Keywords: Optimal control theory
Abstract: Optimal control and filtering problems have a rich history that can be traced back to Queen Dido’s solution to the isoperimetric problem in 814 BC, the solution to the brachistochrone problem by Johann Bernoulli in 1696, and the introduction of the least-squares method by Legendre and Gauss in the late 18th and early 19th centuries.

Modern solutions exploit the calculus of variations and its control perspectives, Pontryagin’s Minimum Principle, Bellman’s Principle of Optimality, and the Wiener and Kalman filters.

While for linear systems all roads lead to the Algebraic Riccati Equation and its dual, we demonstrate that for nonlinear systems one can go well beyond the HJB equation by judiciously exploiting properties of the stable invariant manifold, the associated invariant distribution, and adapted coordinates for the underlying Hamiltonian system. These, in turn, yield PDEs that differ dramatically from the HJB equation and have a deep geometrical meaning.

Finally, we propose an ansatz for developing global solutions based on a fixed-point characterization of optimal feedback strategies and on a geometric relation between the tangent space to the stable manifold and the first order approximation of the Hamiltonian vector field away from its equilibrium.

Keywords: Adaptive control design
Abstract: This is the first presentation in the IFAC WC’26 tutorial session on “Game-Theoretic Control Paradigms for Socio-Technical Networks”, which will introduce the audience to the foundations of game-theoretic methods for control, including static and dynamic games, stochastic games, information structures, and equilibrium solution concepts, with an emphasis on their relevance to socio-technical networks. Two classes of equilibria will be elaborated on, namely Nash and Stackelberg, as well as a mix of the two, where the latter will lead to design of incentives (or “soft” inducement mechanisms) and control of outcome of a hierarchical decision-making process resulting from strategic interactions among multiple agents. In this context, an introduction to “Bayesian persuasion” will be provided, which captures a game-theoretic framework where an informed sender influences a receiver’s actions by designing an appropriate information structure under which a Bayesian receiver ends up with actions that benefit the sender.

Keywords: Adaptive control design

Keywords: Adaptive control design

Keywords: Adaptive control design
Abstract: This tutorial presents cooperative and noncooperative game-theoretic frameworks for modeling, learning, and control in socio-technical systems, where human behavior, incentives, institutions, and social interactions are coupled with cyber-physical and networked infrastructures. The paper reviews strategic, dynamic, cooperative, matching, learning, and feedback-control approaches for analyzing how local decision-making, adaptation, and strategic interactions shape collective system outcomes. The tutorial further develops feedback-learning and incentive-design perspectives that connect equilibrium analysis with adaptation, distributed control, and mechanism design under information and coordination constraints.We also examine resilience and security challenges arising from adversarial behavior, misinformation, disruptions, and cascading failures in interconnected socio-technical networks. Finally, we discuss emerging research directions at the intersection of game theory, control, learning, and network science for resilient and adaptive socio-technical systems.

Keywords: Cyber-physical-social systems in enterprises, Enterprise interoperability, Model-driven enterprise-system engineering
Abstract: Cyber-physical systems (CPS) embrace cybernetic and physical components in dynamic interactions. CPS modelling involved multiple views of the system from different disciplines. Interoperability of CPS comprises coordinating data exchange and operation between heterogeneous components and systems. Due to the multidisciplinary nature of CPS, the independence of its components, and its complex behavior, interoperability approaches tend to focus on a specific level of abstraction and a single type of interoperability. Whereas a holistic view is expected to provide a more accurate representation of reality. The aim of this paper is to highlight the lack of an integrated perspective on CPS interoperability. First, we identify how the usage of different models contributes to achieving CPS interoperability. Then, we propose a pre-conceptual schema to show CPS elements and its relationships involved in interoperability from a general perspective. A complete characterization of CPS interoperability is required to include essential aspects in a unified model abstraction.

Keywords: Digital transformation, Intelligent manufacturing systems
Abstract: A digital twin is defined as an organised set of models that accurately represent a physical entity in the real world in order to meet specific industrial uses. Continuously updated using real data, it offers a level of precision and granularity tailored to operational needs. This virtual model can integrate the shapes, states, functions, processes, behaviours and dynamic data of the equipment under study, while reflecting its environment. However, precise calibration between the virtual twin and its physical counterpart remains a major challenge, mainly due to the limitations of current industrial IoT approaches, which are often costly, complex and unreliable. To overcome these constraints, this research proposes the integration of neuromorphic machine vision, a technology characterised by high temporal resolution and low latency, enabling automatic synchronisation of the digital twin via discrete event system modelling. This approach aims to reduce the gap between the virtual and the real, improve calibration accuracy and optimise operational efficiency in complex industrial environments. The study highlights the potential of event-based vision systems, combined with machine learning algorithms, to capture and interpret the behaviour of physical equipment in real time. By replacing heavy IoT instrumentation with intelligent visual observation, this method offers a more economical, robust and adaptable solution, contributing to the emergence of a more connected and efficient industry of the future.

Keywords: Human-technology integration in manufacturing, Robotics in manufacturing systems
Abstract: International organizations report concerning statistics regarding the inclusion of people with disabilities in the labor market, underscoring the need for effective inclusive solutions. Industry 4.0 has accelerated technological advances, including collaborative robots (cobots), whose design enables safer and improved interaction with human workers than non-collaborative solutions. Hence, cobots have the potential to support workers with disabilities (DWs), reinforcing the human-centric orientation emphasized in Industry 5.0 and contributing to more inclusive workplaces. This topic has attracted growing scholarly interest, with studies addressing diverse goals such as developing cobot-based assistance systems for DWs or examining user acceptance. Research also varies in the categories of disabilities and impairments, industrial applications, or cobot technologies involved. Such heterogeneity has resulted in a fragmented body of knowledge that may hinder broader implementation efforts. To address this gap, this study conducts a Systematic Literature Mapping (SLM) to review, structure, and synthesize existing research. The review showed that developing Human-Robot Collaboration (HRC) systems and improving the human-cobot alignment are the most prevalent research goals. Assembly tasks emerge as the most common application area, with frequent focus on robotic arms. The findings can support researchers in identifying promising research directions and assist practitioners in introducing cobots to better include DWs in industrial settings.

Keywords: Industrial artificial intelligence, Manufacturing plant simulation, control and optimization, Simulation and optimization in production, operations and services
Abstract: Abstract: Municipal solid waste incineration (MSWI) process is a complex industrial process characterized by high nonlinearity and nonstationary dynamics, making it difficult to achieve optimum operation. To solve this problem, a meta-knowledge transfer-based dynamic operation optimization (MKT-DOO) method is proposed for the MSWI process. First, the data stream learning is employed with online elastic weight consolidation incremental update strategy and attention mechanism to construct ensemble surrogate models. Then, the time-varying objective functions can be approximated accurately. Second, a dynamic multi-objective particle swarm optimization algorithm based on transfer learning is proposed to derive the optimal solutions of the manipulated variables. To reduce negative transfer, a meta-knowledge transfer strategy is designed to address the issue of task-specific knowledge differing significantly across transfer tasks caused by drastic fluctuations in operating conditions. Finally, the effectiveness of the proposed method operation optimization is validated by real industrial data.

Keywords: Intelligent manufacturing systems, Cyber-physical production systems, Smart production and logistics in manufacturing
Abstract: Modern manufacturing systems require control architectures capable of bridging the gap between flexible high-level planning and the immediate low-level execution of the manufacturing process. This paper proposes a Resource agent architecture that links the planning and execution layers by integrating a Belief-Desire-Intention agent into the Skill Orchestration Agent framework. Thereby, enabling agent-based planning combined with skill-based execution. A shared knowledge base, structured by the Capability-Service-Skill model, ensures semantic coherence between capabilities and skills across all control levels. This architecture enables autonomous and decentralized production planning and execution.

Keywords: Intelligent manufacturing systems, Smart production and logistics in manufacturing
Abstract: During the design of a product, shapes are defined using complex mathematical functions. However, these must eventually be approximated by lines/arc segments. Under traditional Computer-Aided Manufacturing (CAM), this is done individually for each part. Thus, the approximations can be inconsistent, which results in poor assembly. We propose a workflow and a datamodel to generate toolpaths knowing final product assembly information. This allows parts that are meant to be assembled to share common machining toolpaths. We generated 9261 part assemblies for two use cases. Results show that the Shared Approximation Method (SAM) eliminates mismatches in assemblies regardless of the approximation quality.

Keywords: Manufacturing plant simulation, control and optimization, Complex dynamic systems, Large-scale complex systems
Abstract: Extracting knowledge from observed data regarding complex systemic behavior is closely associated with system identification methodology, where inherent uncertainty in model development necessitates stochastic formulations. Addressing stochastic identification tasks requires appropriate quantifiers of statistical association among variables. The most widely used quantifier, the ordinary (Pearson) correlation, may vanish even when a deterministic functional relationship exists between the variables of interest. Dependence measures termed “consistent”, which equal zero only when two random variables are statistically independent, provide a more comprehensive representation of inter-variable relationships. However, additional considerations such as normalization constraints and compatibility with Gaussian assumptions introduce further complexity. To address these challenges, this work adopts the maximum correntropy coefficient. This measure captures affine associations between pairs of random variables and enables computationally tractable procedures for stochastic system identification. Since an affine mapping constitutes a nonlinear transformation, the systems considered should be classified as nonlinear, despite their relatively simple nonlinearity. The nonlinear behavior examined arises primarily from the complex probabilistic interdependencies among model variables. This study develops a framework for constructing piecewise-affine stochastic models, aiming to identify and precisely quantify the stochastic relationships between model inputs and outputs.

Keywords: Manufacturing plant simulation, control and optimization, Manufacturing engineering and management, Advanced manufacturing and remanufacturing technologies
Abstract: This paper provides a detailed examination of the calculation of technical and economic indicators (TEI) for nuclear power plants, focusing on methodology, algorithms, and implementation as a specialized software module for analyzing and quantifying the thermal efficiency of nuclear power plant units. The paper presents the theoretical foundations and practical aspects of using TEI to monitor the efficiency of thermodynamic conversion of thermal energy generated in the core of a nuclear reactor. Methodological approaches to TEI calculation, data processing algorithms, and methods for visualizing analytical results are considered. Particular attention is paid to assessing the energy efficiency of both individual equipment and the unit as a whole.

Keywords: Model-driven enterprise-system engineering, Enterprise interoperability, Digital transformation
Abstract: Increasing complexity of modern enterprise systems and the demand for automation and interoperability require consistent and semantically validated models in Model-Based Systems Engineering (MBSE). The Object Constraint Language (OCL) supports formal definition of such constraint validations. However, MBSE models and OCL constraints are typically managed in separate tools, causing manual effort during model constraint application and result interpretation. To address this gap, this paper proposes an approach to managing OCL constraints and their validation results through Asset Administration Shells (a well-established technology for interoperability in enterprise systems). The methodology is demonstrated through a fictional industrial scenario, and to support reproducibility, all artifacts are publicly available in a GitHub repository.

Keywords: MBSE, OCL, AAS, Semantic Constraint Modeling, AutomationML

Keywords: Reinforcement learning and deep learning in control, AI-driven modeling and control, Knowledge-based and data-driven control
Abstract: Process systems often impose several state and input constraints owing to safety and environmental limitations. There is an increasing interest in deploying reinforcement learning-based controllers to achieve the goal of autonomous process systems. Standard reinforcement learning algorithms lack the provision to impose hard state constraints. This impedes their applicability in safety-critical process systems, where constraint violations can have catastrophic consequences. To this end, we characterize the concept of safe set as a maximal control invariant set, and ensure that exploration and exploitation occur within the safe set. We propose an action replacement-based reinforcement learning approach that can effectively prevent violation of state constraints while learning the control policy. More specifically, we propose a model-based safety filter that replaces the potentially unsafe control action suggested by the conventional reinforcement learning controller with the safe control action such that the replaced control input drives the system to safe states. In this work, we integrate this safety filter with the deep deterministic policy gradient algorithm to learn the control policy. We demonstrate the efficacy of the proposed approach on a double integrator system, showing that the proposed action replacement strategy provides a safety guarantee.

Keywords: Reinforcement learning and deep learning in control, AI-driven modeling and control, Machine learning for modeling and prediction
Abstract: We propose a hybrid reinforcement and self-supervised learning approach for accelerating generalized Benders decomposition. On the master side, we employ a graph-based reinforcement learning agent that operates on a bipartite graph representation of the master problem and is equipped with a verification mechanism to either partially or fully solve it. On the subproblem side, a physics-informed neural network, trained to approximate solutions that satisfy the Karush--Kuhn--Tucker conditions via self-supervision, takes the values of the integer variables as input and produces primal--dual pairs for Benders cut construction. The proposed framework is evaluated on a mixed-integer nonlinear programming case study, where it achieves a 52% reduction in solution time relative to classical GBD while preserving convergence behavior.

Keywords: Reinforcement learning and deep learning in control, Control architecture for multi agent systems
Abstract: Offline reinforcement learning allows control policies to be learned directly from data without online interaction, making it suitable for safety-critical tasks. Recent studies have applied diffusion models to offline reinforcement learning to leverage their strong capacity for modeling complex data distributions. However, existing approaches primarily focus on single-agent settings, leaving the safety challenges in multi-agent environments largely unexplored. In this work, we propose a safe offline multi-agent reinforcement learning algorithm that embeds neural individual control barrier functions into the diffusion model to enhance safety during trajectory generation, with control policies recovered through inverse dynamics. We evaluate our algorithm across diverse benchmarks, demonstrating substantial safety improvements while maintaining competitive rewards.

Keywords: Reinforcement learning and deep learning in control, Control architecture for multi agent systems, Safety and security in networked control
Abstract: This paper proposes a primal-dual based safe multi-agent reinforcement learning (MARL) framework that integrates Transformer-driven graph neural networks (GNNs) and Lagrangian method, termed G-MATrans-Lagr, to enable safe and scalable cooperation among agents under limited communication. The approach adopts Lagrangian multipliers to optimize the reward and cost in a hybrid objective function, and a Transformer-based GNN is utilized to aggregate local observations into expressive graph representations, facilitating effective information sharing among neighboring agents. Experimental validation on multi-UAV navigation task demonstrates that G-MATrans-Lagr achieves superior performance compared with the latest MARL and safe control baselines, maintaining higher performance and lower safety costs across varying agent scales. The results showcase our method’s ability to balance efficiency and safety while enhancing scalability for complex multi-agent systems. Besides, we open source our code at https://github.com/finleygou/G-MAT-Lagr.

Keywords: Reinforcement learning and deep learning in control, Knowledge-based and data-driven control, AI-driven modeling and control
Abstract: This paper proposes a simulation-based reinforcement learning algorithm for controlling systems with uncertain and varying system parameters. While simulators are useful for safely learning control policies, the reality gap remains a major challenge. To alleviate this challenge, we propose a two-stage algorithm. First, multiple control policies are learned for systems with different system parameters in a simulator. Second, for a real system, the control policies are adaptively switched using an online convex optimization algorithm based on observations. The proposed approach mitigates the learning difficulty of training a single policy to handle all possible system parameters and enables lightweight online adaptation.

Keywords: Reinforcement learning and deep learning in control, Knowledge-based and data-driven control, AI-driven modeling and control
Abstract: We study visual domain transfer for end-to-end imitation learning in a realistic and challenging setting where target-domain data are strictly off-policy, expert-free, and scarce. We first provide a theoretical analysis showing that the target-domain imitation loss can be upper bounded by the source-domain loss plus a state-conditional latent KL divergence between source and target observation models. Guided by this result, we propose State- Conditional Adversarial Learning (SCAL), an off-policy adversarial framework that aligns latent distributions conditioned on system state using a discriminator-based estimator of the conditional KL term. Experiments on visually diverse autonomous driving environments built on the BARC–CARLA simulator demonstrate that SCAL achieves robust transfer and strong sample efficiency.

Keywords: Reinforcement learning and deep learning in control, Machine learning for modeling and prediction
Abstract: This paper proposes a memory-enhanced PPO-GRU reinforcement learning framework for autonomous beyond-visual-range (BVR) air combat under partial observability. The BVR air-combat scenario is formulated as a partially observable Markov decision process, and the framework integrates recurrent memory, progressive curriculum learning, and an auxiliary prediction module to improve long-horizon tactical decision-making under intermittent observations. Experimental results show that the proposed agent achieves an 89.2% final win rate and outperforms feedforward PPO, SAC, and DDPG baselines under the same observation, reward, and action settings.

Keywords: Remote control, Networking for internet of things, Networking for teleoperation
Abstract: This work presents a Digital Twin-enhanced tele-operated locomotion system for an articulated quadruped robot, easy-to-deploy, and designed to calibration walking movements. A geometric approach is developed to solve the inverse kinematics for a three-joint leg model, thereby accurately deriving the required joint angles from desired foot coordinates. Central to this enhancement is a digital twin implementation within CoppeliaSim software, which provides a virtual testing ground for predictive analysis and optimization of the control algorithms, significantly accelerating development and improving system robustness. Commercial servomotors, actuated based on these calculated angles, are controlled by a mobile application developed in .NET MAUI. This application facilitates remote operation and telemetry monitoring through secure MQTT communication via HiveMQ Cloud. The refined control equations, initially validated through the digital twin, are then thoroughly tested on a 3D-printed physical prototype utilizing an ESP32 microcontroller. The results show the feasibility of communication and quadruped robot calibration in runtime, while offering an integrated and scalable solution, supported by a simulation-driven physical prototyping.

Keywords: Remote control, Virtualized and cloud-based control architectures, Remote data acquisition and fusion
Abstract: This paper presents an FPGA Remote Laboratory that enables students and hobbyists to conduct real hardware experiments on a Digilent (2025) Arty Z7-20 board through a web interface. The platform combines MQTT based control, RDP virtual access, multi peripheral hardware, and live video feedback to provide a hands-on FPGA learning environment beyond simulation-only approaches. The system achieves 120–180 ms control latency and supports up to five concurrent sessions. It offers a scalable and low-cost model for remote FPGA and embedded systems education, supporting self-paced experimentation and practical understanding.

Keywords: Robotics in manufacturing systems, Industry X.0 for production and logistics, Human-technology integration in manufacturing
Abstract: The use of collaborative robots (cobots) in production systems is no longer a vision of the future, but a practical solution for human-robot collaboration. This paper provides a literature review on the role of cobots in the transition from Industry 4.0 to Industry 5.0. The review is based on Web of Science, Scopus, and Google Scholar searches using terms related to cobots, HRC, Industry 4.0/5.0, safety, HMI, mass customization, and mass personalization. The study shows how cobots connect Industry 4.0, a digitized and automation-focused industry, with Industry 5.0, a human-centered industry, by combining AI-driven customization, safe physical interaction and HMI-based operator support. From a production management perspective, implementation is also seen as a managerial and technological enabler of mass personalization, bottleneck mitigation, and manufacturing-as-a-service models. The contribution is to merge market trends, security features, and implementation logic into a conceptual argument for cobots as a driver of contemporary production transformation.

Keywords: Stochastic optimal control problems, Model predictive control
Abstract: Safe motion planning in uncertain, time-varying environments is challenging because the safe region can change unpredictably across planning steps, often causing a loss of recursive feasibility. In this work, we present a Probabilistic Recursively Feasible Model Predictive Control (PRF-MPC) framework that guarantees recursive feasibility with a specified probability. We introduce properties that an ideal predictor should satisfy to ensure distributional consistency, and use these properties to derive closed-form expressions for the means and covariances of trajectories predicted at future time steps. Building on this analysis, we construct safety constraints that ensure, with high probability, that the current safe set is contained within the safe sets at future time steps, thereby probabilistically guaranteeing recursive feasibility. Simulation results on a lane-change scenario demonstrate that the proposed method significantly improves recursive feasibility.

Keywords: Supervision and testing, Fault detection and isolation, Design methods for data-based control
Abstract: This paper presents an integrated procedure for designing, diagnosing and recovering offshore wind turbine operation under faulty conditions. The main contribution is not a stand-alone control or diagnosis algorithm, but a reproducible co-design workflow in which controller tuning, residual-based diagnosis and recovery actions are selected together and assessed against common safety and performance requirements. The procedure is applied to a benchmark floating offshore wind farm represented by an aero-hydro-servo-elastic digital twin. Candidate supervisory settings are first obtained from an energy-load trade-off. Diagnosis thresholds and isolation rules are then tuned on separate healthy and faulty scenarios, and the resulting decisions trigger recovery actions via safe derating and command reconfiguration. The complete closed loop is tested under multiple wind conditions, noisy measurements and injected sensor and actuator faults. The results show that the integrated strategy improves availability, reduces downtime and shortens post-fault recovery episodes while preserving load-sensitive operational margins. The study also clarifies how diagnostic delay, false alarms, and missed detections affect feasible recovery, thereby making the links between design choices, diagnosability and safe operation explicit. This provides a traceable route from design intent to evidence-based operation, suitable for further validation on higher-fidelity models and field data.

Keywords: Sustainable and circular manufacturing systems, Sustainable and circular supply chain and production, Cyber-physical production systems
Abstract: The transition to a Circular Economy (CE) requires structured, interoperable data across product life cycles. The Asset Administration Shell (AAS), as the Industry 4.0 digital representation standard, provides this foundation, yet CE-relevant data points remain insufficiently defined. This paper asks: How can nano-level CE data points be formally integrated into the AAS? We present a methodology to identify and classify nano-level CE data, map them to modular AAS submodels, and produce a reusable template for Digital Product Passports and digital twins. The approach enhances data exchange, supports future CE requirements, and is scalable to higher CE levels.

Keywords: System identification and adaptive control of distributed parameter systems, Backstepping control of distributed parameter systems
Abstract: We are considering the problem of observer design for heat partial difference equations (PDEs) with distributed delay in actuator. Distributed delays are generally assumed to be uniformly distributed, i.e., their kernel functions are constant and perfectly known. The main novelty of this study lies in letting the actuator delay kernel function (DKF) not to be necessarily constant or known. These considerations make the observer design problem under study a new problem never studied in the past. Making use of the backstepping design method and a suitable decoupling transformation, we develop an adaptive observer that provides online estimates of the PDE state and the actuator DKF. We first show that the L^2-norm of the DKF estimation error exponentially converges to zero, under a well-defined persistent excitation (PE) depending only on the input signal. Then, we show that the PDE state estimation error in turn exponentially converges to zero.

Keywords: AI-driven modeling and control, AI tools in automation engineering and operation, AI in networked control
Abstract: We characterize the achievable satisfaction region of real-world generative machine intelligence systems under compute, architecture, training, adaptation, and budget constraints. The result defines an alignment capacity metric that quantifies how many user preferences can be met to a target quality and frequency. By expressing this capacity as an explicit resource-allocation optimization driven by user-specific expectile utility, the theorem reveals clean Pareto frontiers between coverage, quality, and reliability, and provides sharp conditions for when universality is not achievable. The framework offers actionable guidance for maximizing user satisfaction and quality-of-experience in deployed machine intelligence systems.

Keywords: AI-driven modeling and control, Knowledge-based and data-driven control, Machine learning for modeling and prediction
Abstract: In this paper we propose a novel physics-informed neural network framework for solving general first-order delay differential equations. Our approach combines a differentiable history switch, a trial-solution formulation that explicitly enforces history constraints, and a segmented collocation strategy to stabilize gradient propagation across large temporal domains. The method enables a scalable and physics-consistent approximation of delay differential equation solutions while maintaining continuity across subintervals. Numerical experiments demonstrate the effectiveness of the proposed method.

Keywords: AI-driven modeling and control, Machine learning for modeling and prediction, Knowledge-based and data-driven control
Abstract: We introduce Perron--Frobenius Operator Matching (PFOM), a generative framework that matches density evolution via the integral PF operator, subsuming flow, diffusion, and jump models. We prove that among Bregman divergences, only Kullback--Leibler divergence preserves equality between density-level and sample-conditioned objectives, yielding a practical loss equivalent to Koopman path matching. We further develop Nesterov-accelerated training and sampling that stabilize discretization and accelerate convergence. %On Gaussian mixtures and two-moons, PFOM achieves faster KL/W_2/MMD decrease and improved wall-clock efficiency with empirical validation. PFOM unifies operator-theoretic identification with modern generative modeling and opens paths to adaptive dictionaries and high-dimensional applications.

Keywords: AI-driven modeling and control, Machine learning for modeling and prediction, Reinforcement learning and deep learning in control
Abstract: The transition from monolithic to multi-component neural architectures in advanced neural network controllers poses substantial challenges due to the high computational complexity of the latter. Conventional model compression techniques for complexity reduction, such as structured pruning based on norm-based metrics to estimate the relative importance of distinct parameter groups, often fail to capture functional significance. This paper introduces a component-aware pruning framework that utilizes gradient information to compute three distinct importance metrics during training: Gradient Accumulation, Fisher Information, and Bayesian Uncertainty. Experimental results with an autoencoder and a TD-MPC agent demonstrate that the proposed framework reveals critical structural dependencies and dynamic shifts in importance that static heuristics often miss, supporting more informed compression decisions.

Keywords: AI-driven modeling and control, Reinforcement learning and deep learning in control, Cyber physical systems
Abstract: This paper presents a unified benchmarking framework to compare model-free reinforcement learning (RL) controllers on a nonlinear cyber-physical system (CPS) under false data injection and denial-of-service attacks. Four reward functions—exponential, progressive, Lyapunov-descent, and linear are analysed across two controller architectures (RL-PID,RL-MPC) and two learning algorithms (PPO, DDPG) using eight Key Performance Indicators covering tracking error, computational cost, and resilience. The Lyapunov reward yields the best resilience and lowest tracking error; the exponential mode provides a strong accuracy–robustness trade-off. Progressive and linear rewards converge faster but are less robust under attacks. RL-MPC achieves superior steady-state resilience, whereas RL-PID requires significantly less training time and is better suited for embedded deployment. These results demonstrate that reward shaping is a central design lever for model-free RL in CPS security, and provide actionable guidance for practitioners selecting controller and reward configurations.

Keywords: Cloud control and robotics, Networking for teleoperation
Abstract: This article presents a novel framework for real-time Light Detection and Ranging (LiDAR) data transmission that leverages rate-adaptive technologies and point cloud encoding methods to ensure low-latency and low-loss data streaming. The proposed framework is intended for, but not limited to, robotic applications that require real-time data transmission over the internet for offloaded processing. Specifically, the Low Latency, Low Loss, Scalable Throughput (L4S)-enabled SCReAM v2 transmission framework is extended to incorporate the Draco geometry compression algorithm, enabling dynamic compression of high-bitrate 3D LiDAR data according to the sensed channel capacity and network load. The low-latency 3D LiDAR streaming system is designed to maintain minimal end-to-end delay while constraining encoding errors to meet the accuracy requirements of robotic applications. We demonstrate the effectiveness of the proposed method through real-world experiments conducted over a public 5G network across multi-kilometer urban environments. The low-latency and low-loss requirements are preserved, while real-time offloading and evaluation of 3D SLAM algorithms are used to validate the framework’s performance in practical use cases.

Keywords: Control software architecture, Information models for control engineering, Virtualized and cloud-based control architectures
Abstract: A common assumption when designing control systems is periodic sampling and actuation. As a consequence of this periodicity, unnecessary control delays may be caused. In this paper we show how performance can be improved if, rather than waiting for periodicity, control systems actuate and sample as soon as possible. The performance evaluations are done using stochastic analysis of a large number of processes, comparisons to continuous controllers in simulations, and implementation on a ball and beam system.

Keywords: Cyber physical systems, AI tools in automation engineering and operation, Knowledge-based and data-driven control
Abstract: This paper evaluates the capacity of off-the-shelf Large Language Models to infer human-interpretable cyber-physical system states from multivariate time-series data in a zero-shot setting. Using the JIGSAWS surgical benchmark, we prompt the model with lightweight per-state kinematic summaries. Across tasks, these summaries produce consistent, though modest, improvements in semantic alignment, as reflected by cosine similarity and ranking metrics. The effects are strongly task-dependent, yet the observed performance gains indicate that LLMs do extract meaningful structure from kinematic time series despite the absence of domain adaptation or supervision.

Keywords: Digital twins for cyber physical systems, AI tools in automation engineering and operation
Abstract: This paper presents an adaptive framework that integrates Machine Learning Operations (MLOps) with the Asset Administration Shell (AAS) to maintain the reliability of Alarm Flood Classification (AFC) models under changing alarm configurations. The AAS serves as a vendor-independent interface for semantically typed configuration revisions and change events, which automatically trigger a change-aware MLOps pipeline for AFC model evaluation, retraining, and redeployment. Alarm data are regenerated using the updated configuration and compared with prior results, while models are selectively redeployed based on performance thresholds. Experiments on two industrial datasets with 200 perturbed configurations demonstrate that static models degrade strongly with increasing configuration change, whereas the proposed method maintains stable accuracy while reducing unnecessary retraining by up to 30%.

Keywords: Digital twins for cyber physical systems, Cyber physical systems, Remote data acquisition and fusion
Abstract: A real-time method for cyber-attack detection based on zonotopic state estimation is presented in this work for a smart cyber-physical system with energy management. The proposed approach employs set-based zonotopic Kalman filtering to explicitly account for bounded process and measurement uncertainties while ensuring consistency under adversarial conditions. By combining residual bound violation with secure control logic, the method enables reliable attack detection and prevents the propagation of corrupted data into the energy management and relay actuation layer. The proposed work is validated through real-time experimental results, which demonstrate improved attack detection, reduced false alarms, and secure energy management operations in the presence of cyber attacks.

Keywords: Digital twins for cyber physical systems, Cyber physical systems, Soft computing and robust intelligent control
Abstract: Digital Twins (DTs) are increasingly invoked to pilot Systems-of-Systems (SoS), yet how they are built and what value they actually deliver at SoS scale remains unclear. We review 19 studies to examine scope, implementation, application domains, complexity drivers, DT roles, and supporting properties for SoS piloting. No study reports a fully implemented SoS-wide DT, i.e. most replicate only parts. Roles concentrate on experimentation–simulation and control–orchestration, with governance and assurance rising, while pure monitoring is rare. We identify interoperability, composition and SoS-level Verification and Validation (V&V) as key gaps and propose a role–capacity crosswalk and metrics to guide future deployments.

Keywords: Digital twins for cyber physical systems, Fuzzy and neural systems in control, Knowledge-based and data-driven control
Abstract: Digital twins (DTs) are a key enabler of Industry 5.0's objective to reconcile operational performance with sustainability and human well-being. However, there is no widely adopted and reproducible evaluation framework for assessing the contributions of DT to these objectives. To address this gap, we first conducted a systematic literature review to identify current practices and limitations, then present a practical, modular six-step evaluation framework that calculates a single, interpretable score for a DT instance by jointly evaluating three explicit pillars: sustainability (environmental, economic, and social), resilience, and human-centricity. The framework combines expert elicitation using a triangular fuzzy number analytical hierarchy process (TFN-AHP), objective weighting using Shannon entropy, and epistemic uncertainty modeling through spherical fuzzy sets. An optional PROMETHEE II module enables pairwise ranking across alternatives. We demonstrate the robustness of the framework through a sensitivity analysis and five synthetic case studies, with all datasets and evaluation scripts published to support reproducibility.

Keywords: Expert systems and cognitive-based control, AI tools in automation engineering and operation, Intelligent human-machine interaction
Abstract: The use of Bayesian Optimization (BO) to tune engineering systems is increasing. Conventional BO requires an objective function, which is often difficult to define and rarely captures expert judgment. Preferential Bayesian Optimization (PBO) addresses this limitation by using preference selections. We show that, after applying PBO, a data-driven cost function can be extracted that captures expert preferences, removing the human operator from the loop when safety constraints are well-defined and enabling fully automated tuning while still emulating expert decision-making. By mapping from well-defined features rather than raw control settings, this cost function becomes transferable across operating conditions, provided that the new conditions remain sufficiently covered in the feature space.

Keywords: Intelligent human-machine interaction
Abstract: This paper presents the Sensing Pod, a compact on-device AI sensor node integrating fall detection, localization support, and wake-word recognition for indoor service environments. Low-resolution thermal and audio data are processed entirely on-device using lightweight learn ing pipelines, enabling real-time inference while preserving user privacy. IR-marker signaling improves robot localization without additional hardware. In addition, centroid-based thermalfeatures enable reliable identification of user falls, and a robust three-class wake-word model ensures dependable voice activation under natural pronunciation variability. These results demonstrate that practical safety monitoring and human–robot interaction can be achieved with low-cost sensors, making the Sensing Pod a scalable infrastructure component for future service-robot deployments.

Keywords: Intelligent human-machine interaction, Data fusion and mining in control, Information models for control engineering
Abstract: With the increase of space debris, efficient spacecraft damage detection and assessment have become increasingly important. This study proposes a hypervelocity impact damage identification method based on multi-objective feature extraction. An adaptive classification algorithm driven by transient thermal response (TTR) density information is first used for unsupervised separation of different damage types. A multi-objective optimization model is then established to balance intra-class representativeness and inter-class difference, where MOEA/D with dynamic weight vector adjustment is adopted to optimize typical TTRs under an irregular Pareto front Finally, the selected high-quality TTRs are used to reconstruct infrared images. Experimental results demonstrate that the proposed method enhances defect features and improves image discriminability for spacecraft damage assessment.

Keywords: IT/OT-security in automation systems, AI tools in automation engineering and operation, Service-architectures for control systems
Abstract: Industrial Control Systems (ICS) need security measures aligned with evolving regulations, but manually linking laws, standards, and threat intelligence is slow and inconsistent. We propose an automated framework integrating the Cyber Resilience Act, IEC 62443, and MITRE ATT&CK for ICS into an accountable database via Latent Dirichlet Allocation (LDA), providing the knowledge base for Retrieval-Augmented Generation (RAG) of countermeasures. On a ground truth of 3,330 candidate pairs labeled by three-LLM consensus, the LDA-based linkage achieves Recall@5 of 0.527 (law--standard) and 0.454 (standard--countermeasure), outperforming BERT-base by 11.3 and 18.5 points respectively at lower computational cost and higher interpretability.

Keywords: IT/OT-security in automation systems, Safety and security in networked control
Abstract: Growing regulatory pressure increases the need for field‑level certificate management. In modular plants, operators typically integrate only a module-level interface, breaking the implicit assumption of direct connectivity between field devices and plant public key infrastructure assumed in current solutions. This paper examines whether overlay networks can provide a lightweight, decentralized substrate for zero‑touch certificate management within modules. Classical overlays are evaluated, and three (Chord, Kademlia, CAN) were selected for a proof of concept assessing resource efficiency and feasibility for automation systems. The results show that overlays provide a viable, protocol‑independent foundation for certificate management in modular plants.

Keywords: Knowledge-based and data-driven control, AI-driven modeling and control, Machine learning for modeling and prediction
Abstract: Monotonicity is a key requirement for trustworthy machine learning in high-stakes applications, where predictions must align with domain knowledge and human intuition. While deep neural networks excel at modeling complex non-linear relationships, they lack inherent guarantees of monotonic behavior. Existing approaches enforce monotonicity through either hard architectural constraints, which limit expressiveness, or soft regularization penalties, which lack robust guarantees. We introduce Mixup Buffer, a training technique that significantly enhances soft monotonicity enforcement by maintaining a dynamic replay buffer of synthetic constraint-violating samples. By forcing the model to repeatedly confront its worst violations through targeted retraining, Mixup Buffer drives optimization toward solutions with superior monotonic compliance. Extensive experiments across five benchmark datasets demonstrate that Mixup Buffer achieves state-of-the-art monotonicity performance for a soft optimization approach, both in-distribution and out-of-distribution, without sacrificing predictive performance.

Keywords: Knowledge-based and data-driven control, Bio-inspired algorithms and optimization-based control
Abstract: In interactive systems, feedback is often provided as preferences over queried options rather than precise scores. In this work, we propose a preference-based optimization algorithm that relies on noisy two-point comparisons. At each iteration, the algorithm employs a uniform-sphere perturbation to generate a perturbed action and queries the resulting loss comparison to estimate a descent direction. We demonstrate that, under standard smoothness and bounded variance assumptions, the algorithm converges to a stationary point when the smoothing and step size parameters are properly chosen. Numerical experiments on an LQG system demonstrate the effectiveness of the preference-based optimization algorithm with comparison feedback.

Keywords: Machine learning for modeling and prediction, AI-driven modeling and control
Abstract: Due to the scarcity of labeled data and inherently nonlinear, time-varying dynamic nature of industrial processes, achieving accurate prediction of key variables remains a major challenge. To address scenarios with only a few labeled samples but numerous raw measurements, we propose a semi-supervised collaborative masking and regularization-driven (SS-CMR) model for industrial soft sensor. We first design a dual-view masked autoencoder to emulate realistic missing-data patterns and learn robust temporal representations via self-supervised learning. During fine-tuning, a random clustering-based regularization strategy is introduced to further stabilize the latent space and mitigate overfitting. In addition, a hybrid predictor combining a deep neural network and a factorization machine is constructed to jointly capture nonlinear dependencies and interactive effects among process variables. We evaluated the performance of SS-CMR on an industrial study. The results show that the proposed approach consistently outperforms existing methods, confirming its effectiveness as a promising soft sensor solution under label-limited conditions.

Keywords: Machine learning for modeling and prediction, AI-driven modeling and control, Intelligent human-machine interaction
Abstract: Steel-surface defect detection is crucial for quality control in industrial manufacturing. However, prevailing object detection models based on deep learning still struggle with defects with large range of scale variation, moreover, pooling-based down-sampling often erases fine details and causes missed detections, especially when the defects have high similarity to the normal background. To address these issues, we propose Wavelet-DilatedNet, a novel detection framework that introduces two plug-and-play modules on top of the DEIM-DFINE-n baseline. (i) A Multi-Layered Dilated Reparameterized Convolution (MDRC) module which captures multi-scale defect features by fusing parallel dilated convolutions with re-parameterization. (ii) A Two-Stage Wavelet Transform Down-sampling (TWTD) module that cascades Haar wavelet decomposition and inversed Haar wavelet transform to preserve weak edges and textures during feature reduction. Besides, experiments on the high-resolution public dataset GC10-DET show that Wavelet-Dilated Net achieves 37.1% mAP@50:95 and 72.1% mAP@50, surpassing the baseline by 2.6% and 5.8%, respectively, while outperforming other state-of-the-art methods.

Keywords: Model driven engineering of control systems, Information models for control engineering, Control software architecture
Abstract: This paper is aimed at practitioners and offers an analysis of the effect of sampling-time jitter, i.e. the error produced by execution-time inaccuracies. It proposes a reinterpretation of jitter-afflicted linear time-invariant systems as equivalent jitter-free analogs. By constructing a perceived system that absorbs the effects of timing perturbations into its dynamics, we find an affine scaling of the system matrices with respect to jitter. Moreover, in the Laplace domain, jitter can be interpreted as a frequency scaling. The main result of this paper shows that the effects of jitter can be transferred to a time-variation of the continuous system dynamics. Consequently, the overall system can be analysed by the standard sampled-data control theory with constant sampling period, which is demonstrated by the robustness analysis of feedback loops with jitter.

Keywords: Reinforcement learning and deep learning in control
Abstract: Normalizing Flow (NF) models have recently emerged as a powerful class of generative models capable of learning expressive probability distributions through invertible transformations. In Reinforcement Learning (RL), most of the modern algorithms rely on distributions typically parameterized as Gaussian or deterministic. While these choices facilitate tractable optimization, they can severely limit the expressiveness of learned policies. In environments where optimal behaviors require multimodal action distributions, such restrictions can hinder both learning efficiency and final performance. A promising way to address these limitations is through more flexible generative models that can accurately capture complex probability distributions. This study investigates the application of Normalizing Flow architectures to RL tasks, both in single-agent and multi-agent environments. In particular, it is assessed that NFs are capable to model policies that converge to the Nash equilibrium in a two-player zero-sum game scenario, unlike deterministic policies.

Keywords: Reinforcement learning and deep learning in control, AI-driven modeling and control, AI tools in automation engineering and operation
Abstract: Reinforcement learning (RL)-based controllers provide excellent control characteristics for power-output stabilization of wind turbines but require large training datasets, while LQR controllers are suboptimal away from the linearization point. This paper proposes a hybrid collective pitch control (CPC) architecture combining an LQR and Twin Delayed Deep Deterministic Policy Gradient (TD3) controller. The LQR controller guides the TD3 agent during training, while the TD3 controller learns to compensate for the nonlinear dynamics not captured during linearization. Results show that the LQR+TD3 hybrid controller improves performance and reduces steady-state error compared with individual LQR and TD3 controllers.

Keywords: Applications of optimal control, Learning methods for optimal control
Abstract: This paper proposes a novel framework for safe and coordinated multi-agent navigation under communication constraints. Traditional multi-agent reinforcement learning methods often struggle to ensure safety and coordination in partially observable environments with limited bandwidth. The proposed R-MADDPG–HORCBF framework integrates Recurrent Multi-Agent Deep Deterministic Policy Gradient (R-MADDPG) with High-Order Robust Control Barrier Functions (HORCBFs). Specifically, a recurrent actor-critic network is employed to capture temporal dependencies, while a differentiable RCBF layer is incorporated to enforce safety constraints in real time. Simulation results in multi-vehicle navigation scenarios demonstrate that the proposed framework significantly enhances both safety and communication efficiency, highlighting its strong potential for real-world deployment in safety-critical systems.

Keywords: Applications of optimal control, Numerical methods for optimal control, Application of nonlinear analysis and design
Abstract: The increasing deployment of wind turbines and the limited availability of suitable installation areas motivate the integration of multiple wind-energy-harvesting technologies. Airborne Wind Energy Systems (AWES), capable of accessing high-altitude wind resources, offer a promising complement to conventional Horizontal-Axis Wind Turbines (HAWTs). This work presents an optimal path-planning algorithm for AWES operating within the wake of HAWTs. A simplified wake model is employed to estimate wind speed deficits behind the turbine and is incorporated directly into the trajectory optimization scheme. Simulation results show that lemniscate flight paths exhibit less sensitivity to wake effects compared to circular trajectories. The results demonstrate the potential of wake-aware path planning to improve AWES performance in multi-technology wind farm environments.

Keywords: Applications of optimal control, Optimal control of hybrid systems, Fault detection and isolation
Abstract: To address the low efficiency of fastener assembly inspection for aircraft power distribution boxes, the reliance on manual expertise, and the poor adaptability to small targets and diverse assembly specifications, this paper presents a two stage automatic inspection method that combines an RT-DETR based detection network with template comparison. We build a dataset of 4,125 images of power distribution box fasteners, use RT-DETR to obtain class labels and bounding box priors for each assembly position, and design a global image matching method constrained by keypoints and annotation boxes to align template boxes with detection results and perform consistency assessment. Experiments show that the RT-DETR detector achieves an mAP50 of 0.9925 on the constructed dataset, with mean precision and recall of 0.9862 and 0.9844, respectively. Experimental results on multi view inspection images show that the proposed framework can reliably identify missing and misinstalled fasteners and reduce reliance on manual inspection, indicating strong potential for engineering applications.

Keywords: Applications of optimal control, Real-time optimal control, Application of nonlinear analysis and design
Abstract: This paper investigates state-dependent Riccati equation (SDRE) feedback for practical self-righting of an asymmetric two-link falling-cat model. The velocity-input nonholonomic model is augmented with virtual angular-acceleration inputs to better align the control layer with torque-driven actuation. Three state-dependent coefficient (SDC) parameterizations are constructed, and their pointwise controllability conditions are characterized through a PBH-based analysis. Comparative simulations for a static-drop maneuver show that the parameterization preserving the dominant spin dynamics yields faster convergence and smoother inputs, whereas the alternatives either fail near the zero-velocity manifold or violate the bending constraint.

Keywords: Control of complex systems, Design methods for data-based control, Linear systems
Abstract: Approximate simulation-based hierarchical control (ASHC), in brief, is a technique used for simplifying the control design of a complex system with an a priori known output discrepancy bound. Current ASHC methods are based on state feedback, which hinders the possibility of developing data-driven enhancements. To overcome this difficulty, in this paper, we present a novel output-feedback ASHC framework when online state feedback is not possible. Furthermore, we propose a direct data-driven enhancement. While the proposed data-driven results still rely on the state data, the results of this paper can be seen as a stepping stone in developing a fully input-output data-driven method for solving the ASHC problem. All results are illustrated by means of a numerical example.

Keywords: Control of complex systems, Parametric optimization, Robustness analysis
Abstract: A state space pole placement approach is proposed to design PID controllers for high-order processes. The method makes use of a single parameter to determine the locations of the closed-loop poles, thus a (high-order) PID controller can be tuned with this parameter. Tuning rules of PID/PIDD2 controllers are then derived for typical stable, integrating and unstable process models. The tuned rules are applied to the benchmark processes. Simulation results show that the tuning rules can achieve compromise among disturbance rejection, robustness, and noise attenuation.

Keywords: Differential or dynamic games, Control of hybrid systems, Optimal control of hybrid systems
Abstract: Many real-world systems, such as robotics and cyber defense, rely on hierarchical decision processes where a strategic layer sets long-term configurations and a tactical layer executes fast-time actions, leading to a leader–follower structure with asymmetric information and temporally coupled interactions that may fall outside classical Stackelberg models. To address this gap, we introduce hylomorphic dynamic programming (HDP) for hierarchical control. HDP operates between an anamorphism, which unfolds strategic choices into tactical consequences by solving inner dynamic programs, and a catamorphism, which folds tactical outcomes into strategic values. This hylomorphic recursion provides a consistent and computationally tractable framework of the associated dynamic Stackelberg equilibrium.

Keywords: Differential or dynamic games, Optimal control theory, Analytic design
Abstract: This paper proposes a comprehensive analysis framework and optimal strategies for the Attacker-Defender-Target (ADT) differential game. The game involves three agents with simple kinematic models, where the attacker has a speed advantage. Based on Pontryagin’s minimum principle, this paper establishes a unified Hamiltonian framework for both scenarios where the attacker wins and the defender wins. The study proves that each agent's optimal strategy manifests as constant-velocity rectilinear motion towards a specific interception point. Drawing upon the geometric theory of Apollonius circles, analytical equations for determining the optimal interception point are derived. Furthermore, by analyzing the relative positions of the two Apollonius circles—between the attacker and defender, and between the attacker and target—this paper provides strict geometric criteria for dividing the game’s winning regions.Finally, numerical simulations are implemented to validate the theoretic results.

Keywords: Differential or dynamic games, Real-time optimal control, Stability of nonlinear systems
Abstract: This paper addresses a zero-sum differential game with a quadratic cost functional for controlling nonlinear plants under bounded disturbances, modelled by ordinary differential equations with state feedback. A diffeomorphic coordinate transformation linearizes the system, yielding a model with constant parameters and a transformed cost functional featuring state-dependent weighting matrices. Optimal strategies are derived from the Bellman–Isaacs equation, which leads to a state-dependent Riccati-type equation. In the infinite-horizon case the problem reduces to a state-dependent Riccati equation (SDRE), which is solved numerically, yielding a suboptimal regulator that guarantees asymptotic stability. The control and disturbance inputs are combined into a single regulator, and the inverse transformation recovers the original controls. An example based on the Lotka–Volterra predator–prey model illustrates the effectiveness of the proposed method.

Keywords: Fault detection and isolation, Observer design, Application of nonlinear analysis and design
Abstract: This paper presents a framework for fault estimation and root–cause identification (RCI) in nonlinear systems that avoids the structural difficulties of nonlinear unknown–input observers. We construct a collapsed model that merges nonlinearities and unknown faults into aggregated input channels, and propose a robust L_2 filter to estimate the resulting lifted state. We show that the lifted dynamics remain well posed and that filter existence requires only a weak zero-frequency input-observability condition, milder than full input observability. Individual fault components are then recovered through simple algebraic extractor maps. For RCI, we introduce a dictionary-based filter that compares the estimated trajectory against a library of candidate fault signatures and scores each by how well it explains the observed fault behaviour. The approach is illustrated on a three-tank benchmark.

Keywords: Fault detection and isolation, Positive linear systems, Observer design
Abstract: The paper deals with the properties of a fault detection filter when applied to a class of continuous-time linear systems with dynamics specified by a system matrix with an overlapped ostensible Metzler structure. The proposed solution reduces to the use of diagonal stabilization in the synthesis of the state observer and uses orthogonal transformation to construct a model with reduced order dynamics in the form of an ostensible Metzler matrix and the separation principle to generate a hidden strictly Metzler matrix for the synthesis conditions. This approach creates a unified framework that covers the compactness of parametric constraints on Metzler matrices and their diagonal quadratic stability. Using a structural model of a fixed-wing unmanned aerial vehicle to validate the method shows that the proposed approach provides high sensitivity of the fault detection filter for actuator fault detection.

Keywords: Large-scale and networked optimization problems, Convex optimization
Abstract: This paper addresses distributed composite optimization, where standard algorithms suffer from significant communication overhead and computational burden. We propose DC-ADMM-LU, a novel framework that achieves both communication and computation efficiency through local updates. The key innovation is leveraging ADMM's variable splitting to decouple the expensive proximal operator from frequent local computations, while each client performs multiple lightweight, explicit update steps. An integrated variance-reduction mechanism ensures rigorous error control across local iterations. We establish the first linear convergence guarantee for multi-step local-update ADMM in the distributed stochastic setting, without restrictive assumptions. Numerical experiments confirm superior performance.

Keywords: Learning methods for optimal control, Adaptive control design
Abstract: This paper proposes a novel adaptive learning control approach for attitude tracking of unmanned aerial vehicles (UAVs) subject to safety constraints and unknown disturbances with relaxed persistence of excitation (PE) conditions. We first formalize the robust optimal attitude tracking problem with a zero-sum game structure. Then, a modified reward function that consists of a control barrier function (CBF) is presented, which prevents the system states from violating the prescribed safety boundaries. To solve this optimization problem, a critic adaptive dynamic programming (ADP) framework is employed to approximate the solution of Hamilton-Jacobi-Issac (HJI) equation, thus obtaining the approximated optimal control. Unlike the existing gradient-descent learning methods, we transform the weight learning problem into a parameter estimation problem, which is further solved by a novel estimator design using dynamic regression extension and mixing (DREM) and generalized parameter estimation based observer (GPEBO) techniques. The main advantage of this method lies in that it not only relaxes the strict PE conditions for parameter convergence but also provides specific implementation solutions, thereby enhancing its applicability in real-world scenarios. Rigorous theoretical analysis and numerical simulations demonstrate the effectiveness and superiority of our proposed method.

Keywords: Learning methods for optimal control, Numerical methods for optimal control, Applications of optimal control
Abstract: A physics-informed neural network (PINN) approach for solving hyperbolic infinite-horizon Hamilton--Jacobi--Bellman (HJB) equations arising in nonlinear optimal regulator problems is proposed in this paper. The method simultaneously learns the value function and the optimal feedback control law through two coupled neural networks, trained to satisfy the continuous-time HJB equation and the optimality conditions for the control. We then apply the method to the closed-loop control of a quadrotor UAV and a high-dimensional reduced model of a nonlinear heat equation. The proposed PINN approach proves capable of overcoming the curse of dimensionality problem. Finally, the application of the proposed PINN approach is discussed for solving the optimal nonlinear estimation problem.

Keywords: Learning methods for optimal control, Optimization-based estimation and control, Adaptive control design
Abstract: This article proposes a novel predefined--time observer--identifier--based optimal tracking control framework for robotic systems with unknown states and uncertain dynamics subject to prescribed state constraints. Till now, most of the existing results on optimal control approaches for uncertain robotic systems require full--state information in the identifier and controller design, which is often invalid in practical scenarios. To address this issue, a predefined--time dynamic regression extension and mixing (PTDREM) method is proposed to design an observer--identifier that can simultaneously estimate unmeasurable system states and uncertain model parameters. Then, a new predefined--time prescribed performance control (PTPPC) scheme is developed under the framework of optimized backstepping technique. With this scheme, the tracking error is guaranteed to be constrained to a prearranged vicinity of origin within a predefined time. In contrast to previous studies, the proposed framework not only achieves the convergence of all closed-loop signals, but also allows that the upper bounds of convergence time for the observer--identifier and controller can all be adjusted through separate design parameters, thus ensuring global predefined--time stability (GPTS). Finally, simulation results demonstrate the effectiveness of the proposed observer--identifier--based control method.

Keywords: Learning methods for optimal control, Stability of nonlinear systems, Uncertain systems
Abstract: Despite the widespread use of PID controllers in engineering practice, designing optimal PID parameters has long been regarded as a challenging problem in both theory and practice, particularly when faced with uncertain nonlinear dynamical systems. Based on the authors' PID control theory established recently for MIMO nonlinear uncertain systems (Zhao and Guo, 2022), which provides a concrete PID parameter set for global stability of PID controlled systems, this paper further proposes a near-optimal PID tuning method, where only input-output (zeroth-order) data on the control performance is available. The tuning method is formulated as a constrained optimization problem and solved by an iterative learning algorithm, referred to as HRS-KW algorithm, that combines a hysteretic random search with the Kiefer–Wolfowitz algorithm, aiming at utilizing the advantages of both global exploration and local gradient acceleration. This method operates without requiring precise structural knowledge of the system dynamics, yet its almost sure convergence to an epsilon-optimal solution for the PID parameters can be guaranteed in theory while ensuring closed-loop system stability. Simulation results illustrate that our HRS-KW algorithm outperforms other related optimization methods, exhibiting better convergence to the prescribed epsilon-optimal performance set.

Keywords: Linear systems
Abstract: This paper proposes a computational procedure for designing a static output feedback (SOF) gain matrix for multi-input multi-output (MIMO) systems using a continuation (or homotopy) method. We regard the characteristic equations for the closed-loop SOF system as simultaneous nonlinear equations with respect to the gain elements for a given set of desired poles. We then derive differential equations from the characteristic equations based on the continuation approach. By integrating the differential equations from known initial poles to desired poles, we can obtain a gain matrix that assigns the closed-loop poles to the desired ones. From the rank of a derivative matrix in the differential equation, we can know if all or part of the designated closed-loop poles are assignable. The method is also extended to dynamic control design, particularly PID control. The effectiveness of the proposed procedure is demonstrated through flight control design for an unstable aircraft and its numerical simulation.

Keywords: Linear systems, Control in neuroscience, Optimal control theory
Abstract: Electrical brain stimulation relies on externally applied currents to modulate neural activity, but safety constraints require each stimulation cycle to be charge-balanced, enforcing a zero net injected charge. However, how such charge-balanced stimulation works remains poorly understood. This paper investigates the ability of charge-balanced inputs to steer state trajectories in discrete-time linear systems. Motivated by both open-loop and adaptive neurostimulation protocols, we study two practically relevant input structures: periodic (repetitive) charge-balanced inputs and non-repetitive charge-balanced inputs. For each case, we derive novel reachability and controllability conditions. The theoretical results are further validated through numerical demonstrations of minimum-energy control input design.

Keywords: Linear systems, Controller constraints and structure, Linear time-delay systems
Abstract: All stabilizing PID controllers form a set in three-dimensional space. A novel viewpoint to its boundary as a ruled surface (or surfaces) being cut with 3D planes is presented. The characterization, being not too new, contributes to an understanding of the stability set as the whole, instead of the classical view as a stack of 2D slices, say, on the P-coefficient. The viewpoint gives clear insight on the structure of the PID stability region, and, in particular, splits its boundary into continuous parts. It is followed by natural 2D unwrapping of the stability set boundary. It also correctly handles pure imaginary zeros in transfer function. A wireframe 3D visualization reveals the structure of the stability set. The presentation is valid both for ideal and filtered PID controllers, as well as for time-delay systems and other linear systems. Finally, based on the viewpoint, a simple formula for stability (fragility) radius is provided.

Keywords: Linear systems, Optimal control theory
Abstract: Addressing the strong dynamic coupling between the looper and gauge control systems in hot rolling, this paper proposes a coordinated control scheme based on an enhanced inverse linear quadratic (ILQ) theory. The proposed design systematically constructs the adjustable gain matrix Π and establishes an autonomous optimization framework integrating swarm intelligence. Furthermore, disturbance observer-based robust control (DOBRC) is innovatively incorporated, forming a composite control architecture. Simulation results demonstrate that the proposed scheme significantly improves the suppression of external mismatched disturbances and enhances robustness against model uncertainties.

Keywords: Linear systems, Robust estimation, Uncertain systems
Abstract: This research focuses on fragility analysis of PID controllers. The problem considered is to find a complete stabilizing set for each parameter of a given PID controller. The proposed solution is based on the classical frequency-domain stability criterion -- the Nyquist criterion. The procedure utilizes a known robust analysis method, the so-called ``breaking by parameter'' technique, which enables the study of robust (here, stabilizing) properties for an individual system parameter. Applying this technique to PID controller parameters solves the fragility analysis problem. The main result is presented as an analytical procedure for individual PID parameters.

Keywords: Linear systems, Structural and geometric control, Numerical methods for optimal control
Abstract: This work aims to provide numerically efficient computational techniques for recent results from data-driven geometric control. First, an overview of recent results on the data-driven disturbance decoupling problem (D4P) from (Naveen Mukesh et al., 2025) is presented. These results use multiple noisy output trajectories collected from the system instead of system matrices. Then, numerically efficient subspace computational methods that use only input-output data are developed to verify the solvability condition for the disturbance decoupling problem (DDP). The proposed numerical method uses the LQ decomposition to perform the required subspace computations. Subsequently, from the ``noisy'' output data, the largest controlled invariant subspace contained in the nullspace of the output matrix and a corresponding feedback matrix that solves the DDP are also computed numerically using LQ decomposition. Lastly, efficient computation techniques for computing the largest controlled invariant subspace contained in the nullspace of the output matrix and the smallest conditioned invariant subspace containing the range space of the input matrix, from exact noise-free data collected from the system, are presented.

Keywords: Linear time-delay systems
Abstract: The spectrum of a discrete-time delay system (DTDS) with linear-time-invariant (LTI) dynamics is of the discontinuity nature, when the delay tau is treated as a free parameter. This is a long-standing obstacle for directly keeping track of the stability property in the whole delay parameter space. This work proposes an intuitive frequency-domain framework to solve this problem. First, we construct the characteristic entire function for a DTDS, whose spectrum has the equivalence relation with that of the characteristic function. Second, we propose the continuity property of unstable roots for the characteristic entire function. Therefore, the spectrum of the characteristic function is replaced by that of the characteristic entire function, and the discontinuity issue is fully solved, which allows for an available and direct way to study the stability w.r.t. a free tau. Finally, within our new framework, a general idea for analyzing the stability in the whole delay parameter space, the tau-decomposition idea for DTDS, is provided.

Keywords: Task and motion planning, Aerial, field, and marine robotics
Abstract: This paper presents a path-based integer linear programming formulation for the sweep coverage problem, in which points of interest of a given area, i.e., an indoor farming field, must be covered by mobile sensors, subject to redundancy and sensing range constraints. A price-and-branch algorithm, whose pricing subproblem is formulated as a generalized orienteering problem, is employed to compute primal and dual bounds. For a simplified variant of the problem, a convex-hull-based destroy-and-repair heuristic is designed for the warm start and acceleration of column generation. The effectiveness of the proposed approach is discussed through computational experiments.

Keywords: Data-driven robust control, Sliding mode control, Model predictive control
Abstract: This paper proposes a robust data-driven control for nonlinear systems by integrating model-free predictive control (MFPC) with discrete-time sliding mode control (DTSMC). While MFPC effectively optimizes control actions using historical input-output data without an explicit model and Sliding mode augmentation is introduced to ensure the system stability and to address the steady-state error issue due to the unmodeled uncertainty. Simulation results on a nonlinear system demonstrate that the proposed method significantly reduces the error compared to conventional PID and pure MFPC, yielding faster settling times and superior tracking performance.

Keywords: Model predictive control of hybrid systems, Multi-agent systems, Optimal control of discrete event and hybrid systems
Abstract: Recently, it has become increasingly important to develop crowd behavior models which reflect the states of pedestrian attention. An example of these models is the behavior of smartphone-distracted pedestrians, whose reduced attention increases the risk of accidents. In this study, we focus on a hybrid system model which represents the pedestrian behavior with model predictive control, and reflect the state of attentions by adjusting the recalculation cycle of the model predictive control. The features of the proposed models are discussed based on simulation results.

Keywords: Model predictive control, Applications of optimal control, Convex optimization
Abstract: This paper presents a model predictive control (MPC) framework for autonomous vehicle overtaking of a slower preceding vehicle. A virtual vehicle, defined as a copy of the following vehicle assuming no slower preceding vehicle, is used to generate a nominal reference trajectory. The MPC optimizes the rates of steering and acceleration commands to obtain smooth maneuvers, while safety is enforced through a rectangular safety region constraint with slack variables to maintain feasibility. Simulation results with a linear vehicle model demonstrate smooth lane change and lane return while satisfying input-rate and safety constraints.

Keywords: Model predictive control, Convex optimization, Large-scale and networked optimization problems
Abstract: Water Distribution Networks (WDNs) require efficient operational strategies to ensure reliable and energy-efficient water supply. Model Predictive Control (MPC) approaches for scheduling pumps in WDNs often lead to Mixed-Integer Nonlinear Programs (MINLPs) due to combinatorial pump configurations which hinder the computational tractability. This paper presents a decomposition-based framework for a class of single-sourced, branched WDNs with a single pumping station composed of multiple variable-frequency-drive pumps (VFDs). The pumping station is replaced with a discharge pressure decision variable, allowing the network to optimize for an energy-dissipation surrogate for pumping work. Under suitable relaxations, the resulting problem becomes convex and yields a well-defined optimal solution. The actual pump operation is then recovered through a set of smaller nonlinear programs corresponding to different numbers of active pumps. Such a decomposition has also been integrated into MPC experiments to enable control of WDNs.

Keywords: Model predictive control, Uncertain systems, Stability of nonlinear systems
Abstract: Certainty-equivalence MPC (CE-MPC) is widely used despite lacking performance guarantees under model mismatch. This paper discusses the analysis framework proposed by Liu et al. (2026), which provides theoretical stability and optimality guarantees for CE-MPC of nonlinear systems, and applies it to scenarios involving (non-smooth) additive model mismatch. The core of the analysis framework is a perturbation analysis of the MPC value function for quadratic stage costs. A sufficient condition on the mismatch level ensuring stability is presented, followed by a competitive-ratio performance bound quantifying the suboptimality of CE-MPC relative to the infinite-horizon optimal controller with perfect model knowledge. The results explicitly characterize the joint effect of the prediction horizon and the mismatch on the stability and infinite-horizon performance of CE-MPC, and they are particularly useful for designing CE-MPC using surrogate models, e.g., neural network-based models.

Keywords: Optimal control theory, Model predictive control, Saturation and discontinuity
Abstract: Sparse optimal control is considered in the infinite horizon. In the literature, sparse control has been considered mostly in a finite horizon so that it is formulated into a finite-dimensional optimization problem. It is shown in this paper that an optimal solution of the infinite-horizon sparse control problem can be obtained through a solution of some finite-horizon subproblem. This is due to sparsity of the optimal solution in the sense that the optimal control input is constantly equal to zero at its tail. An estimate is given on the horizon length required for this subproblem to guarantee optimality in the infinite horizon.

Keywords: AI and learning-based control for automotive systems, Automotive system identification and modelling, Electric and solar vehicles
Abstract: This study proposes a Neural State-Space Model Predictive Control (NSS-MPC) strategy for integrated EV charging–thermal management under low-temperature slow-charging conditions. An NSS prediction model was identified using data generated from an AMESim-based EV charging–thermal management system model and integrated into a MATLAB-based MPC framework. The proposed controller uses compressor speed as the manipulated variable, incorporates ambient temperature and charging current as measured disturbances, and optimizes preheating timing to satisfy HVB temperature requirements without PTC heater use. To distinguish the effect of heater elimination from the effect of predictive preheating control, the proposed strategy was compared with Baseline (HP+PTC) and Baseline (HP) cases. AMESim–MATLAB co-simulation results showed that the proposed NSS-MPC achieved the highest charging efficiency and final SOC and the lowest carbon emissions among the compared cases.

Keywords: Data-driven control theory, Machine and deep learning for system identification, Learning methods for control
Abstract: We study residual-based, data-driven control for nonlinear discrete-time systems with unknown dynamics. Rather than identifying full plant dynamics, we learn the residual between a nominal model and the true input-output map and inject this correction into a model-reference controller. A Model-Reference Gaussian Process Regression (MR-GPR) module estimates a one-step update that absorbs nominal-plant mismatch; the resulting law is an output-feedback controller built from finite input-output windows. Under finite-window reconstructability, practical internal stability, and a deterministic residual-approximation condition, the closed loop admits an explicit class-K output ultimate bound in terms of the residual approximation error. The Gaussian process is used as a smooth nonparametric approximator; its predictive variance is reported only diagnostically and is not used in the controller or theorem. On a battery-manufacturing-motivated coating example with state-dependent mismatch, the design improves over nominal-only control, and a dataset-size sweep clarifies the accuracy-computation trade-off.

Keywords: Data-driven control theory, Linear system identification, Time series modeling
Abstract: In this paper, we consider the problem of designing input signals for an unknown linear time-invariant system in such a way that the resulting input-state data, in the noise-free setting, is suitable for identification or stabilization. We will take into account prior knowledge on system-theoretic properties of the system, in particular, controllability and stabilizability. For this, we extend the notion of universal inputs to incorporate prior knowledge on the system. An input is called universal for identification (resp., stabilization) if, when applied to any system complying with the prior knowledge, it results in data suitable for identification (resp., stabilization) regardless of the initial condition. We provide a full characterization of such universal inputs. In addition, we discuss online experiment design using prior knowledge, and we study cases where this approach results in the shortest possible experiment for identification and stabilization.

Keywords: Data-driven control theory
Abstract: This paper proposes Image-Driven Control (IDC), an approach to vision-based feedback that operates directly in the image space. IDC offers a direct end-to-end solution in the behavioral framework. Targeting a limited subset of vision-based feedback problems, including applications in manufacturing, the IDC design builds on a computationally advantageous subspace predictive control formulation and offset-free tracking to enable reference tracking in the image space under input and pixel-level constraints. IDC provides a control-theoretic alternative to vision-based feedback systems, enabling improved transparency for analysis and tuning of end-to-end control. It offers straightforward experiment design, an advantage in applications where data collection is expensive. The effectiveness of IDC is shown in a simulation example using a nonlinear high-fidelity model of the heating phase of the thermoforming process.

Keywords: Data-driven control theory, Consensus, Control of networks
Abstract: For a collection of homogeneous LTI systems that is interconnected by a protocol, given the network topology and the system model, one may obtain a feedback gain to synchronize the network. However, the model-based methods cannot be applied in case the system model is unknown. Therefore, in this paper, we study the data-driven synchronization problem for homogeneous networks. In particular, given a collection of LTI systems, we collect the input-state data from one individual system. Then, given the network topology, we provide data-based necessary and sufficient conditions for synchronizability. Once the conditions are satisfied, one can also obtain a feedback gain directly from data to synchronize the network with the corresponding design method provided in this paper. Finally, we illustrate our results with a numerical simulation.

Keywords: Data-driven control theory, Nonlinear system identification, Adaptive gain scheduling autotuning control and switching control
Abstract: This paper presents a Gain-Scheduled Data-Enabled Predictive Control (GS-DeePC) framework for nonlinear systems based on multiple locally linear data representations. Instead of relying on a single global Hankel matrix, the operating range of a measurable scheduling variable is partitioned into regions, and regional Hankel matrices are constructed from persistently exciting data. To ensure smooth transitions between linearization regions and suppress region-induced chattering, composite regions are introduced, merging neighboring data sets and enabling a robust switching mechanism. The proposed method maintains the original DeePC problem structure and requires locally informative data sequences. Extensive experiments on a nonlinear DC-motor with an unbalanced disc demonstrate the significantly improved control performance compared to standard DeePC.

Keywords: Probabilistic and Bayesian methods for system identification, Data-driven control theory, Linear system identification
Abstract: Extending data-driven algorithms based on Willems' fundamental lemma to stochastic data often requires empirical and customized workarounds. This work presents a unified Bayesian framework for linear systems that provides a systematic and general method for handling stochastic data-driven tasks, including smoothing, prediction, and control, via maximum a posteriori estimation. This framework formulates a unified trajectory estimation problem and solves a Bayesian problem that optimally combines trajectory knowledge with trajectory characterization from offline data. This problem generalizes existing data-driven prediction and control algorithms. Numerical examples demonstrate the performance of the unified approach for all three tasks against other data-driven and system identification approaches.

Keywords: Multi-agent systems, Distributed control and estimation, Control under communication constraints
Abstract: This paper proposes a decentralized controller for large-scale heterogeneous multi-agent systems subject to bounded external disturbances, where agents must satisfy Signal Temporal Logic (STL) specifications requiring cooperation among non-communicating agents. To address the lack of direct communication, we employ a decentralized k-hop Prescribed Performance State Observer (k-hop PPSO) to provide each agent with state estimates of those agents it cannot communicate with. By leveraging the performance bounds on the state estimation errors guaranteed by the k-hop PPSO, we first modify the space robustness of the STL tasks to account for these errors, and then exploit the modified robustness to design a decentralized continuous-time feedback controller that ensures satisfaction of the STL tasks even under worst-case estimation errors. A simulation result is provided to validate the proposed framework.

Keywords: Multi-agent systems, Control of networks, Consensus
Abstract: The practical implementation of multi-agent formation control is often hindered by physical sensing limitations. Moving beyond traditional assumptions of omnidirectional motion and undirected topologies, this paper addresses formation control for nonholonomic mobile robots constrained by directed sensing graphs and limited Field-of-View (FOV) sensors. We propose a distributed control framework that guarantees convergence to a desired geometric configuration while maintaining neighbor visibility throughout the transient phase. Specifically, the underlying nominal control law reduces the relative bearing errors under nonholonomic constraints, while the FOV limits are modeled using Control Barrier Functions (CBFs) and incorporated via a Quadratic Programming (QP) formulation. Rigorous stability analysis proves the asymptotic convergence of the constrained agent trajectories to the target formation, despite the directed sensing topology. The numerical simulations verify that the proposed framework successfully achieves the target formation without violating visual connectivity constraints.

Keywords: Control under communication constraints, Quantized systems, Control over networks
Abstract: Digital communication and controller implementation have become ubiquitous in many modern control applications. However, digital control also imposes new challenges on the controller design because it requires that the controller has a finite control range. This raises the question of what fundamental limitations controllers with finite control range have, and whether classical performance objectives can be achieved. We investigate these questions for quadratic performance and dissipativity with quadratic supply rates. This work proves that these properties require an infinite control range if the system possesses an unstable mode affecting the performance channel. Consequently, it is not possible to design digital controllers that can guarantee these properties. The results also imply that dissipativity-based feedback theorems, such as the passivity theorem, cannot be used to guarantee stability of feedback interconnections when the individual systems within the interconnection should be controlled digitally. In view of the significance of digital control, these limitations show the importance to define new notions of control performance which can be achieved by digital controllers and which mathematically describe properties which are practically desired.

Keywords: Multi-agent systems, Distributed control and estimation, Consensus
Abstract: The knowledge of the algebraic connectivity is central in many graph-theory related studies, such as network analysis, maintaining connectivity, and control of multi-agent systems. The computation of this variable is usually centralized, but for some applications it is necessary to obtain it in a distributed way. In this paper, a distributed scheme is proposed to compute the algebraic connectivity using dynamic average consensus algorithms converging in finite time. It is proved that the estimated value asymptotically converges toward the algebraic connectivity for undirected graphs. Numerical simulations show the performance and the advantages of the proposed scheme compared to existing distributed ones.

Keywords: Multi-agent systems, Control over networks
Abstract: In this paper, we propose a control barrier function (CBF) approach to distributed generalized Nash equilibrium (GNE) seeking problems, which ensures feasible set invariance while seeking the equilibrium. We start with singe-integrator agents, and design a CBF-based algorithm that converges asymptotically to the exact GNE of the game without violating the feasibility of the problem along the evolution. Then by introducing a coordinate transformation, we extend the approach to double-integrator agents. The algorithms are developed with full decision information setting. A simulation example is provided to illustrate the applicability of the algorithms.

Keywords: Multi-agent systems, Distributed optimization, Control under communication constraints
Abstract: Satellite constellations in low-Earth orbit are now widespread, enabling positioning, Earth imaging, and communications. In this paper, we address how learning problems can be solved in a distributed manner across these satellite constellations. Specifically, we focus on a federated approach, where satellites collect and locally process data, with the ground station aggregating local models. The goal is to design a novel algorithm that is jointly communication-efficient and accurate. To this end, we employ several mechanisms to reduce the number of communications with the ground station (local training) and their size (compression). We then propose an error feedback mechanism that enhances accuracy, which yields, as a byproduct, an algorithm-agnostic error feedback scheme that can be more broadly applied. We analyze the convergence of the resulting algorithm, and compare it with the state of the art through simulations in a realistic space scenario, showcasing superior performance.

Keywords: Adaptive control of multi-agent systems, Model reference adaptive control
Abstract: Despite the fact that numerous protocols have been proposed for the longitudinal control of multi-vehicle systems, most designs disregard the inherent actuation constraints of vehicles. Ignoring such constraints may lead to disruption in the formation, e.g., loss of cohesiveness or collisions. In this study we address this challenge in the presence of uncertainties in the multi-vehicle dynamics. We solve the problem by introducing a constrained reference model: the idea is to design a reference behavior that prevents the vehicles from reaching their input constraints, so that the formation is not disrupted. Using absolute stability arguments, stability of the adaptive closed-loop system is proven for input constraints described by sector-bounded nonlinearities. The effectiveness of the proposed approach is verified via simulations.

Keywords: Cyber security networked control, Control over networks, Multi-agent systems
Abstract: This paper investigates privacy-preserving distributed cooperative control for multi-agent systems within the framework of differential privacy. In cooperative control, communication noise is inevitable and is typically treated as a disturbance that degrades coordination performance. This work instead reinterprets such noise as a potential privacy-preserving mechanism and develops a linear quadratic regulator (LQR)-based distributed control framework to exploit this property. In the proposed setup, agents communicate over noisy channels whose noise variance depends on the relative state differences between neighbouring agents. It is analytically shown that, under a tree-structured communication topology, the inherent communication noise guarantees bounded (ϵ,δ)-differential privacy for reference signals without requiring additional privacy injection. Meanwhile, the cooperative tracking error remains bounded and converges in both the mean-square and almost-sure senses.

Keywords: Cyber security networked control, Multi-agent systems, Consensus
Abstract: In this paper, we investigate false data injection attacks on multi-agent systems. We consider attacks on agent sensors and the defense of agents on actuators under these attacks. We focus on the intrinsic stealthiness of attacks and the vulnerability of systems to attacks. For this purpose, we propose a signal-centric, detection-oriented stealthiness metric termed stealthiness margin, which constitutes the fundamental limit of attack stealthiness and system vulnerability, and can be computed by solving a zero-sum minimax optimization problem. We solve this problem analytically, explicitly constructing optimal attack and defense signals. The solution indicates that the stealthiness and the vulnerability are closely related to the observability and controllability Gramian matrices of multi-agent systems in terms of agent dynamics and network topology, which can be interpreted from a geometric analysis by generalizing the concept of balanced realization to multi-agent systems.

Keywords: Fault detection and diagnosis, Cyber security networked control, Resilient networked control systems
Abstract: In this paper, we investigate the maximally allowable energy of an attack that can be injected into multi-agent systems with disturbances, while remaining stealthy by passing through the anomaly detector. Specifically, it is formulated as a feedback control problem in which, the attack policy is a feedback of the system disturbance. Then, the attacker maximizes the controller H∞ norm to achieve the allowed energy margin, while satisfying the stealthiness requirement that the H∞ norm of the composite disturbance–attack residual channel stays below the detector’s threshold. Our results demonstrate that the existence of stealthy attacks closely depends on a specific distance metric between the non-minimum phase zeros of the attack and disturbance channels entering the agents. Furthermore, a closed-form solution for the optimal attack policy is derived based on a generalized Hankel operator characterizing the projection of disturbance channels on the attack channels. These results quantify how the agent dynamics and network topologies confine the attack performance in the security of distributed systems.

Keywords: Cyber security networked control, Fault detection and diagnosis
Abstract: The stealth of false data injection attacks (FDIAs) against feedback sensors in linear time-varying (LTV) control systems is investigated. In that regard, the following notions of stealth are pursued: For some finite epsilon > 0, i) an FDIA is deemed epsilon-stealthy if the deviation it produces in the signal that is monitored by the anomaly detector remains epsilon-bounded for all time, and ii) the epsilon-stealthy FDIA is further classified as untraceable if the bounded deviation dissipates over time (asymptotically). For LTV systems that contain a chain of q geq 1 integrators and feedback controllers with non-negative impulse-response kernels, it is proved that polynomial (in time) FDIA signals of degree a—growing unbounded over time for a geq 1—will remain i) epsilon-stealthy, for some finite epsilon > 0, if a leq q, and ii) untraceable, if a < q. These results are obtained using the theory of linear Volterra integral equations.

Keywords: Markov decision process, Consensus and reinforcement learning control
Abstract: In this paper, we consider an N-to-1 pursuit-evasion problem with a limited field of view, where multiple pursuers make decisions independently. The problem is formulated as a partially observable stochastic pursuit-evasion game with history information, and the existence of a Nash equilibrium is established by constructing a saddle point of an auxiliary two-player zero-sum game. For strategy computation, we adopt a sliding window approach based on recent observations and present a self-play reinforcement learning algorithm to compute the corresponding strategies. Moreover, the effectiveness of the proposed method is validated through a numerical example.

Keywords: Markov decision process, Control under communication constraints, Data-driven control theory
Abstract: This paper investigates the problem of data-driven event-triggered control for discrete-time Markovian jump systems (MJSs) with generally uncertain transition probabilities (TPs). Considering the scenario where the measured state is transmitted through a network, a novel data-driven framework is developed that directly constructs stabilizing state-feedback controllers and a triggering policy from offline input-state data without requiring explicit system identification. First, ignoring the network, we derive sufficient conditions for designing mode- dependent controllers that guarantee stochastic stability solely from collected data via linear matrix inequalities (LMIs). Then, incorporating the event-triggered mechanism, a triggering policy is proposed to reduce communication load while preserving closed-loop stability. We further establish a verifiable sufficient condition under which the designed triggering policy reduces to time-triggered transmission. Finally, a numerical example is provided to illustrate the effectiveness of the proposed method.

Keywords: Markov decision process, Multi-agent systems
Abstract: Collective cooperation drives natural, social, and economic systems, with its evolutionary game study a priority. Though human interactions form complex networks, current public goods game research is limited to deterministic evolution with static or instantly replaced players. In reality, interactions involve randomness and finite lifecycles, leaving cooperation in such dynamic evolutionary networks unaddressed. This paper investigates cooperation evolution through a random entry-exit mechanism, developing an overlapping generations model where finite-lived players undergo evolutionary processes influenced by cooperative dynamics. Our analysis shows that properly accounting for update randomness enables a stable cooperative equilibrium given sufficient population size and update frequency.

Keywords: Markov decision process, Supervisory control and automata, Synthesis of stochastic systems
Abstract: We study the Stochastic Shortest Path (SSP) problem for autonomous systems with mixed max-sum cost aggregations under Linear Temporal Logic constraints. Classical SSP formulations rely on sum-aggregated costs, which are suitable for cumulative quantities such as time or energy but fail to capture bottleneck-style objectives where performance is determined by the worst single event along a trajectory. To address this limitation, we introduce max-aggregated objectives that minimize the maximum one-step cost. We show that standard Bellman equations on the original state space are not directly applicable and propose an augmented MDP that tracks the running maximum cost. We also identify a cyclic policy phenomenon caused by zero marginal cost under max-aggregation, and resolve it via a finite-horizon formulation. To handle high-level task specifications, we construct a product MDP from the stochastic system and the automaton corresponding to the LTL formula. A lexicographic value iteration algorithm is then developed to optimize mixed max-sum objectives under lexicographic ordering. Gridworld case studies demonstrate the effectiveness of the framework.

Keywords: Security for stochastic systems, Markov decision process, Stochastic control
Abstract: The adoption of reinforcement learning for critical infrastructure defense introduces a vulnerability where sophisticated attackers can strategically exploit the defense algorithm's learning dynamics. While prior work addresses this vulnerability in the context of repeated normal-form games, its extension to the stochastic games remains an open research gap. We close this gap by examining stochastic security games between an RL defender and an omniscient attacker, utilizing a tractable linear influence network model. To overcome the structural limitations of prior methods, we propose and apply neuro-dynamic programming. Our experimental results demonstrate that the omniscient attacker can significantly outperform a naive defender, highlighting the critical vulnerability introduced by the learning dynamics and the effectiveness of the proposed strategy.

Keywords: Statistical analysis, Markov decision process, Statistical inference
Abstract: This study evaluates a Stochastic Automata Network (SAN) framework for short-term forecasting of sky conditions. Observations from the Reims--Prunay station were used to build a stochastic model coupling temperature, relative humidity, atmospheric pressure, and cloud cover through data-driven transition rules. Independent validation for 2025 shows that the model performs best in identifying coarse sky regimes, particularly clear-sky conditions, while intermediate cloud states remain challenging. The SAN behaves primarily as a statistical regime classifier rather than a physical cloud model. Although its accuracy is lower than that of Numerical Weather Prediction (NWP) systems, the model generates daily forecasts in seconds, highlighting its potential for forecasting in computationally constrained environments.

Keywords: Optimal control of discrete event and hybrid systems, Discrete event modeling and simulation, Reachability analysis, verification and abstraction of hybrid systems
Abstract: We investigate the task and motion planning problem for dynamic systems under signal temporal logic (STL) specifications. Existing works on STL control synthesis mainly focus on generating plans that satisfy properties over a single executed trajectory. In this work, we consider the planning problem for hyperproperties evaluated over a set of possible trajectories, which naturally arise in information-flow control problems. Specifically, we study discrete-time dynamic systems and employ the recently developed temporal logic HyperSTL as the new objective for planning. To solve this problem, we propose a novel recursive counterexampleguided synthesis approach capable of effectively handling HyperSTL specifications with multiple alternating quantifiers. The proposed method is not only applicable to planning but also extends to HyperSTL model checking for discrete-time dynamic systems. Finally, we present case studies on security-preserving planning and ambiguity-free planning to demonstrate the effectiveness of the proposed HyperSTL planning framework.

Keywords: Optimal control of discrete event and hybrid systems, Event-based control
Abstract: This paper proposes a constraint-switching approach for Nonlinear Model Predictive Control (NMPC) to address the challenges of large prediction horizons coupled with recursively feasible constraints in particular for design frameworks which imply exploration goals. The K-invariant sets and their parameterization is introduced as a substitute for controlled invariant sets. It is shown that such sets can be effectively constructed through an external selection and verification module, making them available along the prediction horizon. The availability of K-invariant sets and their enforcement as constraint in the receding optimization for exploration purposes leads to a switching mechanism. The recursive feasibility is guaranteed under the framework, and an example of navigation in partially known environment is provided to demonstrate the advantages.

Keywords: Optimal control of discrete event and hybrid systems, Markov decision process, Security for stochastic systems
Abstract: We formalize a proactive defense that dynamically redeploys service instances (e.g., containers) based on their age to interrupt attacks. We model the problem using weakly coupled Markov decision processes with a per-time availability constraint to maintain service access through a pool of low-risk servers. We find ergodic stationary policies that minimize operational and energy costs for standalone servers and extend the result with an algorithm to control the servers' redeployment while maintaining the per-time availability constraint in expectation.

Keywords: Optimal control of discrete event and hybrid systems, Neural and fuzzy adaptive control, Data-driven control theory
Abstract: Model-based lateral control for autonomous vehicles faces challenges with unmeasurable parameters and nonlinearities. This paper proposes a robust data-driven control to identify vehicle lateral dynamics from driving data. The Koopman operator lifts nonlinear dynamics into a linear space, but its observable function selection is challenging and experience-dependent. A neural network is employed to learn the embedding function and Koopman operator. Based on Serret-Frenet kinematics and deep neural Koopman model, the control design achieves dissipativity under steering constraints, with conditions formulated as linear matrix inequalities. Moreover, an adaptive sliding mode approach attenuates adverse influence of actuator uncertainties. CarSim-Simulink co-simulations validate the proposed control's effectiveness.

Keywords: Optimal control of discrete event and hybrid systems, Reachability analysis, verification and abstraction of hybrid systems, Stochastic hybrid systems
Abstract: Discrete-time stochastic systems with continuous spaces are hard to verify and control due to the curse of dimensionality. We propose an abstraction-based framework with robust dynamic programming mappings that synthesize controllers with provable temporal-logic satisfaction lower bounds via approximate stochastic simulation relations. Exploiting decoupled dynamics, we reveal a Canonical Polyadic Decomposition tensor structure in value functions, enabling scalable dynamic programming. The method provides correct-by-design probabilistic guarantees and is validated on continuous-state linear stochastic systems.

Keywords: Optimal control of discrete event and hybrid systems, Supervisory control and automata, Discrete event modeling and simulation
Abstract: In this paper, we investigate the motion planning problem for Signal Temporal Logic (STL) specifications in the presence of uncontrollable agents. Existing works mainly address this problem in a robust control setting by assuming the uncontrollable agents are adversarial and accounting for the worst-case scenario. While this approach ensures safety, it can be overly conservative in scenarios where uncontrollable agents have their own objectives that are not entirely opposed to the system’s goals. Motivated by this limitation, we propose a new framework for STL motion planning within the Stackelberg game setting. Specifically, we assume that the system controller, acting as the leader, first commits to a plan, after which the uncontrollable agents, acting as followers, take a best response based on the committed plan and their own objectives. Our goal is to synthesize a control sequence for the leader such that, for any rational followers producing a best response, the leader’s STL task is guaranteed to be satisfied. We present an effective solution to this problem by transforming it into a single-stage optimization problem and leveraging counter-example guided synthesis techniques. We demonstrate that the proposed approach is sound and identify conditions under which it is optimal. Simulation results are also provided to illustrate the effectiveness of the proposed framework.

Keywords: Convex optimization, Lyapunov methods, Stability of nonlinear systems
Abstract: In a real Hilbert space setting, we investigate the asymptotic properties of the solutions of the classical continuous steepest descent method with convex-like potential. Despite the absence of convexity, we show that the solutions preserve the remarkable minimizing properties typically associated with convex functions. In particular, we find that the values of the convex-like potential decay asymptotically at a sublinear rate. If, moreover, the potential function is weakly lower semi-continuous, we prove that the solutions weakly converge toward a minimizer. Under a quadratic growth condition on the convex-like potential, we further provide a strong convergence result for the solutions along with an exponential decay rate of the function values. Numerical experiments illustrate our theoretical findings.

Keywords: Convex optimization, Robustness analysis
Abstract: This work studies the linear convergence of an accelerated scheme of the Alternating Direction Method of Multipliers (ADMM) for strongly convex and Lipschitz-smooth problems. We use the methodology of expressing the accelerated ADMM as a Lur'e system, i.e., an interconnection of a linear dynamical system in feedback with a slope-restricted operator, and we use Integral Quadratic Constraints to establish linear convergence. We leverage this machinery to systematically explore parameter tuning heuristics, including Nesterov-inspired choices and configurations identified via grid search, and analyze their impact on the convergence rate. Our new bounds show improved linear convergence rates compared to the vanilla algorithm and previously proposed accelerated variants, which is also empirically validated on a LASSO regression benchmark.

Keywords: Convex optimization, Optimization-based estimation and control, Sliding mode control
Abstract: This paper introduces a control-theoretic framework for solving local minimax optimisation problems by formulating invariant ridge-following dynamics within the paradigm of sliding mode control. The proposed textit{Sliding Follow-the-Ridge} (SFR) algorithm reinterprets classical Gradient Descent–Ascent (GDA) schemes as a sliding manifold control problem, where the gradient field defines the manifold representing the first-order optimality condition. The reaching law is designed in such a way that the sliding manifold remains asymptotically stable only when the second-order necessary condition for a true minimax equilibrium is satisfied, thereby preventing convergence to non-minimax stationary points. The resulting dynamics ensure finite-time convergence to local minimax solutions, while preserving numerical stability and low computational cost. Comparative evaluations against the Follow-the-Ridge (FR) algorithm demonstrate that SFR achieves faster convergence, shorter trajectory paths and improved robustness. By embedding minimax optimisation in a rigorous control-theoretic structure, SFR establishes a principled bridge between sliding mode control and nonconvex game optimisation.

Keywords: Convex optimization, Adaptive control design, Linear systems
Abstract: In this paper, we address the problem of achieving fast convergence in gradient descent for quadratic functions without relying on a-priori knowledge of global function parameters. Inspired by adaptive stepsize algorithms for smooth convex functions, we propose a computationally lightweight strategy based on running estimates of minimal and maximal local curvatures. We prove that our proposed algorithm converges to the optimal constant stepsize which achieves the fastest convergence. Simulations show that the convergence rate achieved by our proposed algorithm is comparable or superior to recent adaptive approaches both in the quadratic case under consideration and in a preliminary test on logistic classification.

Keywords: Convex optimization, Numerical methods for optimal control, Model predictive control
Abstract: The Proportional-Integral Projected Gradient (PIPG) method has demonstrated to be an efficient first-order method for quadratic programs (QP) in practice, which only uses vector operations and one projection per iteration. However, it lacks linear convergence guarantees for strongly convex QPs, which is a key property for comparable first-order methods. This gap limits both theoretical understanding and practical confidence in its use. To this end, this paper rigorously proves that PIPG achieves global linear convergence for such problems with explicit convergence rate. To further accelerate its convergence, we propose an adaptive step-size rule with periodic restarts. Numerical experiments on a model predictive control problem show that the enhanced PIPG converges faster than the-state-of-the-art first-order methods, while maintaining its signature simplicity and low per-iteration cost.

Keywords: Convex optimization
Abstract: Resource allocation is a fundamental problem in Industrial Internet of Things (IIoT) systems, in which devices work together under limited communication bandwidth to complete diverse tasks. This paper proposes a communication-efficient distributed optimization algorithm tailored for problems with coupled constraints. To tackle coupled constraints, we solve the problem via its dual counterpart, and develop a compressed version. Difference compression and a dynamic scaling factor are then introduced to mitigate compression errors. We show that the proposed algorithm converges linearly for strongly convex and smooth objectives. Numerical simulations verify the theoretical results and demonstrate the efficiency and robustness of the proposed algorithm.

Keywords: Learning methods for optimal control
Abstract: Deep neural networks are increasingly used as effective parameterizations of control policies in various learning-based control paradigms. For continuous-time optimal control problems (OCPs), which are central to many decision-making tasks, control policy learning can be cast as neural ordinary differential equation (NODE) problems wherein state and control constraints are naturally accommodated. This paper presents a NODE approach to solving continuous-time OCPs for the case of stabilizing a known constrained nonlinear system around a target state. The approach, termed Lyapunov-NODE control (L-NODEC), uses a novel Lyapunov loss formulation that incorporates an exponentially-stabilizing control Lyapunov function to learn a state-feedback neural control policy, bridging the gap of solving continuous-time OCPs via NODEs with stability guarantees. The proposed Lyapunov loss allows L-NODEC to guarantee exponential stability of the controlled system, as well as its adversarial robustness to perturbations to the initial state. The performance of L-NODEC is illustrated on a double integrator, where it effectively stabilizes the controlled system around the target state despite perturbations to the initial state and it reduces the inference time necessary to reach the target.

Keywords: Learning methods for optimal control
Abstract: We propose a variational quantum control framework for high-fidelity state transfer in spin chains under noisy and resource-constrained quantum hardware conditions. The method maps a continuous-time optimal control problem into a Trotterized, physics-informed parameterized quantum circuit, enabling hybrid quantum-classical optimization of control parameters. We investigate two control parameterizations: a local scheme with site-wise parameters and a compact global scheme with a low-dimensional shared structure. Numerical results on XXZ spin chains show that both parameterizations achieve near-unit fidelity in the noiseless setting. Under depolarizing and dephasing noise, the global parameterization exhibits improved robustness, training stability, and parameter efficiency. These findings highlight an expressivity-robustness trade-off and demonstrate the advantages of structured low-dimensional ansätze for variational quantum control in NISQ regimes.

Keywords: Learning methods for optimal control, Applications of optimal control
Abstract: This paper presents a safety-certified reinforcement learning framework for autonomous racing that minimizes lap times while strictly enforcing safety and velocity constraints. To address the brittleness and lack of safety guarantees in standard Deep Reinforcement Learning (DRL), we propose a multi-stage architecture. In this framework, a Proximal Policy Optimization (PPO) policy optimizes racing performance, while a structured safety filtering stack combining Control Barrier Functions (CBF), Control Lyapunov Functions (CLF), and a feasibility-enforcing mechanism, enforces track limits and velocity constraints while guaranteeing episode completion. This filtering stack restricts exploration to admissible regions, stabilizing training, and preventing constraint violations. Simulations on 1:10-scale tracks demonstrate near-optimal trajectories and successful zero-shot generalization to unseen environments, without requiring computationally expensive offline trajectory optimization.

Keywords: Learning methods for optimal control, Applications of optimal control, Model predictive control
Abstract: This paper presents the design and experimental validation of a Meta-Reinforcement Learning approach for temperature control in a real building with not perfectly known thermal dynamics. The proposed architecture consists of three main modules: an encoder, a controller, and an adapter. In particular, the encoder projects the building’s uncertain parameters into a low-dimensional latent representation that provides contextual information to the downstream controller, enabling it to interpret the current building dynamics and apply the most suitable temperature control strategy to optimize closed-loop performance. The whole control architecture is trained in simulation on a control-oriented model across the range of dynamics induced by parameter uncertainties. Experimental results on a real house located at the SYSLAB Risø Campus of the Technical University of Denmark demonstrate that the Meta-Reinforcement Learning approach is feasible in practice and improves energy efficiency after just three days of data collection.

Keywords: Learning methods for optimal control, Applications of optimal control, Stochastic optimal control problems
Abstract: In this article, we study the problem of rooftop PV installation for a renewable energy community, while considering a multi-stage, multi-investor setting, subject to uncertainty in the membership status of the community end-users. We model the evolution of the membership status as a Markov chain and formulate the investor's decision making as a stochastic optimal control problem; with coupling in the objective function due to non-linear proportional sharing mechanism. We then translate the problem as a partially observable Markov decision process and propose a reinforcement learning based solution. We perform Monte Carlo simulations using real-world household dataset and validate the proposed framework.

Keywords: Learning methods for optimal control, Design methods for data-based control, Applications of optimal control
Abstract: This paper presents the adaptation of established Verification & Validation (V&V) practices for classical flight control to a data-driven Reinforcement Learning (RL) based Gust Load Alleviation (GLA) control system, with the aim of addressing the simulation to reality gap in RL based aeroelastic control applications. The control law was developed using the Soft ActorCritic (SAC) algorithm as an end-to-end deep neural network that maps sensor measurements to actuator commands, and was trained offline using data from a linear Aeroservoelastic (ASE) model. The resulting fixed-policy controller is then subjected to a V&V workflow comprising closed-loop verification through simulation analyses on a high-fidelity nonlinear ASE model, followed by experimental validation through wind tunnel testing. Validation was conducted on a flexible wing demonstrator under harmonic gust excitation generated by cylindrical gust generators. Performance was evaluated against a set of control design specifications that cover nominal load alleviation, robustness to unstructured and parametric uncertainties, actuator and safety constraints. Wind tunnel results demonstrated that the RL controller attained a maximum Wing Root Bending Moment (WRBM) reduction of 65% and 80% using single and dual-actuator configurations, respectively, at the first wing bending mode frequency.

Keywords: Linear fractional-order systems, Linear systems, Linear time-delay systems
Abstract: In this work, a design methodology for fractional-order tilted derivative (FOTID) controllers is introduced for fractional-order systems with inherent time delay. The approach is built upon a delayed Bode transfer function (DBTF) framework, which enables systematic compensation of the phase lag induced by dead-time while ensuring a well-shaped open-loop frequency response. As a result, the proposed FOTID controllers achieve enhanced robustness against both gain and delay variations, while preserving the desired frequency-domain performance characteristics. The effectiveness of the proposed FOTID controllers is validated through simulation studies and comparative analyses against established fractional order PID controllers reported in the literature.

Keywords: Controller constraints and structure, Control of complex systems
Abstract: In this paper, we present a ratio control architecture where fractional-order proportional-integral-derivative (FOPID) controllers are employed. In particular, we extend the dynamic blend station method and show that the general architecture can be successfully applied in a fractional setting, achieving perfect ratio tracking when the set-point changes and improving load disturbance rejection performance. In addition, a simplified technique is proposed to provide greater flexibility in the overall design and reduced computational cost. Simulation results demonstrate the effectiveness of the proposed methodology.

Keywords: Robust control applications, Robustness analysis, Linear fractional-order systems
Abstract: Most fractional-order PID tuning research focuses on gain robustness; however, few papers address the critical issue of robustness to time constant variations. A design procedure for robust fractional order PID controllers under time constant variations is presented in this study, with a focus on both first and second order plus dead time processes. Partial derivatives are used in the design method to specify the robustness condition. The resulting system of nonlinear equations is solved by a graphical approach. The numerical examples based on biomedical systems are employed to validate the performance of the developed method in ensuring robustness to varying time constants. Comparative closed-loop simulation results for PID controllers are presented.

Keywords: Linear fractional-order systems, Linear systems, Decentralized control
Abstract: Closed-loop control of anesthesia faces fundamental limitations when modeled as multiple-input single-output (MISO) systems, where patient responses may lead to non-unique or physiologically ambiguous operating points. To ensure an interpretable multivariable formulation, this work proposes a 2×2 multiple-input multiple-output (MIMO) control framework that simultaneously regulates clinical Bispectral Index (BIS) and the Nociception Level Index (NOL) using Propofol and Remifentanil infusion as inputs. The framework leverages Response Surface Models (RSM) for sedation–nociception dose mapping and defines a unique operating point through analytical RSM-based (BIS–NOL) characterization. For induction, population-based decentralized fractional-order PI (FOPI) loops are tuned using frequency-domain specifications. For maintenance, a disturbance-benchmarking profile, specific to the high-risk procedure of liver transplantation, is proposed and designed in this work to test controller robustness against abrupt and recurrent hemodynamic perturbations. Simulation results confirm that a population-based, decentralized fractional-order PI controller ensures fast convergence, accurate reference tracking, and robustness against inter-patient variability and external disturbances across ten patient parameterizations.

Keywords: Control in system biology, Adaptive control design, Nonlinear time-delay systems
Abstract: Multivariable control using steady-state decoupling is computationally attractive but theoretically limited by residual dynamic cross-coupling at physiological frequencies. This paper demonstrates that event-based execution fundamentally alters this limitation through temporal decorrelation, achieving 41% coupling reduction in a 4 × 4 anesthesia system despite using only steady-state decoupling. Four independent fractional-order PI/PID controllers, one per decoupled loop, are compared under fixed-rate and event-based execution strategies across 24 virtual patients. Event-triggered updates decorrelate substantially faster than fixed-rate execution (threshold-based: 35.6% improvement, p < 0.001; integral timescale: 32.0% improvement, p < 0.001), disrupting temporal correlation chains that amplify cross-coupling beyond steady-state predictions. This mechanism yields 46.6% faster system-wide induction, 43.7% improved disturbance rejection, and 80% computational cost reduction compared to fixed-rate discrete-time implementation with identical controllers. Results suggest that execution strategy is a critical design variable for multi-input multi-output control, demonstrating that intelligent sampling can overcome fundamental architectural limitations without complex dynamic models.

Keywords: Linear fractional-order systems
Abstract: This paper proposes an improved frequency domain identification method for stable fractional-order systems of commensurate-order. The notion of normalized root stability is introduced as optimization constraint to enforce system stability. This notion is applied to the identification of models for fast corrector magnets designed for PETRA IV, the fourth generation synchrotron light source currently under development at Deutsches Elektronen-Synchrotron (DESY). The performance of the proposed method is validated against reference methods for fractional-order identification.

Keywords: Adaptive control design, Robust controller synthesis, Linear systems
Abstract: The classical self-tuning regulator (STR) relies on the certainty equivalence hypothesis, which often leads to performance degradation or even instability when parameter estimation is not sufficiently accurate. This paper proposes a self-tuning disturbance observer (DOB) as an inner-loop controller for STR, which is a variant of Q-filter-based DOB. The proposed self-tuning DOB forces the inner-loop system to behave like the model estimated by a recursive least squares algorithm instead of a nominal model, so that the certainty equivalent hypothesis actually meets dynamically from the point of view of the outer-loop STR. A stability condition is analyzed by the singular perturbation theory, which clarifies the interaction between the fast inner-loop dynamics of DOB and the slow parameter adaptation process.

Keywords: Optimal control theory, Applications of optimal control, Uncertain systems
Abstract: This paper addresses the infinite horizon optimal control problem under time-varying external disturbances through a variational-based framework. Beyond model uncertainty and the strong coupling between the estimator and the optimal controller, additional challenges will arise in the infinite horizon problems due to the requirements for cost convergence and persistent stability. By integrating the disturbance estimation into the controller design, a complete analytical characterization of the dependence of optimality loss on estimation error is established. It is demonstrated that the variation in the optimal solution is a linear functional of the disturbance estimation error. Furthermore, both the variations in the optimal solution and the optimal cost can be quantitatively assessed by the estimation inaccuracy. Finally, the effectiveness of the proposed method is shown by numerical simulations.

Keywords: Optimal control theory, Lagrangian and Hamiltonian systems, Real-time optimal control
Abstract: This paper addresses the trajectory optimization problem for flight vehicles under free-final-time conditions. The severe nonlinearity of the vehicle dynamics, together with uncertainty in the model parameters—interpreted as deviation from their nominal values—makes real-time trajectory optimization particularly challenging. To overcome this difficulty, we propose the extended neighboring optimal control (ENOC) framework that unifies initial-state deviation, terminal-condition deviation, and parameter deviation within a second-order variation analysis, thereby yielding a neighboring-optimal feedback law capable of compensating both boundary-condition and parameter deviation. Building on this framework, the extended state observer (ESO) is incorporated to estimate the parameter deviation in real time. Simulation results demonstrate that the proposed method maintains near-optimal performance and high terminal accuracy in the presence of parameter deviation as well as initial-state and terminal-condition deviations.

Keywords: Robust controller synthesis, Robust linear matrix inequalities, Uncertain systems
Abstract: This paper addresses the design problem of disturbance observers for a class of uncertain linear plants affected by external disturbances. The disturbance entering the plant is assumed to be generated and modeled by a linear system whose eigenvalues all lie in the closed right-half complex plane. Under this setting, we propose a design method that asymptotically rejects the effect of the modeled disturbance on the closed-loop system, whereas conventional disturbance observers mainly attenuate it in an approximate sense. This is achieved by implicitly embedding an internal model of the disturbance into one of the low-pass filters of the disturbance observer. In contrast to existing internal-model-based designs that require the solvability of certain linear equations and thus restrict the class of disturbances to polynomials in time and/or sinusoids with distinct frequencies, our method does not impose such restrictions. With the proposed disturbance observer, we show that the closed-loop system is robustly stable and provide simulations to validate it.

Keywords: Robust controller synthesis, Distributed robust controller synthesis, Uncertain systems
Abstract: This paper presents a robust cooperative adaptive cruise control (CACC) strategy for heterogeneous vehicle platoons subject to parameter uncertainties and external disturbances. A disturbance observer (DOB) is employed in the inner-loop to compensate for model mismatches, achieving nominal performance recovery. By doing so, the uncertain heterogeneous platoon is effectively treated as a homogeneous system governed by a nominal model. Based on these compensated dynamics, a simple outer-loop CACC is proposed that integrates a dynamic feedforward filter utilizing the desired acceleration of the preceding vehicle with a proportional-derivative (PD) feedback controller to theoretically guarantee string stability. Simulation results validate that the proposed method effectively maintains string stability against severe heterogeneity and disturbances.

Keywords: Robust control applications, Uncertain systems, Robust controller synthesis
Abstract: The inverse model-based disturbance observer (DOB) that incorporates the nominal inverse of a given uncertain plant is a kind of robust controller for estimating and compensating for the combined effect of external disturbance and model uncertainties. Despite the high performance and transparent structure, one of the limitations is that it cannot be directly applied to non-minimum phase systems. In this paper, we try to overcome this limitation by employing a generalized sampler replacing the conventional sampler used in the sampled-data control system. Applying the zero-assignment ability of this new sampler, we obtain a discrete-time model that is of minimum phase so that the disturbance observer design can be applied. In addition, a robust version of the generalized sampler together with its design is introduced to effectively cope with the plant uncertainty. A robust stability condition of the closed-loop system is also proposed, and the effectiveness of the result is validated through numerical simulations.

Keywords: Nonlinearity learning from data, Stability of nonlinear systems, Application of nonlinear analysis and design
Abstract: This work proposes a novel model augmentation framework for learning discrete-time Lur’e-type systems in which the existing linear-time-invariant first-principles equations are augmented with static nonlinearities captured by neural networks. We first present conditions guaranteeing that the resulting augmented model is contractive and admits a unique, bounded steady-state solution to any bounded input. The latter property facilitates training of the neural networks directly based on steady-state responses. Unlike conventional approaches that require constrained enforcement of stability properties during training, our method employs a direct parameterisation technique, enabling scalability to large scale systems. The result is a scalable and contraction enforcing learning framework that improves model accuracy while retaining the inherent properties of the first-principle model. The effectiveness of the approach is demonstrated through a simulation case study of a nonlinear oscillator.

Keywords: Learning methods for optimal control, Output feedback nonlinear control, Robust controller synthesis
Abstract: Designing stabilizing control policies for nonlinear systems while optimizing complex objectives remains a formidable challenge. Neural Networks (NNs), despite their expressive power, can be highly sensitive to small input perturbations and can easily destabilize the closed-loop system. Existing approaches often impose explicit constraints on the controller’s parameters to ensure stability, but this typically leads to extra computational overhead. To address this issue, we leverage recently proposed Structured State-Space Models (SSMs) to parametrize discrete-time control policies for nonlinear systems. Our key contribution is a new free parametrization of Linear Time Invariant (LTI) systems with a prescribed mathcal{L}_2-gain, which we use to construct the L2-Recurrent Unit (L2RU) architecture, an SSM layer that enforces the desired mathcal{L}_2-bound emph{by design}. This result can be leveraged to guarantee closed-loop stability via the small-gain theorem or the so-called performance-boosting framework, independently of the controller’s optimization parameters, thereby enabling fully unconstrained optimization of general nonlinear objectives. Furthermore, the structure induced by the proposed parametrization allows efficient processing of long input sequences, as it is highly parallelizable through algorithms such as parallel scan. We demonstrate the effectiveness of this approach on a formation control task for mobile robots, where the L2RU-based controller ensures collision and obstacle avoidance while maintaining stability and performance.

Keywords: Nonlinearity learning from data, Distributed nonlinear control, Infinite-dimensional multi-agent systems and networks
Abstract: In this paper we consider the limiting case of neural networks (NNs) architectures when the number of neurons in each hidden layer and the number of hidden layers tend to infinity thus forming a continuum, and we derive approximation errors as a function of the number of neurons and/or hidden layers. Firstly, we consider the case of neural networks with a single hidden layer and we derive an infinite width integral neural representation that generalizes existing continuous neural networks (CNNs) representations. Then we extend this to deep residual CNNs that have a finite number of integral hidden layers and residual connections. Secondly, we revisit the relation between neural ODEs and deep residual NNs and we formalize approximation errors via discretization techniques. Then, we merge these two approaches into a unified homogeneous representation of NNs as a Distributed Parameter neural Network (DiPaNet) and we show that most of the existing finite and infinite-dimensional NNs architectures are related via homogenization/discretization with the DiPaNet representation.

Keywords: Nonlinear observers and filters, Observer design, Application of nonlinear analysis and design
Abstract: The Kazantzis-Kravaris-Luenberger (KKL) observer provides a general framework for nonlinear state estimation by immersing the system dynamics into a stable linear or nonlinear latent dynamics. However, the performance of KKL observers relies heavily on the specific choice of these latent dynamics, which is often heuristic. This paper proposes a methodology to learn a KKL observer that combines global stability guarantees with local optimality. We derive a condition on the latent dynamics such that the observer locally mimics the behavior of a Minimum Energy Estimator (Mortensen observer). We then employ Deep Learning to approximate the KKL transformation and the latent dynamics, using neural network architectures that structurally enforce the contraction property. The proposed strategy is validated through numerical simulations on nonlinear benchmarks, demonstrating a good performance in the presence of state and measurement noise.

Keywords: Nonlinearity learning from data, Stability of nonlinear systems, Lyapunov methods
Abstract: Hybrid models combine physics-based models with data-driven components to achieve high accuracy while maintaining interpretability. However, their performance can degrade when extrapolating beyond the training data, often producing physically inconsistent predictions and compromising stability. The aim of this paper is to address this challenge by enforcing consistency with the underlying physics-based model away from the training dataset while maintaining accuracy on it. We propose a framework in which we modify existing hybrid models via a transition mapping that modulates the contribution of the data-driven component. Design constraints are imposed on the transition mapping to guarantee stability-preserving extrapolation based on known stability properties of the physics-based model. To enable scalable application to high-order systems, we introduce an unconstrained parametrization of the transition mapping that satisfies the design constraints by construction for any given hybrid model, regardless of model order or structure. We provide theoretical results establishing well-posedness and stability guarantees inherited from the underlying physics-based model. A simulation-based case study illustrates the effectiveness of the approach.

Keywords: Nonlinearity learning from data, Robust learning systems
Abstract: This paper presents a method that learns a regionally stable recurrent neural network model from a set of input-output data generated by an unknown dynamical system. Relying on generalized sector conditions on the deadzone activation function, we first derive sufficient conditions that guarantee forward invariance on a compact set of the state space for any inputs from a given set. Such regional properties lead to less conservative conditions compared to variants that offer a global form of stability, and are in line with the system data that is only observed regionally. We then present a learning method that imposes the derived conditions for regional stability using a barrier function approach, leading to models equipped with a certificate of regional stability in a subset of the state space and for a given input set. We illustrate our theoretical result with a numerical example and compare it to methods that impose a global form of stability, which fail to identify the system, and with a method that imposes no stability constraints at all, which does not guarantee a stable behavior within any state or input set.

Keywords: Supply chain and logistics engineering, simulation and optimization, Data-driven and AI-based modelling of production and logistics, Manufacturing plant simulation, control and optimization
Abstract: The increasing adoption of digital twins in industrial and logistics systems has profoundly changed how organizations plan and control their operations. However, the reliability of digital twin-based decision support critically depends on the realism of the demand forecasts used as inputs. This paper presents a structured tool-supported methodology for multi-attribute demand forecasting specifically designed to feed digital twins with coherent and simulation-ready data. The approach combines independent time-series forecasting, correlation reconstruction, and heuristic assignment to generate detailed synthetic demand datasets that preserve the interdependencies observed in historical data. The proposed framework is validated through an industrial application in a distribution warehouse, where it supports daily capacity planning decisions. Results demonstrate high forecasting accuracy across key demand attributes, structural coherence with historical distributions, and significant improvements in the realism and usefulness of simulation-based analyses. The study contributes to bridging the methodological gap between forecasting and digital twin integration and outlines research perspectives toward probabilistic and adaptive decision-support frameworks.

Keywords: Supply chain and logistics engineering, simulation and optimization, Supply network dynamics and control, Supply chain management in manufacturing
Abstract: This paper addresses sourcing and transportation planning in an aeronautics supply chain under cost, CO2 emissions, and delivery-risk considerations.We propose the architecture of a hybrid optimization–simulation decision-support framework in which a deterministic mixedinteger linear programming (MILP) model generates sourcing and transport plans, while a discrete-event simulation (DES) model is intended to evaluate their robustness under operational uncertainty. The present paper focuses on the formulation of the MILP decision layer and on the definition of its interaction with an existing DES model previously validated on the same industrial context. Using industrially inspired data, we illustrate how the MILP reacts to changes in service, carbon, and capacity-related parameters, and how such scenario analyses can support sourcing and transport decisions. The automated closed-loop interaction in which DES results are used to update MILP parameters is left for future work.

Keywords: Intelligent manufacturing systems, Simulation and optimization in production, operations and services, Data-driven and AI-based modelling of production and logistics
Abstract: The dynamic nature of manufacturing systems requires agile monitoring approaches to represent and assess the system’s behavior accurately. Discrete-Event-Simulation (DES) models have been demonstrated in the literature to provide proactive insights into manufacturing operations. Despite this prospect, the design and maintenance of DES models requires substantial cost and resources. A minor alteration to the physical process flow can render a model obsolete. This paper addresses this challenge by dynamically extracting an optimized DES workflow from large manufacturing datasets, using it to run simulations on these data and identify the most advantageous improvement scenario. A Heterogeneous Graph Transformer (HGT) is trained on real and simulation-generated data to predict key performance indicators such as waiting time, service time, total duration, and resource utilization. The model achieves high accuracy, with R2 scores up to 0.98 for machine-usage prediction. The proposed pipeline serves as a scalable surrogate for system understanding, enabling integration into optimization and scheduling applications.

Keywords: Cyber-physical production systems, Smart production and logistics in manufacturing, Manufacturing plant simulation, control and optimization
Abstract: Matrix-structured manufacturing systems (MMS), in which products flow through stations arranged in a matrix-shaped grid, have recently gained attention as an alternative to traditional assembly lines for mixed-model assembly. As MMS allow multiple possible product routes through the system even for identical products, planning material supply to stations becomes considerably more complex. Despite this, material supply planning and its interdependencies with system design have not yet been systematically addressed for MMS. This article presents an approach for material supply planning in MMS. To this end, a mathematical optimization model is presented. Based on the model's implementation as a mixed-integer programming model, numerical examples demonstrate the effectiveness of the proposed approach.

Keywords: Sustainable and circular supply chain and production, Supply chain management in manufacturing
Abstract: Circular manufacturing systems are challenged by uncertainties in the quality, quantity, and timing of returned products, which complicate planning and condition assessment in reverse logistics. This paper proposes an AI-enabled Decision Support System (AIxDSS) designed to support decision-making under such uncertainty through predictive analytics and explainable machine-learning models. The AIxDSS evaluates component reusability, derives human-interpretable rules, and supports routing decisions for reuse, remanufacture, or recycling. A case study on electronic component recovery demonstrates how decision-tree models and feature analysis improve transparency, prediction reliability, and decision consistency. The approach contributes to Zero-Defect Re-Manufacturing (ZDRM) by enabling early defect prevention and quality-oriented process control. The results show that integrating explainable predictive models into DSS architectures can enhance uncertainty management, operator understanding, and overall efficiency in circular manufacturing environments.

Keywords: Intelligent manufacturing systems, AI-based enterprise systems, Cyber-physical production systems
Abstract: Component detection is crucial for the autonomy of intelligent production lines. However, the high-precision component detection remains challenging due to cluttered back-grounds, multi-view variability, and heterogeneous target appearances. This paper introduces Adaptive Inception-YOLO, a detection framework tailored for an intelligent manufacturing line. Firstly, an Inception-style backbone with large-kernel depthwise convolutions is built based on YOLOv11 to enlarge the effective receptive field and improve small-object perception in complex backgrounds. To improve robustness, we propose a dynamic multi-branch aggregation module with a learning gate, enabling the network to adaptively weight multi-scale branches according to input features. An experiment is carried out based on a real-world multi-view production-line dataset. Experimental results demonstrate the effectiveness of the proposed method in multi-view component detection with the mAP@50 of 0.9848, mAP@75 of 0.9786, and mAP@[0.50:0.95] of 0.8746, exceeding the performance of both the YOLOv11 baseline and a static Inception-enhanced variant.

Keywords: Digital enterprise, Human-centered production and logistics, Manufacturing engineering and management
Abstract: Digital Twins (DTs) increasingly require robust human-centered perception to support monitoring, safety, and autonomous decision-making across diverse real-world environments, including smart industries, for reliable and continuous operation. We introduce WhoFi, a Wi-Fi–based person re-identification (Re-ID) model that offers an alternative to vision-based methods, which often suffer from occlusion, visual degradation, and privacy constraints. By extracting biometric signatures from Channel State Information (CSI) and encoding them through a Transformer-based architecture, WhoFi enables resilient and privacy-friendly Re-ID. Evaluation on the NTU-Fi dataset, currently a key reference benchmark for complex Wi-Fi sensing tasks, demonstrates its effectiveness within adaptive, human-aware DT ecosystems.

Keywords: Intelligent manufacturing systems, Cyber-physical production systems, Industrial artificial intelligence
Abstract: Digital Twins (DTs) are increasingly used in manufacturing and supply chain systems to support real-time monitoring and decision-making. As these DTs become essential for operational continuity, their reliability under data, communication, or cyber disruptions becomes a critical requirement. While many studies focus on resilience by DTs how they help physical systems anticipate and recover from disruptions, there is limited attention to the resilience of DTs themselves, that is, their ability to remain accurate, functional, and trustworthy under uncertainty. This paper introduces the concept of a Resilient Digital Twin. Five resilience dimensions are identified: data, model, synchronization, architectural resilience, and cyber. These dimensions are mapped to the classical resilience capabilities of detection, response, recovery, and adaptation, providing a structured understanding of how resilience manifests within DT ecosystems.

Keywords: Industrial artificial intelligence, Manufacturing engineering and management, Digital transformation
Abstract: Industry 5.0 emphasises human-centred manufacturing and the use of cognitive Artificial Intelligence to support collaboration between experts and intelligent systems. In advanced manufacturing, technical process planning still depends heavily on manual interpretation of drawings, material information, production constraints, and expert knowledge. This paper presents a neuro-symbolic framework that integrates multimodal large language models, computer vision, and ontology-based reasoning to support manufacturing plan generation. The proposed system extracts structured data from technical drawings, classifies part geometry using a YOLO-based model, and populates a domain ontology enriched with semantic rules and tacit knowledge. The results indicate that combining neural perception with symbolic reasoning can improve decision support, reduce manual effort, and generate manufacturing sequences consistent with expert-developed plans.

Keywords: Cyber-physical production systems, Digital enterprise, Industry X.0 for production and logistics
Abstract: This paper presents a systematic mapping study on how Semantic Web technologies support cognition in digital twins and Industry 4.0/5.0 systems. A SCOPUS search (2019–2025) identifies 20 studies combining ontologies, knowledge graphs, reasoning frameworks, and semantic integration tools with cognitive capabilities in industrial or robotic applications. Cognitive functions are classified using taxonomies by Metzler and Neisser. Results show strong emphasis on perception, memory, and low-level reasoning, while higher-order cognition is less developed. Semantic technologies dominate knowledge representation, but advanced reasoning and neuro-symbolic methods remain limited. The study highlights key gaps and opportunities for cognitive-ready semantic frameworks.

Keywords: Digital transformation, Industrial artificial intelligence, Digital enterprise
Abstract: Automated Feature Recognition (AFR) is a critical enabler of the digital thread in advanced manufacturing, translating low-level geometric data from CAD models into high-level semantic information for process planning, cost estimation, and inspection. Traditional AFR methods, often relying on rule-based or graph-based algorithms, struggle with robustness and adaptability when faced with the geometric complexity and variability of modern aerospace components. This paper introduces a novel hybrid AI-symbolic framework that integrates traditional geometric analysis with advanced artificial intelligence techniques, including multi-agent systems and ensemble learning. We present a comprehensive benchmarking suite of five distinct feature extraction methodologies, ranging from simple geometric parsers to sophisticated multi-agent AI systems. These methods were evaluated on a corpus of aerospace parts represented in the STEP format. The experimental results demonstrate that our top-performing hybrid model, the "Improved Multi-Agent AI," achieves a mean F1-score of 0.89 and an accuracy of 96.5%, significantly outperforming both traditional symbolic proxies and simpler AI-based extractors. This work demonstrates the synergistic potential of combining symbolic reasoning with generative AI to create a more robust, accurate, and versatile AFR solution for the manufacturing industry.

Keywords: Industrial artificial intelligence, Manufacturing prognostics and health management, Maintenance engineering, management and services
Abstract: The rise of Industry 4.0 technologies has revolutionized industries, enabled seamless data access, and fostered data-driven methodologies for improving key production processes such as maintenance. Predictive maintenance has notably advanced by aligning decisions with real-time system degradation. However, data-driven approaches confront challenges such as data availability and complexity, particularly at the system level. Most approaches address component-level issues, but system complexity exacerbates problems. In the realm of predictive maintenance, this paper proposes a framework for addressing drift detection and diagnosis in heterogeneous manufacturing processes. The originality of the paper is twofold. First, this paper proposes algorithms for handling drift detection and diagnosing heterogeneous processes. Second, the proposed framework leverages several machine learning techniques (e.g., novelty detection, ensemble learning, and continuous learning) and algorithms (e.g., K-Nearest Neighbors, Support Vector Machine, Random Forest and Long-Short Term Memory) for enabling the concrete implementation and scalability of drift detection and diagnostics on industrial processes. The effectiveness of the proposed framework is validated through metrics such as accuracy, precision, recall, F1-score, and variance. Furthermore, this paper demonstrates the relevance of combining machine learning and deep learning algorithms in a production process of SEW USOCOME, a French manufacturer of electric gearmotors and a market leader. The results indicate a satisfactory level of accuracy in detecting and diagnosing drifts, and the adaptive learning loop effectively identifies new drift and nominal profiles, thereby validating the robustness of the framework in real industrial settings.

Keywords: Model-predictive and optimization-based control in chemical processes, Industrial applications of process control, Advanced process control
Abstract: With the aim to lower the barrier for industries to adopt model predictive controllers (MPC), we derive analytical expressions for the terms in the cost matrices of an MPC formulation for a first-order system, such that the control signal exactly matches the output from a PI-controller when constraints are inactive. We solve this controller-matching problem with an inverse optimal control formulation by requiring no state-input cross-terms in the cost function, resulting in a MPC formulation with terminal costs. The resulting controller is validated with simulations based on a control scenario from the Swedish mining industry.

Keywords: Advanced process control, Model-predictive and optimization-based control in chemical processes, Industrial applications of chemical process control
Abstract: The core computational algorithm in Model Predictive Control (MPC) is based on real-time optimization with respect to operational constraints. This optimization problem is commonly solved using a quadratic programming algorithm. This paper proposes a solution of the optimization problem using a prime-dual Hildreth's algorithm with respect to interval constraints. The utilization of interval constraints reduces the number of constraints in half. More importantly, the algorithm is exceedingly simple for real-time implementation. A MATLAB program is presented in this paper for those who wish to try the proposed approach. end{abstract}

Keywords: Model-predictive and optimization-based control in chemical processes, Advanced process control, Real-time optimization and control in chemical processes
Abstract: The development of MPC schemes based on dual control framework (DMPC) has attracted significant attention over the last few years. Most of the approaches available in the literature are concerned with solving the target tracking problem rather than disturbance rejection. The unmeasured disturbances are correlated in time, and using their temporal relationships in controller synthesis can help in improving the regulatory performance. The ARMAX models provide a parsimonious representation of such autocorrelated signals. Therefore, in this work, we propose to use multiple MISO ARMAX models to improve the regulatory control performance of adaptive dual MPC (ADMPC) schemes. Using the concept of excitation horizon, the future predictions in the stochastic optimal control problem are split into the excitation and control components. For approximating the excitation term, a sampling-based approach using the unscented transformations is used to arrive at a computationally tractable ADMPC formulations. The efficacy of the proposed ADMPC schemes is evaluated by conducting simulation studies on the benchmark quadruple tank process. The simulation studies reveal that the proposed ADMPC schemes have the edge over ADMPC schemes that employ models with OE structure while dealing with unmeasured disturbances.

Keywords: Model-predictive and optimization-based control in chemical processes, Industrial applications of chemical process control, Advanced process control
Abstract: In the process industries, MPC (Model Predictive Control) is typically implemented as a two-stage controller with a Linear Program (LP) steady-state optimizer that generates economically optimal targets for the MPC algorithm. Abnormal behaviors in industrial LP optimizers are often difficult to rationalize, especially when a large number of manipulated variables (MVs) and controlled variables (CVs) are involved. We introduce a novel, post-hoc LP explainability method by recasting the role of shadow prices in the LP solution as an attribution mechanism for MV-CV relationships. The core idea is that the shadow price of a constrained CV is not just an intrinsic property of the LP solution, but can be split into contributions from individual unconstrained MVs and resolved into one-to-one MV-CV pairings using a linear sum assignment algorithm. The proposed MV-CV pairing framework serves as a practical explainability tool for online LP-MPC systems, enabling practitioners to diagnose suboptimal constraints and verify alignment of the controller's behavior with its original design.

Keywords: Process modeling, identification, and estimation techniques, Advanced process control
Abstract: This paper investigates a closed-loop system identification method for run-to-run controlled semiconductor processes in the presence of sampled and delayed metrology. It is well known that system identification under closed-loop operation can lead to biased estimates if the effect of feedback is ignored. However, we derive a condition involving sampling interval and measurement delay under which accurate identification is possible under closed-loop environment. We assume a static process model with an integrated first-order moving average (IMA(1,1)) disturbance and an exponentially weighted moving average (EWMA) controller. A mathematical analysis is first carried out for a single-input single-output (SISO) control system, from which an identifiability condition that depends on the sampling interval and measurement delay is derived. Furthermore, simulation studies of both SISO and multiple-input multiple-output (MIMO) control systems are presented to demonstrate the validity of the derived condition.

Keywords: Motion control for AVs, Autonomous vehicles, Trajectory tracking and path following for AVs
Abstract: Adaptive Cruise Control (ACC) is an Advanced Driver Assistance System (ADAS) that enhances vehicle safety by regulating ego throttle and brake inputs to sustain a desired speed or a safe following distance. Conventional ACC often depends on costly radar sensors or computationally intensive learning-based perception to track surrounding vehicles. This paper proposes a low-cost, vision-based ACC framework that relies solely on a monocular camera for vehicle detection, tracking, and ego control. The method uses a model-based high-gain observer with the YOLO computer-vision algorithm to estimate surrounding vehicles’ trajectories directly from ego-vehicle camera frames. These estimates are incorporated into a Model Predictive Control (MPC) scheme to achieve real-time ACC functionality. The proposed vision-based observer–MPC technique is validated in the CARLA simulation environment by demonstrating stability, real-time feasibility, and applicability to practical real-world driving scenarios.

Keywords: Distributed optimization for smart grids, Distributed optimization and control for smart cities
Abstract: This paper proposes a distributed estimation method for the voltage and phase angle of power systems. We assume that each node measures active and reactive power with Gaussian noise. We construct an optimization-based approach to state estimation, based on maximum likelihood. The first-order necessary conditions of the optimization problems are equivalent to systems of non-linear equations.

To calculate these equations, we use a sequential procedure inspired by the Gauss-Seidel method and a parallel procedure inspired by the Jacobi method. In addition, we propose a modified parallel procedure such that each node uses the two previous steps’ estimates of its two-hop neighbors. We evaluate the effectiveness of the proposed method through numerical experiments using five IEEE bus system models.

Keywords: Electrical distribution systems, Electrical transmission systems, Control and management of energy systems
Abstract: High penetration of renewable energy resources increases operating variability and challenges the regulation capability of grid-following inverters, placing greater burden on synchronous generators. While grid-forming inverters enhance stability, their droop coefficients are typically tuned heuristically, remain fixed in operation, and do not account for temporally varying uncertainty in power sources and loads. This paper derives the steady-state relationship between droop parameters and network dynamics and introduces a sensitivity metric for bounded, time-varying load perturbations. A worst-case robust optimization with adaptive droop updating is developed. The proposed framework enables seamless temporal functional adjustment and improves voltage-frequency regulation and robustness under dynamically evolving uncertainty. In simulations, we evaluate two test cases with different perturbation levels to illustrate time-varying uncertainties. The proposed method achieves performance improvements of 27.08% and 4.76% compared with a fixed-droop baseline.

Keywords: Power plant control, Control and management of energy systems
Abstract: Permanent magnet synchronous generators (PMSGs) are widely used in microgrid, wind, and tidal power systems to provide reliable and efficient renewable energy generation. However, their inherent nonlinear dynamics and state constraints, arising from physical and safety limits, pose significant challenges for conventional controllers. Proportional–Integral–Derivative (PID) control cannot explicitly handle this complexity, and traditional model predictive control (MPC) approaches with fixed weighting matrices often fail to ensure stability when the operating conditions of the PMSG vary. To address these issues, we propose a novel terminal control law and terminal cost function design for the nonlinear MPC (NMPC) that explicitly accounts for the nonlinear dynamics and constraints of the PMSG. Simulation results verify the effectiveness of the proposed scheme and highlight its advantages over linear MPC, showing the potential to go beyond linear MPC for micro-grid control while still guaranteeing performance.

Keywords: Power systems stability
Abstract: The recent rapid proliferation of renewable energy is fundamentally changing the dynamic operations of power systems, necessitating new approaches to assess stability for these highly nonlinear systems. In this paper, we prove that synchronous machine systems, modeled in the nonlinear dq frame, possess fundamental dissipativity properties. Specifically, we show passivity from current input to voltage output and a nonlinear negative imaginary property from torque input to rotor angle output. For the nonlinear system shifted around an equilibrium point, we derive explicit conditions for both passivity and the NI property to hold. Finally, we demonstrate that interconnection with passive droop controllers preserves these dissipativity properties with identical supply rates, thereby ensuring closed-loop stability.

Keywords: Power systems stability
Abstract: This paper investigates equilibrium points and stability in two synchronous machine configurations: (i) a single generator with an impedance load and (ii) two interconnected machines with co-located loads. We consider both abc and dq reference frames to show that the equilibrium condition reduces to a cubic polynomial in the single-machine case and to an 18th-degree polynomial in the two-machine case. For the single-machine system, Lyapunov stability analysis and linearization based stability analysis are carried out. For the two-machine system, local stability is assessed through linearization and eigenvalue analysis. Illustrative examples confirm the existence of multiple equilibria and illustrate the impact of parameter variation on stability. Our results provide insight into the stability of synchronous machine systems.

Keywords: Power systems stability, Wind power, Fault-tolerant control methods
Abstract: This paper proposes an event-triggered global sliding mode control (ET-GSMC) strategy for the grid side converter (GSC) of a doubly fed induction generator (DFIG)-based wind energy conversion system (WECS). Unlike the standard sliding mode control (SMC) approach, ET-GSMC establishes the DC-link voltage stability from the initial moment by eliminating the reaching phase and mitigates the chattering phenomenon by reducing the switching frequency. Hence, the proposed controller allows the DC-link capacitor to assist the rotor side converter (RSC) for seamless bidirectional power transfer with the grid. The effectiveness of the proposed ET-GSMC is assessed by computer experiments in the MATLAB/Simulink environment. The results are further compared with those of the standard SMC. The assessment was performed under various faulty conditions, including single-line-to-ground fault, three phase symmetrical fault and balanced load variation. The results confirmed the effectiveness and improved dynamic performance of the DFIG-based WECS in the presence of faulty conditions.

Keywords: Energy management systems, Forecasting of power supply and demand, Big data and machine learning applied to smart cities
Abstract: Non-intrusive load monitoring (NILM) faces significant challenges in maintaining performance when deployed to unseen users or new environments due to limited generalization capability. Accordingly, this paper proposes FedStacking-NILM, an enhanced framework that integrates federated learning with stacking ensemble methodology for non-intrusive load monitoring. The framework employs three federated learning algorithms, including FedAvg, FedProx, and FedAC, as base learners to capture diverse load characteristics while preserving data privacy. A Kolmogorov-Arnold Network (KAN) based meta learner then effectively integrates the base learners' outputs through sophisticated nonlinear fusion. Comprehensive evaluations on public datasets demonstrate the effectiveness of FedStacking-NILM, which enhances identification accuracy when tested on unseen households. The framework maintains robust performance across diverse appliance types while ensuring data privacy protection.

Keywords: Energy market
Abstract: We propose a tractable convex hull pricing (CHP) approximation for multi-period electricity pricing. The method represents single-period non-convexities using the exact convex biconjugate of the bus-level cost function, while intertemporal constraints are handled via continuous relaxation of commitment variables, as in Extended Locational Marginal Pricing (ELMP). Simulations on a modified PJM 5-bus system show that the hybrid formulation can reduce lost opportunity cost uplift relative to ELMP in test cases with nonlinear variable generation costs, while remaining more tractable than a discretized CHP benchmark.

Keywords: Energy market, Power systems stability, Energy management systems
Abstract: This paper presents a learning framework for economic dispatch in a day-ahead electricity market subject to two main sources of uncertainty in (net) loads and cost bids. Uncertain loads add significant challenges to power balance, threatening power system safety, while uncertain cost bids may render system operation economically inefficient. To jointly tackle the two issues, we propose a robust economic dispatch formulation that hedges against worst-case bid deviations from actual costs and further provide statistical feasibility guarantee for solutions to satisfy all physical and operational constraints. Without the knowledge of true distributions of these randomness, we develop a data-driven approach that in parallel learns feasible regions and robust optimal decisions. In particular, we employ conformal prediction with an affine recourse policy to ensure feasibility with statistical validity. We then adopt an end-to-end learning framework that embed a robust optimization layer in training to acquire dispatch decisions directly from historical data of realized marginal costs. This approach mitigates cost bid risks, respects constraints (with high probability), and, more importantly, represents a full paradigm shift from prediction-based optimization to task-oriented learning. Numerical simulations on the IEEE 39-bus system validate that our framework robustly reduces system costs in a variety of scenarios with feasible dispatch decisions.

Keywords: Forecasting of power supply and demand
Abstract: This paper proposes a 24-hour ahead load forecasting framework that integrates the Temporal Fusion Transformer (TFT) with the Grey Wolf Optimizer (GWO) to address forecasting challenges in Taiwan’s power system, which is affected by renewable energy integration, industrial variability, and extreme weather. The TFT component captures multi-scale temporal dependencies and provides interpretability, while the GWO is employed for systematic hyperparameter optimization to improve accuracy and robustness. The framework is validated using nine years (2017~2025) of Taiwan Power Company operational data, including load profiles, meteorological variables, and economic indicators. Compared with benchmark models, the proposed TFT-GWO achieves superior results, with MAE (mean absolute error), RMSE (root mean squared error), and R2 (coefficient of determination) of 0.0145, 0.0220, and 0.9390, respectively. The results demonstrate that the proposed approach supports more reliable unit commitment, reserve allocation, and day-ahead market operations, highlighting the effectiveness of combining transformer-based architectures with metaheuristic optimization in power system forecasting.

Keywords: Power systems stability, Electrical transmission systems, Electrical distribution systems
Abstract: Stability phenomena induced by outer converter control loops are attributed to slow converter-driven stability. These can occur, when the converter input admittance exhibits non-passive frequency bands in the lower frequency range. This paper focuses on the two-channel negative feedback interconnection (NFI) structure of the converter-grid interaction (CGI) and evaluates suitable passivity-based stability theorems capable of addressing this lack of passivity by leveraging directional and relative passivity indices. The present work identifies the most appropriate stability theorem for systems exhibiting non-passive behavior. In particular, the so-called Small R_f Theorem is extended to be applicable to a broader class of such systems, addressing a gap in existing literature. The extended theorem is then employed to analyze the stability of CGI under subsynchronous resonance (SSR), highlighting its practical relevance.

Keywords: Forecasting of power supply and demand, Energy management systems, Demand response
Abstract: Accurate energy forecasting is essential for reducing demand charges. This paper benchmarks eXtreme Gradient Boosting (XGBoost), Long Short-Term Memory (LSTM), CNN-LSTM-Attention, and Temporal Fusion Transformer (TFT) using four uncertainty quantification methods. Validated across multiple commercial buildings, results reveal an accuracy-calibration paradox: TFT delivers state-of-the-art accuracy but severe under-coverage. In contrast, XGBoost with Quantile Regression (QR) provides the best operational trade-off with adaptive intervals. While Conformal Prediction (CP) ensures safety, it lacks sharpness. We demonstrate that operational reliability requires balancing point accuracy with calibrated uncertainty to guide risk-aware energy management.

Keywords: Process modeling, identification, and estimation techniques, Interaction between design and control in processes, Machine learning and artificial intelligence in chemical process control
Abstract: Machine learning models are increasingly used to model chemical process systems, yet they often lack principled uncertainty quantification and mechanisms to enforce physical constraints. We propose a probabilistic neural network framework that guarantees satisfaction of linear equality constraints within a given tolerance, while capturing aleatoric uncertainty. Compared to state-of-the-art methods, our formulation demonstrates improved predictive accuracy, uncertainty calibration, and adherence to constraints on reduced data. It also demonstrates competitive performance, but with significantly faster training times when evaluated on large data regimes. We evaluated this on two batch reactor case studies, enforcing mass balances.

Keywords: Process modeling, identification, and estimation techniques, Machine learning and artificial intelligence in chemical process control, Industrial applications of chemical process control
Abstract: Vapor compression cycles (VCCs) are fundamental to heating, cooling, and thermal management systems, but their phase-changing heat exchangers make dynamic modeling challenging for real-time optimization and control. Moving boundary (MB) models provide compact and physically interpretable representations, yet variations in active phase regions introduce mode-switching and non-smooth dynamics. This study proposes a phase-continuous hybrid modeling framework that combines an MB-based physical model with a multilayer perceptron (MLP)-based latent parameter estimator, yielding a single model structure without explicit switching logic. Latent parameters are estimated from the current operating condition and substituted into the reduced mass and energy balance equations. The proposed model is validated against a high-fidelity MATLAB/Simscape reference simulation. On a temporally segmented test set, the proposed model accurately reproduces the reference simulation with reduced computational time. Moreover, compared with a black-box MLP baseline, it retains physical interpretability and conservation-law consistency, while maintaining competitive prediction accuracy. This framework provides a numerically stable and physically structured basis for future optimization and control applications.

Keywords: Process modeling, identification, and estimation techniques, Advanced process control, Machine learning and artificial intelligence in chemical process control
Abstract: In this paper, we propose a data-enabled moving horizon estimation (MHE) approach for nonlinear systems. While the approach is formulated by leveraging Koopman theory, its implementation does not require explicit Koopman modeling. Lifting functions are learned from the state and input data of the original nonlinear system to project the system trajectories into the lifted space, where the resulting trajectories implicitly describe the Koopman representation for the original nonlinear system. A convex data-enabled MHE formulation is developed to provide real-time state estimates of the Koopman representation, from which the states of the nonlinear system can be reconstructed. Sufficient conditions are derived to ensure the stability of the estimation error. The effectiveness of the proposed method is illustrated using a membrane-based biological wastewater treatment process.

Keywords: Control of multi-scale, distributed, and particulate systems, Machine learning and artificial intelligence in chemical process control, Model-predictive and optimization-based control in chemical processes
Abstract: This study investigates the application of Model Predictive Control (MPC) to regulate agglomerate size and porosity in the fluidized bed spray agglomeration process. To enable effective control, a data-driven modeling approach based on Koopman theory is employed. A coordinate transformation, approximated by a neural network, is used to map measured data to a lifted space where the process dynamics are rendered linear. This facilitates the use of computationally efficient linear MPC despite the inherent non-linearity of the process system. The identified Koopman model is benchmarked against a conventional linear model obtained via N4SID subspace identification. The models are compared based on their prediction error on an independent test data set. A subsequent simulation study assesses the full MPC controller performance for setpoint tracking and disturbance attenuation tested on a nonlinear population balance model. The results demonstrate that the Koopman-based model significantly outperforms the N4SID model in both predictive accuracy and overall controller performance, validating the Koopman framework as a highly effective method for controlling complex agglomeration processes.

Keywords: Process modeling, identification, and estimation techniques, Machine learning and artificial intelligence in chemical process control, Hydrogen systems for energy generation and storage
Abstract: In this paper, we propose a data-driven control strategy for thermal and water management in a proton exchange membrane fuel cell. Based on data from a mathematical model, we obtained a linear time-invariant model using the Koopman operator framework. This linearization does not depend on a specific operating point and is valid across the entire range of training data. This enables the use of linear optimal control techniques for the underlying nonlinear system. Finally, we designed a Linear Quadratic Integral controller to achieve temperature and humidity set-point tracking and to effectively reject disturbances caused by load changes.

Keywords: Biomedical system modeling, identification, and simulation, Biomedical signal measurement and processing
Abstract: Auditory attention decoding (AAD) is typically treated as a backward system-identification problem, reconstructing speech features from electroencephalography (EEG). A central challenge is defining the “true” speech signal, since decoders often are trained on clean speech unavailable in real-world settings. We compare two ways of modeling speech mixture effects - clean attended/ignored mixtures, and four mixture methods: time-difference-of-arrival (TDOA)-based microphone estimates, room impulse response (RIR)-filtered speech, RIR-filtered EEG, and raw microphone signals. Our results show that ignored speech remains strongly represented in neural responses, with implications for future AAD models and adaptive noise-reduction strategies in closed-loop hearing aids.

Keywords: Dynamics and control of biologically motivated nonlinear systems, Modelling, parameter identification and state estimation in biosystems, Biological networks inference and modelling
Abstract: Early and noninvasive identification of Alzheimer's disease (AD) and its prodromal stage, mild cognitive impairment (MCI), is essential for enabling early clinical intervention. We propose a stage-sensitive analytical framework incorporating three nonlinear complexity measures—Higuchi's fractal dimension (HFD), spectral entropy (SE), and wavelet entropy (WE)—derived from resting-state functional near-infrared spectroscopy (fNIRS) recordings. Prefrontal fNIRS signals were acquired from 83 participants (AD: 19, MCI: 37, healthy controls (HC): 27), and complexity features were comprehensively characterized across hemoglobin types. Across groups, HFD exhibited a monotonic decline from HC through MCI to AD, while SE and WE were elevated in the AD group, indicative of increased signal irregularity. Classification models trained on a biologically informed core-channel feature set outperformed their full-channel counterparts, attaining an area under the receiver operating characteristic curve (AUC) of 0.889 with minimal fold-to-fold variability. Furthermore, the extracted complexity features showed strong associations with Mini-Mental State Examination (MMSE) scores, highlighting their clinical utility. Collectively, these results provide provide evidence that nonlinear complexity analysis of resting-state fNIRS signals can expose stage-specific neurovascular disruptions spanning the Alzheimer's continuum. The identification of robust, anatomically interpretable channel-level biomarkers positions resting-state fNIRS as a clinically applicable modality—moving beyond a supporting role toward active utility in the early diagnosis and disease staging of AD.

Keywords: Control of physiological and clinical variables, Pharmacokinetics, tracer kinetic modelling and drug delivery
Abstract: Propofol infusion during surgical procedures is a challenging task that aims to achieve and maintain clinical hypnosis: a state of unconsciousness to avoid intra-operative recollection. Within this paper, a closed-loop depth of hypnosis (DoH) controller based on H-infinity optimal control is proposed. Inter-patient variability is identified and utilized for the controller design to ensure robustness to modeling errors. The strategy is evaluated in silico considering modeling errors and external disturbances due to common surgical events. Results suggest the proposed controller is able to induce and maintain an appropriate level of hypnosis in the face of inter-patient and intra-operative variability.

Keywords: Healthcare management, disease control, critical care, Real time monitoring and control of environmental systems
Abstract: This study investigates the correlation between Near-Infrared Spectroscopy (NIRS) and transcutaneous oxygen pressure (T cP O2) for monitoring tissue oxygenation and wound heal- ing progression in diabetic feet. Signal-derived features, including area under the curve (AUC), reperfusion recovery speed, and peak amplitudes, were extracted and analyzed longitudinally to assess healing trajectories. The results demonstrate that NIRS and T cP O2 measurements provide complementary insights into microvascular function and can serve as reliable indicators of wound healing status. These findings highlight the potential of integrating non-invasive, signal-based monitoring techniques into personalized diabetic foot care and clinical decision- making.

Keywords: Artificial pancreas or organs, Healthcare management, disease control, critical care, Decision support and control in medicine
Abstract: Diabetes affects millions globally, especially in type 1 cases where precise blood glucose control is crucial. Current methods, relying on patient monitoring and insulin administration, often fall short. Hence, there's interest in using reinforcement learning (RL) to optimize management.

RL, a type of machine learning, shows promise in adjusting insulin dosages based on feedback. Our study examines a Proximal Policy Optimization (PPO) agent trained on average patient data, showing its effectiveness across diverse patient profiles. Our findings highlight the adaptability of PPO-based controllers in managing blood glucose levels effectively.

Keywords: Decision support and control in medicine, Artificial pancreas or organs, Control of physiological and clinical variables
Abstract: Meal timing is a critical behavioral determinant of glycemic control in type 1 diabetes (T1D), yet its characterization in free-living conditions remains limited. Understanding daily eating patterns may provide insights into glucose variability and inform personalized therapeutic strategies. This study aimed to derive eating behavior (meal timing and carbohydrate content) in adults with T1D using real-world dietary logs from the T1DEXI dataset and evaluate the effect of these behaviors on glucose control via in-silico testing. A total of 477 participants provided free-living records of breakfast, lunch, and dinner over an average of 16 days. Meal timing distributions were analyzed, and stratification was performed using k-means clustering based on temporal meal variability. Identified clusters were subsequently incorporated into an in-silico framework to assess their impact on simulated glycemic control. Fifteen distinct meal-timing phenotypes emerged. As a proof of concept, three representative phenotypes, nominal meal spacing (S1), compressed daytime (S2), and late dinner (S3), were evaluated in the UVa/Padova simulator using the UniBE hybrid closed-loop controller (10 virtual adults for 10 days per phenotype). Glycemic safety remained high across scenarios, with median time in range of 94.8% for S1, 92.1% for S2, and 95.3% for S3, and time below range approximating 0%. Modest but consistent increases in time above range and glucose variability were observed in S2 compared to S1 and S3. Behavioral phenotyping of meal timing revealed distinct clusters with quantifiable differences in glycemic outcomes. In-silico validation underscores their potential utility for personalized diabetes care and supports the integration of behavioral metrics into digital therapeutic strategies.

Keywords: Artificial pancreas or organs, Biomedical system modeling, identification, and simulation, Decision support and control in medicine
Abstract: Accurate meal detection and carbohydrate content (CHO) estimation are key aspects for effective closed-loop insulin delivery in Type 1 diabetes. In literature, effective approaches rely on the estimates of glycemia and its derivatives via a Kalman Filter (KF), evaluated with conditional logic. Recently, also machine learning techniques have shown promising results for this application. This work proposes an algorithm enabling fully automated meal bolus delivery. To estimate CHO, the KF estimations have been evaluated leveraging a Convolutional Neural Network (CNN) thus avoiding any case-specific conditional logic. Then, CHO estimations have been used to calculate and inject insulin meal boluses. The proposed CNN approach is compared with a reference reported in literature using the 100 in silico adults patients of the UVA/Padova simulator. Results demonstrate superior performance both in meal estimation capability (F1 score of 84.21% vs 64.15%) and in closed-loop performance (Tr 83.43% vs 75.58%, p-value < 0.001), highlighting the strong potential of the proposed method for meal detection and CHO estimation in fully automated closed loop applications, avoiding datasetspecific tuning..

Keywords: Artificial pancreas or organs, Biomedical system modeling, identification, and simulation, Control of physiological and clinical variables
Abstract: The next-generation neural-network-based automated insulin delivery (AID) systems rely on compact and informative state representations to achieve safe and efficient closed-loop control. Thus, understanding the intrinsic dimensionality of these AID state vectors is critical in explaining, validating, and improving NN-based insulin dosage algorithms. In this work, we apply beta-Variational Autoencoders (beta-VAEs) as a theoretically grounded probabilistic method to estimate this intrinsic dimensionality. By sweeping over a range of regularization coefficients beta, we analyze the reconstruction error, the total Kullback--Leibler (KL) divergence, and the degree of latent dimension collapse. This combined way allows us to identify an appropriate balance between reconstruction fidelity and the latent space regularization. The beta-sweep result shows an optimal region, beta in [10^{-3}, 10^{-1}]. Focusing on this range, we observe a clear emph{elbow point} when encoding the original state space, an 8-dimensional Kalman-filtered state representation inferred from the past 6 hours of glucose and insulin dynamics. Our findings indicate that these states lie on an approximately 4-dimensional nonlinear manifold. Thus, the proposed beta-VAE framework provides a compact, manageable, and generalizable latent representation for data-driven biomedical control systems.

Keywords: Pharmacokinetics, tracer kinetic modelling and drug delivery, Biomedical system modeling, identification, and simulation, Artificial pancreas or organs
Abstract: Most glucagon pharmacokinetics mathematical models consider a linear dose-plasma concentration relationship. However, evidence from clinical data suggests a nonlinear linkage between subcutaneous glucagon and plasma glucagon concentration. This work explores this connection using data from four different clinical trials, including doses from 100 to 500 ug, and a total of 44 participants with type 1 diabetes. To this end, first, a pharmacokinetics model is identified. Then, statistical mixed-effects models are exploited to characterize the relation between the identified parameters and glucagon doses. The results highlight a nonlinear relationship between the dose amount and the glucagon clearance from plasma. Taking this nonlinearity into consideration could improve mathematical models of glucagon pharmacokinetics or help characterize these patterns in other drugs.

Keywords: Aerial and space robotics, Aerospace mission control and operations, Urban air mobility
Abstract: This paper investigates the flight control of an over-actuated thrust-vectoring UAV equipped with four 2-DoF rotatable thrusters. To enable thrust vector control, a full-pose controller is employed as the high-level controller, and a novel Exact Bundled Redistributed (EBR) control allocation method is proposed to allocate the desired wrench command exactly to each actuator. The algorithm then determines the required thrust vectors, force magnitudes, and deflection angles for all thrusters. The proposed EBR control allocation method is developed based on the admissible force space of four vectored thrusters, where each actuator computes intersections on the bundled local admissible force space (LAFS). The overall framework integrates Truncated Wrench Allocation (TWA) and Post-Torque Enhancement (PTE), ultimately obtaining an exact allocation solution while ensuring direction preservation. Simulation results illustrate the feasibility of the proposed method by comparing feasible and infeasible control wrench commands with respect to the attainable force space, as well as through maneuver tracking along a square trajectory, thereby validating the performance of the EBR control allocation method in thrust-vectoring UAV applications.

Keywords: Guidance, navigation and control of aircraft and spacecraft, Aerospace mission control and operations, Aerial and space robotics
Abstract: This paper presents the design and flight test validation of a co-designed guidance and control system for a fixed-wing unmanned aerial vehicle (UAV). The co-design of all integrated control loops enables increased performance capabilities which are especially important for highly dynamic UAV missions. The control design is based on a mixed-sensitivity control approach to enable robust inner loop controls in the presence of uncertainties. The guidance design implements a look-ahead path following algorithm which enables superior performance metrics while considering the computational constraints present for small UAV systems. The co-designed guidance and control approach is applied to a small fixed-wing UAV and flight tested mimicking a highly dynamic urban flight trajectory. The flight-test results not only validate the seamless implementation of the developed algorithms on the available flight-control computer, but also demonstrate their robust performance during real-world operations.

Keywords: Guidance, navigation and control of aircraft and spacecraft
Abstract: Incremental Nonlinear Dynamic Inversion (INDI) is effective for UAV attitude control but remains highly vulnerable to Control Effectiveness Matrix (CEM) inaccuracies during actuator faults. {Furthermore, conventional continuous adaptation schemes impose prohibitive computational loads on embedded processors.} To overcome these challenges, this paper presents a resource-aware fault-tolerant attitude control framework based on Adaptive INDI (AINDI). A data-driven CEM reconstruction method is developed using pseudo-partial derivatives together with Recursive Least Squares (RLS) to compensate for aerodynamic uncertainties and actuator faults in real time. To reduce computational overhead and inherently prevent parameter drift, an Event-Triggered Mechanism (ETM) is incorporated to update the estimator only when the tracking error exceeds a prescribed threshold. Crucially, a Lyapunov-based robust auxiliary control law with a dynamic adaptive gain is designed to rigorously guarantee Uniformly Ultimately Bounded (UUB) stability, explicitly accounting for event-triggered residual errors. Numerical simulations demonstrate that the proposed approach delivers superior tracking accuracy and robustness against severe gusts and actuator jamming compared to conventional methods.