheme by varying initial conditions, introducing external disturbances, and increasing the initial sample time.

Keywords: Mechatronic system estimation, identification, control, High-performance motion control systems, Mechatronics for robotic systems
Abstract: We present a practical identification-to-control workflow for the attitude (roll, pitch) and depth (heave) channels of a BlueROV2 using the Modified Relay Feedback Test (MRFT). MRFT enables the generation of test oscillations at predetermined phase lags of the plant, producing oscillation data (amplitude–frequency) across phases set by the MRFT parameter. The induced oscillations occur in the second and third quadrants of the complex plane; notably, while a linearized physics-based model with force as input cannot yield phase lags below −180◦, the MRFT response in the second quadrant reveals the presence of additional actuator and sensor dynamics, critical for accurate controller design. The measured oscillation data are mapped to frequency-response points via the describing-function method, providing comprehensive dynamic samples inclusive of these effects. These points enable Nyquist-domain fitting for system identification, from which a compact low-order complex-domain model is obtained that balances magnitude and phase errors. Using this identified model, a Type-B PID controller is tuned. The resulting gains deliver accurate set-point tracking in experiments with and without the presence of waves, demonstrating the robustness inherent to MRFT-based identification. Overall, the pipeline—MRFT sampling at several phase lags, compact complex-fit identification, and ITAE-based PID tuning—offers a fast, instrumentation-light path from field data to implementable attitude and depth control on ROV platforms. A demonstration video of the experimental results is available at Ibrahim (2025).

Keywords: Mechatronic system estimation, identification, control, Mechatronic system modeling, design, optimization
Abstract: This paper presents a parametric data-driven feedforward control method for horizontal boom motion of a rotary crane system. System parameter data are obtained as a linear least square error minimization solution. An angular velocity-based potential field predicts parameter values corresponding to instantaneous reference velocities. The parameter predictions are used as inverse dynamics to generate the feedforward control signal. Computation time and experimental results illustrate the proposed method's effectiveness practicality in tracking performance.

Keywords: Mechatronic system estimation, identification, control, Mechatronic system modeling, design, optimization
Abstract: Transmission electron microscopes (TEMs) enable atomic-scale imaging but suffer from aberrations caused by lens imperfections and environmental conditions, reducing image quality. These aberrations can be compensated by adjusting electromagnetic lenses, but this requires accurate estimates of the aberration coefficients, which can drift over time. This paper introduces a method for the estimation of aberrations in TEM by leveraging the relationship between an induced tilt of the electron beam and the resulting image shift. The method uses a Kalman filter (KF) to estimate the aberration coefficients from a sequence of image shifts, while accounting for the drift of the aberrations over time. The applied tilt sequence is optimized by minimizing the trace of the predicted error covariance in the KF, which corresponds to the A-optimality criterion in experimental design. The resulting non-convex optimization problem is solved using a gradient-based, receding-horizon approach with multi-starts. The proposed method is validated on a real TEM set-up. The results show that optimized patterns significantly outperform naive approaches. A direct comparison with a Zemlin tableau implementation (Zemlin et al., 1978) shows that the proposed method achieves comparable or higher image quality on amorphous specimens, while additionally extending to non-amorphous specimens where the Zemlin tableau cannot operate.

Keywords: Mechatronic system estimation, identification, control, High-performance motion control systems, Aerial, field, and marine robotics
Abstract: This paper addresses the pose tracking of omnidirectional mobile robots subject to external perturbations. We consider the mobile robot as a rigid body whose configuration space is the Special Euclidean group SE(2). We propose a nonlinear disturbance observer to estimate the external perturbations. The estimated signals are combined with a nonlinear trajectory tracking law that guarantees asymptotic convergence to the reference time-varying pose. Experimental results validate the proposed control strategy.

Keywords: Mechatronic system estimation, identification, control, Mechatronics for robotic systems
Abstract: This paper investigates balance control for underactuated wheeled bipedal robots (WBRs), where conventional feedforward compensation is insufficient for disturbance rejection. To simultaneously handle external disturbances and parameter uncertainties, an event-triggered composite compensation framework is proposed. Specifically, a nonlinear disturbance observer is first employed to estimate the lumped disturbance. An event-triggered mechanism then filters this estimation to facilitate accurate parameter identification. The resulting parameters are employed to refine the reference commands of the model predictive controller (MPC), thus integrating underactuation-aware compensation with recursive optimization. The effectiveness of the proposed approach is validated through comprehensive hardware experiments.

Keywords: Mechatronic system fault detection, diagnostics, hardware-in-the-loop simulation, Mechatronics for mobility systems, Mechatronic system estimation, identification, control
Abstract: Mobile robots, like the ultra-flat overrunable (UFO) robot platform, used in automotive active safety tests, currently lack self-diagnostic capabilities necessary to detect present hardware defects. This circumstance can lead to more severe failures, causing expensive repairs and operational downtime. This work proposes, for the first time, a reconstruction-based time-series anomaly detection model for these mobile robots, considering defect classes such as unevenly worn full-rubber tires or damaged dampers. Unlike prior publications, the proposed approach leverages the vast quantities of unlabeled data generated during routine operation through a simple pre-training step. Furthermore, it optimizes the hyperparameters of the implemented gated recurrent unit-based variational autoencoder (GRU-VAE) and evaluates both a stateless, windowed training approach and one using truncated backpropagation through time (TBPTT). The model’s generalization capabilities are demonstrated by successfully detecting six defect types, with four of them not present in the data used for hyperparameter optimization and threshold selection. This is validated using a test set collected from five system instances at various points over a period of several months, achieving an F1 score of 0.936, indicating strong practical viability.

Keywords: Mechatronic system modeling, design, optimization, High-performance motion control systems, Mechatronics for robotic systems
Abstract: In this paper, we present a cable-driven parallel robot (CDPR) capable of catching darts at arbitrary, predefined segments of a standard tournament dartboard, repeatedly and with high accuracy. The robot employs a previously published novel lightweight CDPR design that minimizes moving masses, thereby enabling high accelerations and precise positioning. To fully exploit the potential of this design, an efficient solution to the forward kinematics problem is necessary. We propose a computationally efficient algorithm for forward kinematics, provide its derivation and analysis, and validate the approach through experiments with the dart-catching robot. A video demonstration is available online: https://youtu.be/_WwZZbF93H4

Keywords: Mechatronic system estimation, identification, control, Robot perception and sensing, Aerial, field, and marine robotics
Abstract: Precise state estimation for heavy-duty machinery requires accurate sensor placement; however, manual calibration is error-prone. We propose a probabilistic framework for estimating the extrinsic parameters, namely lever arms and joint axes, using only IMU data. By unifying rigid-body constraints within a factor graph, our method accounts for measurement noise and resolves translational gauge ambiguity via geometric regularization. High-fidelity simulations of a forestry grapple demonstrate robust performance, achieving sub-degree axis accuracy (RMSE 0.244^circ) and sub-centimeter lever-arm precision (RMSE 5.29 mm). This formulation enables a plug-and-play magnetometer-free solution and demonstrates the feasibility of self-calibration for industrial systems.

Keywords: Mechatronic system estimation, identification, control, Mechatronics for robotic systems, Smart structures and vibration control
Abstract: This paper presents a filtered high-order control barrier function (FHOCBF) for control-affine systems, applied to enforce safety constraints on the deflection of flexible payloads. A reduced-order model is constructed using the floating-frame-of-reference formulation combined with Craig–Bampton reduction to capture dominant elastic modes. An extended Kalman filter reconstructs the rigid–flexible states from collocated force measurements. To mitigate vibration during motion, a trajectory optimization scheme incorporating input shaping produces vibration-aware reference commands. Safety constraints on payload deflection are enforced through a FHOCBF, which prevents excessive deformation while avoiding the vibration amplification typically caused by classical HOCBFs. Simulation studies and experiments demonstrate that the proposed filtered safety-critical architecture achieves precise, vibration-aware, and deflection-constrained control of flexible payloads.

Keywords: Healthcare management, disease control, critical care, Decision support and control in medicine, Intensive and chronic care or treatment
Abstract: Economic savings achieved through targeted isolation avoid additional disease burdens and effectively address the disease-economy trade-offs in epidemic control. In this study, we use phase-space analysis to derive the explicit solution of the optimal control problem that minimize the infection peak given budget limitation. The optimal policy obtained is an adaptive control where the isolation rate dynamically adjusts according to the current epidemic state. We show that targeted isolation control policy achieves the same infection peak as transmission reduction policies under equivalent budgets, while avoiding broad socio-economic disruptions. Additionally, we show through numerical simulations that the control resolves the epidemic faster and reduces total infections. This demonstrates that targeted isolation can strike a balance between public health and economic stability, offering actionable insights for public health decisions moving forward

Keywords: Optimal control theory, Learning methods for optimal control, Numerical methods for optimal control
Abstract: Policy gradient methods are a powerful family of reinforcement learning algorithms for continuous control that optimize a policy directly. However, standard first-order methods often converge slowly. Second-order methods can accelerate learning by using curvature information, but they are typically expensive to compute. The linear quadratic regulator (LQR) is a practical setting in which key quantities, such as the policy gradient, admit closed-form expressions. In this work, we develop second-order policy gradient algorithms for LQR by deriving explicit formulas for both the approximate and exact Hessians used in Gauss--Newton and Newton methods, respectively. Numerical experiments show a faster convergence rate for the proposed second-order approach over the standard first-order policy gradient baseline.

Keywords: Optimization-based estimation and control, Model validation, Uncertain systems
Abstract: Accurate mode and fault discrimination are essential for ensuring system safety and operational efficiency. However, reliable identification is often challenged by confounding disturbances, bounded uncertainties, and modeling inaccuracies. Although recent advances in machine learning can estimate the likelihood of mode occurrences, these approaches generally fail to provide deterministic guarantees, especially under worst-case conditions. To address this limitation, we propose a probabilistically informed active mode discrimination framework that incorporates mode probability priors into the active input design. A probing strategy is developed for systems with probabilistic modes under bounded uncertainties, where a control objective is integrated to minimize the impact of probing on system performance. Furthermore, deterministic discrimination conditions and feasible minimal-time criteria are established to reduce computational complexity. The overall problem is formulated as a mixed-integer linear/quadratic programming (MILP/MIQP) problem solvable by standard optimization solvers.

Keywords: Real-time optimization and control in chemical processes, Advanced process control, Model-predictive and optimization-based control in chemical processes
Abstract: In this paper we aim to steer the output of a continuous-time dynamical system to a state in which a performance index is minimised. To this end, we first obtain a dynamical system that models the gradient of such performance index with respect to the control input of the plant. In general, such model for the gradient depends on the plant's model, which we assume available. Then, we design a control law that ushers the gradient to zero, and the input and output values to the optimal condition, despite of the presence of unknown inputs which shift the location of the optimum state. Such control law updates continuously as they depend on the instantaneous value of the gradient. To show applicability, we perform the numerical simulation of a fermentation process.

Keywords: Water distribution systems, Big data and machine learning applied to smart cities
Abstract: Chance-constrained optimization (CCOPT) of water distribution systems (WDSs) under uncertainty typically relies on computationally expensive Monte Carlo simulations to evaluate the chance constraints. This paper presents an artificial neural network (ANN)-based CCOPT framework for WDS, in which a trained neural network surrogate replaces Monte Carlo simulations for evaluating chance constraints, enabling real-time capable optimization of both hydraulic and water quality objectives. The method integrates a nonlinear programming (NLP) optimizer with an ANN model to generate operation scenarios for estimating state variables such as pressure and water age. Based on the assessment of constraint violations, a heuristic rule is designed to update the weighting parameter in the objective function. This enables adaptive refinement of the NLP formulation based on the predicted outcomes. Tests on a benchmark WDS model demonstrate the prediction accuracy being higher than 97% and a reduction in computation time over 99% compared to the Monte Carlo based method. This highlights the potential of the proposed approach for real-time CCOPT of large-scale WDSs, providing water utilities with rapid, uncertainty-aware decision support for joint pressure and water age control.

Keywords: Control and optimization for sustainability and energy systems, Model-predictive and optimization-based control in chemical processes, Machine learning and artificial intelligence in chemical process control
Abstract: Green ammonia production systems powered by intermittent renewable energy must meet periodic demand under tight unit and storage constraints. We propose Q-learning-based stochastic model predictive control, a methodology integrating a stochastic model predictive control framework with a Q-function as the terminal cost. The proposed method explicitly enforces hard constraints, effectively manages both short-term and long-term disturbances, and offers significant advantages in terms of on-line computational speed. Simulation results show that the proposed method outperforms Nonlinear Model Predictive Control, Double Deep Q-Network, and Q-learning-based Model Predictive Control baselines. The proposed method achieves the lowest total cost, minimal soft constraint penalties, and eliminates both tank overflow and ammonia demand shortfall, enabling practical, real-time operation.

Keywords: Guidance, navigation and control of aircraft and spacecraft, Condition monitoring and maintenance of aerospace systems, Aerial and space robotics
Abstract: This paper develops a proportional-integral-derivative (PID)-based sliding mode framework with an adaptive barrier mechanism for quadrotor Unmanned Aerial Vehicles (UAVs) operating under uncertainty and actuator faults. The barrier term is used to bound the states and drive them to a small neighborhood of the reference in finite time without needing upper bounds on disturbances. A PID switching manifold accelerates transient response during both the reaching and sliding phases. To limit chattering while maintaining robustness, smooth, continuous control inputs are used. Simulations of a faulted quadrotor under severe disturbances demonstrate accurate tracking, finite-time convergence of the sliding variables, and stable behavior even with reduced actuator efficiency. The results indicate improved transients, lower control effort, and better disturbance rejection compared to a standard adaptive sliding mode control benchmark.

Keywords: Nonlinear and optimal automotive control, Adaptive and robust control of automotive systems, Modeling, supervision, control and diagnosis of automotive systems
Abstract: The electro-hydraulic brake (EHB) system presents considerable control challenges due to model uncertainties, friction, and time-varying nonlinearities. This paper proposes a robust cascade control framework combining a nonlinear triple-step approach and nonlinear integral sliding mode control (NISMC). A data-driven model compensates for time-varying pressure-flow characteristics in the pressure loop. Simultaneously, an extended state observer-based NISMC actively rejects lumped disturbances in the servo loop. Hardware-in-the-loop (HIL) experiments verify the strategy, demonstrating an 18.6% faster response, 30.8% improved accuracy, and 68% reduction in cumulative error, significantly enhancing EHB system robustness.

Keywords: Optimal control theory, Adaptive control design, Uncertain systems
Abstract: This paper presents a policy iteration algorithm for adaptive robust control of nominal continuous-time linear systems with completely unknown dynamics. We utilize a dual-layer filtering architecture to enable derivative-free, on-policy learning without explicit system identification. The algorithm is shown to be inherently robust to bounded, time-varying perturbations in both state and input matrices. We derive explicit, iteration-dependent stability bounds using Lyapunov theory, proving that the algorithm remains uniformly ultimately bounded (UUB), and converge to a bounded neighborhood of the nominal optimal solution. The approach relies only on a finite-interval excitation condition and is verifiable online, ensuring a simple yet robust data-driven control framework.

Keywords: Real-time optimal control
Abstract: The paper addresses the problem of optimal control of fixed-wing unmanned aerial vehicles (UAVs) under varying terminal conditions and external disturbances. A control structure is developed based on Pontryagin's maximum principle, followed by the implementation of an algorithm for adaptive correction of the control structure parameters. Simulations of both single UAV flights and group takeoffs using the leader-follower scheme were conducted. Numerical experiments demonstrate that the developed algorithms ensure high accuracy in meeting terminal conditions and robustness to measurement noise and external disturbances. The results have practical applications in monitoring tasks and the organization of communication networks.

Keywords: numerical methods for optimal control, applications of optimal control, model predictive control
Abstract: Linear Quadratic (LQ) control problems are at the heart of linear control theory and Model Predictive Control (MPC). While performant, standard approaches to solving such problems are inherently serial, limiting real-time scalability despite the parallel computing power available on modern multi-core CPUs. Contributing to addressing this challenge and motivated by ``divide and conquer'' strategies, we present a parallel-in-time approach that solves computationally demanding conic optimal control problems through the use of the alternating direction method of multipliers (ADMM). In particular, we formulate the inner primal update of ADMM as an LQ problem and split the reformulated problem along the time horizon. This enables us to derive a variant of the Riccati recursion using dynamic programming to solve each subproblem in parallel. Numerical benchmarks on two real-world applications demonstrate as much as a 5x speedup compared to existing related approaches on multi-core CPU hardware.

Keywords: Control using FAS approach, Fully-actuated systems in industry
Abstract: Communication delay in permanent magnet synchronous motor servo systems significantly compromise system stability and tracking accuracy. To achieve high-precision motion control of servo motor subject to communication delay and external disturbances, this paper proposes a fully actuated system approach based on a state predictor. First, a lead compensated disturbance observer is designed to estimate disturbances. Second, a reducedorder Luenberger observer is developed to estimate the derivative of disturbance. Next, the disturbance and its derivative are used to compute state prediction compensation, improving prediction accuracy. Finally, based on the predicted states, a fully actuated system control scheme is designed to achieve fast tracking performance in the closed-loop system. Simulation results validate the effectiveness and superiority of the proposed method.

Keywords: Control using FAS approach, Global fully actuated systems, Fully-actuated systems in industry
Abstract: This paper proposes a high-order fully actuated (HOFA) robust control approach for the passively-tilting dual-frame hexacopter (PTDF-H) to improve its motion accuracy and stability. The PTDF-H achieves full 6-DoF control via passive tilting enabled by universal joints, optimizing energy use through internal thrust counteraction. A robust controller is designed based on the HOFA system approach, which ensures precise trajectory tracking and disturbance rejection, with stability verified via Lyapunov theory. Dynamics modeling and control allocation are derived, and simulations demonstrate the effectiveness of the controller.

Keywords: Global fully actuated systems, Control using FAS approach, Fully-actuated systems in industry
Abstract: This paper presents a robust trajectory-tracking controller for a tilt-rotor fully actuated quadrotor. Starting from a six-degree-of-freedom model that includes gyroscopic effects, tilt angles, and tilt rates, the position and attitude dynamics are reformulated as high-order fully actuated systems. A robust controller is then designed to handle bounded force/torque disturbances and the time-varying input mapping. Lyapunov analysis guarantees convergence of tracking errors to adjustable invariant sets. Simulations of circular trajectory tracking under disturbances demonstrate accurate position tracking, attitude stability, and bounded control effort, highlighting the effectiveness of the proposed approach.

Keywords: Predictive control of fully-actuated systems, Global fully actuated systems, Control using FAS approach
Abstract: This paper proposes an iterative learning predictive control (ILPC) strategy based on fully actuated system (FAS) approaches for spacecraft fly-around batch processes. In contrast to conventional approaches that focus solely on time-domain control, the proposed fully-actuated ILPC (FA-ILPC) framework extends the control strategy to time-batch dual physical domain. A time–batch mixed dynamical model for spacecraft fly-around mission is first established. By using the fully actuation of the dual physical domain spacecraft system, the double-difference operators with a regulation system is introduced, then a decoupled mixed-difference linear closed-loop predictive model is constructed. This model incorporates both historical difference information and the current states in the time and batch domains. Based on this formulation, the predictive optimization is converted into two decoupled optimization problems subject to three sets of linear matrix inequality (LMI) constraints, which ensures stability in both time and batch domains. Finally, the simulation results further verify the effectiveness of the proposed FA-ILPC strategy for spacecraft fly-around batch processes.

Keywords: Unidirectionally connected FASs, Sub-fully actuated systems, Control using FAS approach
Abstract: This paper proposes a unidirectionally connected fully actuated system (UC-FAS) approach for the sub-stabilization and tracking control of 6-DOF quadrotors, tackling limitations both in state-space and FAS framework to some extent. The framework systematically converts underactuated quadrotor dynamics into a UC-FAS model, unifying the existing different FAS transformation ways. By eliminating estimation of the high-order derivatives of control inputs, a drawback of current methods, the UC-FAS model simplifies controller design and enables direct eigenstructure assignment for closed-loop dynamics. Simulations demonstrate precise 6-DOF tracking performance. This work bridges theoretical FAS approach advancements with practical implementation needs, offering a standardized paradigm for nonlinear quadrotor control.