The trend of electrification of propulsion systems also introduced all-electric drive in the maritime sector. Maritime all-electric drive systems operate using an energy system containing a variety of components, such as batteries, internal combustion engines, or fuel cells. The introduction of new components in the energy system increases both the flexibility as well as the complexity of the system operation. The most commonly used rule-based control is no longer sufficient to solve the control problem. Consequently, the usage of advanced control strategies in maritime has become a topic of research in recent years. In the operation of a maritime energy system, several objectives are of interest as targets of the optimisation, including cost, emission, or an enlargement of component lifetime. Depending on the choice of objective, the control strategy can differ. By integrating multiple objectives in control, the operation is optimised to find the best working point to fulfil the different interests. This article first reviews the commonly used advanced control structures in the maritime, automotive, and building control sectors. A comparison is used to identify further potential for advanced control usage in marine applications. In addition, the implementation of advanced control is reviewed in architecture and optimisation algorithms. Secondly, the control objectives used in the literature are presented and analysed in terms of their usage and potential of the combination. Thirdly, the currently used validation strategies and published results are reviewed and interpreted in terms of potential and required future work. Lastly, open gaps in the state of research are identified and potential for future work is outlined.
In the shift to sustainable and renewable energy generation, wind energy conversion systems (WECSs) based on the Doubly-Fed Induction Generator (DFIG) have gained significant attention due to their lower converter rating, excellent efficiency, and controlled active and reactive power. However, classical control strategies for DFIG systems face significant challenges, including high nonlinear dynamics, parametric uncertainties, external disturbances, wind speed variability, and growing demand for electrical grid integration. Conventional Artificial Intelligence (AI) techniques, including Neural Networks (NNs) and Fuzzy Logic Controllers (FLCs), have been widely adopted to improve control robustness and system efficiency under severe conditions. Furthermore, optimization algorithms like Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) have been widely used to tune controller parameters and enhance dynamic performance. Besides, machine learning approaches, such as Support Vector Machines and Random Forests, have demonstrated promising capabilities in terms of wind speed prediction and system performance enhancement. More recently, Deep Reinforcement Learning has emerged as a promising approach for intelligent and adaptive control in challenging and complicated situations. This paper presents a comparative review of AI-based control techniques for grid-connected DFIG wind turbines. A unified taxonomy of AI-based DFIG control methods is proposed, and the reviewed approaches are comparatively analyzed in terms of control objectives, computational complexity, implementation feasibility, robustness, dynamic performance, and real-time applicability. Particular attention is given to research developments reported between 2022 and 2026 to determine emerging trends in these AI-based methods. Key challenges and future research directions are discussed to support the development of high-performance and reliable DFIG-based WECSs.
Sampling-based motion planning algorithms such as Rapidly-Exploring Random Tree (RRT) and its variants are widely used in complex environments for autonomous robot navigation; however, they face fixed step-size limitation, inefficient exploration in narrow passages, and localisation clustering that leads to suboptimal paths and excessive iterations. This study proposes a modified RRT framework integrating (i) a dynamic step-size mechanism and (ii) a dynamically expanding circle-based sampling strategy to mitigate localization effects and improve global exploration. The proposed approach enhances both convergence efficiency and path reliability without increasing algorithmic complexity. The method is validated across seven benchmark and custom-designed maps of varying complexity and is compared against RRT, RRT*, Informed RRT*, and A*. Experimental results demonstrate success rate improvements up to 98%, consistent reduction in iteration counts compared to classical RRT variants, and competitive computational time while maintaining near-optimal path lengths. The combined dynamic step-size and dynamic circle RRT achieves superior robustness and convergence efficiency, making it suitable for autonomous robotic navigation in constrained and cluttered environments.
Fault detection and diagnosis (FDD) is essential for ensuring the safe and economic operation of chemical plants, as well as minimizing downtime. This is why research in this area has been active in recent years. Typically, faults are detected and diagnosed only after they have occurred. Because of this the damage caused cannot be prevented. Predictive detection and diagnosis methods could prevent this by detecting and resolving faults before they occur. In this context, drifts and their detection and diagnosis are considered. Another complicating factor is that multiple faults may occur simultaneously and require diagnosis. How many methods address the predictive detection and diagnosis of faults? How effective are FDD methods at detecting and diagnosing simultaneous faults? To answer these questions, a systematic literature review was conducted using the SPAR-4-SLR protocol. Several hybrid approaches were compared based on criteria regarding their hybridity, requirements, diagnostic capabilities, goals, and applications. The methods were evaluated based on their ability to perform predictive FDD and diagnose simultaneous faults. Then, a qualitative decision-making logic was developed to support the selection of methods for different purposes. However, the evaluation results showed that none of the methods could perform early FDD and multiple-fault diagnosis. Conclusions were therefore drawn regarding promising strategies for developing new methods. Future research should focus on overcoming structural limitations and other obstacles to enable early fault detection and multiple-fault diagnosis.
This tutorial provides an overview of the generalized Lyapunov method (GLM) for analyzing input-to-state stability (ISS) of partial differential equations (PDEs). We begin by revisiting the classical Lyapunov method and the standard ISS-Lyapunov theorem, highlighting their limitations when applied to systems with complex boundary disturbances. In contrast, the GLM, based on the concept of generalized Lyapunov functionals (GLFs) that explicitly depend on the external input, offers greater flexibility and efficiency, particularly for PDEs with Dirichlet-type boundary disturbances. The main objective of this tutorial is to demonstrate how to systematically construct GLFs to establish ISS estimates in Lq norms with any q∈[2,∞] for different PDEs. Specifically, we consider three representative classes of PDEs: (i) an N-dimensional nonlinear parabolic equation with mixed nonlinear boundary disturbances, (ii) a first order nonlinear hyperbolic equation with boundary disturbances, and (iii) a second order linear hyperbolic equation, i.e., a wave equation, with boundary damping and disturbances. For each case, we provide step-by-step constructions of appropriate GLFs and derive explicit ISS estimates, illustrating the general applicability of the GLM. Finally, we discuss open challenges and future directions, including the systematic construction of GLFs for broader classes of PDEs and their applications in controller design.
With the rapid advancement of distributed drive electric vehicles (DDEVs), ensuring yaw stability under extreme conditions has become a critical challenge. Although electrification and intelligence have reshaped chassis architectures, the principles of wheel force transmission are unchanged. Current yaw stability control strategies, reliant on estimating rather than measuring tire-road forces, encounter a fundamental bottleneck. The lack of direct wheel force sensing (WFS) primarily causes well-known challenges: (1) insufficient state estimation accuracy under varying road conditions, and (2) limited algorithm adaptability to nonlinear force transmission. While prior reviews cover yaw stability control or sensor technologies separately, few integrate them for DDEVs. To bridge this gap, this paper first analyzes the bottleneck of model dependency. It then reviews yaw stability control strategies for DDEVs, highlighting physical limits. Subsequently, the current status of WFS technologies is summarized, focusing on accuracy, robustness, cost, and real-time constraints. Finally, a comprehensive framework for utilizing wheel force data is presented, illustrating how WFS can (A) empower model-based control, (B) enable advanced data-driven control strategies, and (C) facilitate data applications for offline design. The review identifies unresolved challenges and explores future research direction, emphasizing the potential of treating WFS as a data source for data-driven control frameworks. This paper provides researchers and engineers with an integrated perspective on the convergence of WFS and yaw stability control, offering insights into advancing the safety and reliability of next-generation DDEVs.
Advancements in haptic teleoperation enable robots to perform complex tasks with precision. Key progress lies in compliance control strategies that adjust robot stiffness based on task requirements. Despite notable achievements, challenges remain, including the need for robust algorithms capable of learning from sparse data, managing uncertainties, and ensuring stability during dynamic interactions. Furthermore, integrating compliance control with teleoperation systems introduces issues such as real-time feedback, latency, and the fidelity of haptic sensations. These challenges are covered in depth through our contributions in this review. This paper provides an in-depth analysis of the current state of learning variable compliance control in haptic teleoperation with limited demonstrations, outlining key challenges, research gaps, and future research directions. The initial section offers background information, followed by learning from demonstrations and methodologies for various compliance control techniques. The review covers recent research, including methodologies, algorithms, simulations, data generation processes, findings, and research gaps. This critical review of methods and experimental analysis is conducted on leading approaches, recent applications, advancements, particularly challenges, limitations, and future directions. We also conducted an analytical evaluation of different learning methods using robot task based simulation data derived from demonstrated tasks, training models show that imitation learning methods, particularly Imitation Behaviour Cloning (IBC) and LSTM models, achieved the highest accuracies (99.84%), outperforming reinforcement learning models that demonstrated lower accuracy and greater sensitivity to hyperparameter tuning. We also provided the recommendations of optimal approaches are provided for learning compliance control in haptic teleoperation.
This paper comprehensively reviews the modeling and control of automotive propulsion systems, with a particular focus on the integration of emerging artificial intelligence (AI) techniques. The highlights of this review paper include: (1) This review specially focuses on AI-empowered modeling and control for automotive propulsion systems, filling the blank in existing surveys; (2) Different approaches to leverage AI in propulsion applications are summarized, categorized and analyzed. Prospects for possible future directions are discussed based on state-of-the-art researches; and (3) Case studies and examples are given to demonstrate the effectiveness of AI techniques applied to automotive propulsion systems.The review starts with the introduction of the fundamental principles of propulsion systems for Internal Combustion engine Vehicles (ICVs), Hybrid Electric Vehicles (HEVs) and Battery Electric Vehicles (BEVs), along with discussions of traditional modeling developments and their advantages and disadvantages. Motivations and prospects of leveraging AI techniques for automotive propulsion systems are highlighted. Subsequently, this paper summarizes the modeling approaches into white-, gray- and black-box methods and discusses their respective strengths and limitations. More importantly, this paper defines the AI approaches, outlines its typical methodologies, and highlights its peculiar significance in automotive propulsion systems, where challenges in modeling and control are identified, and solutions provided by emerging AI techniques are discussed. Particularly, various AI applications in propulsion systems are addressed, including direct and indirect data-driven modeling and control techniques such as Gaussian process-based models, neural networks etc. The effectiveness of these techniques are demonstrated through several representative examples. In the sequel, the unique challenges and limitations of AI applications to propulsion systems are discussed, and future research directions are envisioned. Finally, this paper concludes by summarizing the key points and emphasizing the pivotal role of AI in advancing automotive propulsion systems.
Autonomous vehicles (AVs) are poised to redefine future mobility by offering enhanced safety, energy efficiency, and intelligent adaptability. A fundamental component enabling this transformation is trajectory tracking control, which ensures precise path-following despite environmental uncertainties, dynamic road conditions, and sensor noise. This systematic review follows the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) methodology to analyze state-of-the-art trajectory tracking control strategies, categorizing them into traditional, adaptive, and learning-based methods. The study provides a comprehensive assessment of trajectory tracking models, highlighting their strengths, limitations, and applicability in real-world scenarios. Additionally, the review discusses key challenges, such as scalability, real-time adaptability, and the integration of multi-sensor data. By bridging theoretical advancements with practical implementations, this review contributes to the development of more robust, adaptive, and efficient trajectory tracking systems for autonomous mobility.
Control theory has established itself as a fundamental discipline for the analysis and design of dynamical systems. From its classical foundations, including proportional–integral–derivative (PID) control, state–space representations, and stability analysis, it has progressively expanded toward advanced, robust, and predictive control frameworks. The increasing complexity of modern systems, characterized by large-scale integration, data-driven operations, and stringent safety requirements, is reshaping their methodological foundations. However, despite extensive progress, the field of control remains fragmented, and an integrative viewpoint connecting the different topics is still lacking. This survey provides a unified cross-domain perspective that consolidates established principles and emerging intelligent paradigms. Specifically, it addresses this research gap by synthesizing classical control theory with contemporary artificial intelligence (AI)-driven and cyber–physical systems (CPS) methodologies within a mathematically consistent framework. To the best of our knowledge, no prior survey has provided a unified analytical framework that jointly treats all the aforementioned topics. A distinctive contribution is the mathematically rigorous treatment of nonlinear observer design through Laguerre polynomial approximations, positioned in direct comparison with extended Kalman filters (EKF), high-gain observers (HGO), and hybrid estimation methods. Beyond the classical scope, this survey addresses the convergence of control with AI, machine learning (ML), deep learning (DL), Internet of Things (IoT) infrastructures, digital twins, and quantum-edge computing, emphasizing their implications for scalability, adaptability, and resilience. The novelty of this review lies in articulating an integrative framework that bridges robust analytical tools with intelligent, data-driven architectures, highlighting both methodological coherence and cross-domain applicability. Key challenges include managing nonlinear complexity, ensuring robustness under uncertainty, embedding ethical and governance-aware mechanisms, and bridging the gap between theoretical innovation and practical deployment. Future trends indicate a shift towards reinforcement learning (RL)-augmented control, hybrid physics–AI architectures, distributed CPS architectures, and sustainability-driven designs. By combining historical depth with forward-looking integration, this survey serves as both a consolidated reference and a roadmap for the next generation of intelligent and resilient control systems.
Formation control of multi-agent systems (MASs) using bearing measurements has attracted considerable attention in recent years. Compared to positions and displacements, bearing measurements contain the least information, which cost-effective sensors in many practical applications can easily measure. However, a critical challenge arises from the conflict between the achieved formation maneuvers and the incomplete observability of bearings, which lack distance information. Past surveys have primarily focused on methods based on static formation shapes, such as bearing rigidity, but lack a systematic overview of approaches through persistence of excitation (PE) variations in bearings over time. Thus, in this paper, we present a comprehensive survey and introduce a two-level taxonomy for bearing-based formation control methods of MASs. The taxonomy reveals that to achieve shape-invariant maneuvers or both shape-invariant and shape-varying maneuvers, the control methods need to steer the formation frameworks to satisfy rigidity (including bearing rigidity and angle rigidity) and PE (including agent-to-center PE, neighboring-pair PE, and entire-MAS PE) conditions, respectively. This guarantees the observability of the agents’ positions and hence the convergence of the formation tracking errors. We also review approaches that combine bearings with other measurements, organized by increasing flexibility of the achieved maneuvers. The survey also includes an overview of experimental platforms for bearing observation and discusses future challenges and directions in this field.
Feedforward control (FFC), often paired with feedback, is widely applied in industry to improve accuracy under transient conditions and reject disturbances. Despite this, academic attention has been limited, with few comprehensive reviews connecting algorithms to practical applications. This work bridges that gap by surveying classical and advanced FFC strategies, their advantages and drawbacks, and their implementation across diverse sectors.Classical FFC improves tracking and disturbance rejection through methods such as inverse modelling and input shaping, but faces challenges from instability, intensive tuning, and model inaccuracies. Hybrid techniques, including model predictive and active disturbance rejection control, broaden capability yet remain constrained by complexity and sensitivity to nonlinearities.Recent advances introduce look-ahead, adaptive, optimization-based, and data-driven methods. Preview and predictive designs enhance responsiveness but depend on accurate future estimation, while iterative and intelligent approaches reduce modelling requirements at the expense of stability and training demands. Adaptive and optimization-based controllers strengthen robustness but add computational burden and parameter sensitivity.FFC has been deployed across process control, electrical drives, fuel cells, engines, robotics, motion systems, power electronics, and energy systems. These applications consistently show improved tracking and robustness, though adoption is limited by reliance on models, calibration effort, and a lack of hardware validation.Signal acquisition and processing shape the stability and robustness of feedforward-augmented control architectures. A discussion is had reviewing recent advances in noise-aware estimation, delay-robust control, and filtered/learned inverse models to identify practical design strategies for maintaining reliable performance, focusing on medium and high frequency.To demonstrate practical benefits, a feedforward-augmented PID (i.e., feedforward control added to a PID closed loop system) with preview and anti-windup enabled transient testing on absorbing dynamometers, traditionally restricted to steady-state use. Combined with reinforcement learning controllers, this approach reduced error, expanded applicability, and offered a cost-effective alternative to motored dynamometers.
This article surveys classical, machine learning, and data-driven system identification approaches to learn control-relevant and physics-informed models of dynamical systems. In recent years, machine learning approaches have enabled system identification from noisy, high-dimensional, and complex data. However, their utility in control applications is limited by their ability to provide provable guarantees on control-relevant properties. Meanwhile, traditional control theory has identified several properties of physical systems that are useful in analysis and control synthesis, such as dissipativity, monotonicity, energy conservation, and symmetry-preserving structures. In this paper, we postulate that merging system identification algorithms with such control-relevant or physics-informed properties can provide useful inductive bias, enhance explainability, enable control synthesis with provable guarantees, and improve sample complexity. We formulate system identification as an optimization problem where control-relevant properties can be enforced in three ways, namely, direct parameterization (constraining the model structure to satisfy a desired property by construction), soft constraints (encouraging control-relevant properties through regularization or penalty terms), and hard constraints (imposing control-relevant properties as constraints in the optimization problem). Through this lens, we survey methods to learn physics-informed and control-relevant models spanning classical linear and nonlinear system identification techniques, machine learning-based approaches, as well as direct identification through data-driven and behavioral representations. Taken together, these perspectives suggest that control-oriented identification should be viewed not only as a problem of minimizing prediction error, but also as one of selecting model structures and learning methods that preserve the properties needed for downstream analysis and control synthesis. Throughout the paper, we provide several expository examples that are accompanied by code and brief tutorials on a public Github repository. We also describe several challenging directions for future research in this area, including identification in networked, switched, and time-varying systems, experiment design, and bridging the gaps between data-driven, learning-based, and control-oriented system identification.
This paper presents a tutorial and survey on Probabilistic Inference-based Model Predictive Control (PI-MPC). PI-MPC reformulates finite-horizon optimal control as inference over an optimal control distribution expressed as a Boltzmann distribution weighted by a control prior, and generates actions through variational inference. In the tutorial part, we derive this formulation and explain action generation via variational inference, highlighting Model Predictive Path Integral (MPPI) control as a representative algorithm with a closed-form sampling update. In the survey part, we organize existing PI-MPC research around key design dimensions, including prior design, multi-modality, constraint handling, scalability, hardware acceleration, and theoretical analysis. This paper provides a unified conceptual perspective on PI-MPC and a practical entry point for researchers and practitioners in robotics and other control applications.
The partitioning problem is of central relevance for designing and implementing non-centralized Model Predictive Control (MPC) strategies for large-scale systems. These control approaches include decentralized MPC, distributed MPC, hierarchical MPC, and coalitional MPC. Partitioning a system for the application of non-centralized MPC consists of finding the best definition of the subsystems, and their allocation into groups for the definition of local controllers, to maximize the relevant performance indicators. The present survey proposes a novel systematization of the partitioning approaches in the literature in five main classes: optimization-based, algorithmic, community-detection-based, game-theoretic-oriented, and heuristic approaches. A unified graph-theoretical formalism, a mathematical re-formulation of the problem in terms of mixed-integer programming, the novel concepts of predictive partitioning and multi-topological representations, and a methodological formulation of quality metrics are developed to support the classification and further developments of the field. We analyze the different classes of partitioning techniques, and we present an overview of their strengths and limitations, which include a technical discussion about the different approaches. Representative case studies are discussed to illustrate the application of partitioning techniques for non-centralized MPC in various sectors, including power systems, water networks, wind farms, chemical processes, transportation systems, communication networks, industrial automation, smart buildings, and cyber–physical systems. An outlook of future challenges completes the survey.
Controlling nonlinear dynamical systems remains a central challenge in a wide range of applications, particularly when accurate first-principle models are unavailable. Data-driven approaches offer a promising alternative by designing controllers directly from observed trajectories. A wide range of data-driven methods relies on the Koopman-operator framework that enables linear representations of nonlinear dynamics via lifting into higher-dimensional observable spaces. Finite-dimensional approximations, such as extended dynamic mode decomposition (EDMD) and its controlled variants, make prediction and feedback control tractable but introduce approximation errors that must be accounted for to provide rigorous closed-loop guarantees. This survey provides a systematic overview of Koopman-based control, emphasizing the connection between data-driven surrogate models, approximation errors, controller design, and closed-loop guarantees. We review theoretical foundations, error bounds, and both linear and bilinear EDMD-based control schemes, highlighting robust strategies that ensure stability and performance. Finally, we discuss open challenges and future directions at the interface of operator theory, approximation theory, and nonlinear control.
The fields of MPC and RL consider two successful control techniques for Markov decision processes. Both approaches are derived from similar fundamental principles, and both are widely used in practical applications, including robotics, process control, energy systems, and autonomous driving. Despite their similarities, MPC and RL follow distinct paradigms that emerged from diverse communities and different requirements. Various technical discrepancies, particularly the role of an environment model as part of the algorithm, lead to methodologies with nearly complementary advantages. Due to their orthogonal benefits, research interest in combination methods has recently increased significantly, leading to a large and growing set of complex ideas leveraging MPC and RL. This work illuminates the differences, similarities, and fundamentals that allow for different combination algorithms and categorizes existing work accordingly. Particularly, we focus on the versatile actor-critic RL approach as a basis for our categorization and examine how the online optimization approach of MPC can be used to improve the overall closed-loop performance of a policy.
This tutorial paper presents the design of backstepping-based boundary state-feedback controllers, observers, and boundary output-feedback controllers for a class of one-dimensional parabolic equations with space–time-varying reaction coefficients, employing time-invariant closed-form kernels expressed through Bessel functions. Based on the application of the generalized Lyapunov method, this paper also presents how to perform the stability analysis of the involved closed-loop systems and the state estimation error system under the framework of input-to-state stability theory in the presence of Dirichlet boundary disturbances. Numerical simulations are conducted to confirm the results of theoretical analysis and to illustrate the effectiveness of the proposed control scheme.
Traffic simulators are essential for testing autonomous driving algorithms, and they require driver models that accurately emulate human behavior to reflect real traffic conditions. Our study focuses on developing human driver models to be used in these simulators. We address the limitations of fixed driver models, which do not adapt to new information, by introducing an attention-based learning mechanism inspired by human memory. This mechanism is integrated into a multi-type car following model we developed. Unlike existing car following models, our approach allows the ego driver’s decisions to be influenced, without any bias, by both the vehicle in front and the vehicle behind them.We demonstrate the predictive capabilities of the proposed model using real traffic data and provide a comprehensive statistical analysis of the model parameter distributions. This analysis shows how the model captures general behavioral tendencies across different data sets, enhancing the understanding of interactions between human drivers and providing more realistic simulations for testing purposes. Finally, we offer a step-by-step guide for implementing the model in the development of high-fidelity traffic simulators.
