ABSTRACT With the presence of external disturbance and moving targets, how to establish new predefined time stability criteria to achieve prescribed performance under varying sea states is a challenging problem. A new predefined‐time stability lemma is introduced through employing an exponentially decaying function. It has more adjustable parameters for the size of residual set, and can enter a smaller residual set. Subsequently, according to this lemma, a practical predefined‐time prescribed performance control for cooperative target enclosing of unmanned surface vehicles is proposed. A predefined‐time observer is designed to compensate for uncertain parameters and disturbances. Finally, experimental results validate that this control scheme reduces tracking errors when compared to existing methods, and demonstrates robustness against external disturbances.
ABSTRACT This paper addresses the adaptive tracking control problem for a class of uncertain nonlinear systems. Conventional approximation‐based control methods usually employ fuzzy logic systems (FLSs) or neural networks to directly approximate unknown nonlinear functions. However, the resulting approximation error is highly dependent on the selection of basis‐function parameters, which are often empirically chosen and difficult to characterize precisely. This limitation may degrade the approximation capability and consequently affect the tracking performance. To overcome this issue, the states of the original system are augmented, and a fuzzy state observer is developed to construct an indirect fuzzy approximator for the unknown nonlinear functions. In the proposed framework, the approximation error is equivalently represented by the fuzzy observation error, which can be explicitly regulated through the observer gains. Based on the proposed indirect fuzzy approximator and an adaptive design strategy, an indirect‐approximator‐based adaptive tracking controller is developed. Compared with traditional direct approximation‐based methods, the proposed controller can achieve improved tracking performance by driving the tracking error into a smaller residual set. The stability of the closed‐loop system and the convergence property of the tracking error are rigorously analysed. Finally, simulation and experimental results are provided to demonstrate the effectiveness and superiority of the proposed method.
ABSTRACT Thunderstorm gale forecasting can be formulated as a competitive multi‐agent spatio‐temporal decision problem, where meteorological stations, radar grid cells and convective regions interact under nonlinear, time‐varying and uncertain atmospheric conditions. To address this problem, this paper proposes a data‐driven competitive multi‐agent forecasting and fuzzy decision‐control framework for thunderstorm gale warning. First, a dynamic spatio‐temporal graph convolutional forecasting model, termed DSTGFP, is developed for multi‐station wind speed prediction. By integrating dynamic time warping, mutual information and geographical proximity, DSTGFP constructs an adaptive dynamic adjacency matrix to identify competition–cooperation interactions among station agents. Multi‐head attention and multi‐scale time–frequency feature extraction further allocate representation weights across heterogeneous spatio‐temporal patterns. Second, a multi‐source spatio‐temporal attention network, termed MSTA‐UNet, is constructed for radar echo extrapolation by fusing radar reflectivity, atmospheric dynamic variables and static terrain constraints. Finally, Kriging interpolation and fuzzy logic are introduced to construct a joint thunderstorm gale potential index, where wind‐speed and radar‐reflectivity risks are treated as soft competitive decision constraints. Experimental results show that DSTGFP achieves the best wind‐speed forecasting performance. The overall framework achieves a CSI of 0.5641 and a POD of 0.6471, demonstrating its effectiveness in modelling, analysing and controlling competitive multi‐agent interactions for severe convective weather forecasting, together with a FAR of 0.1852
This paper presents a combination of communication, estimation and control techniques for path-following of a radio-controlled (RC) car. The proposed approach integrates: i) a periodic adaptive event-triggered communication (PAETC) mechanism that regulates sensor-to-controller updates according to vehicle dynamics, ii) a non-uniform extended Kalman filter (NUEKF) for state estimation from an irregular measurement signal and iii) a pure pursuit controller (PPC) for path-following with straightforward control action generation. Unlike traditional time-triggered communication, and periodic (static) event-triggered strategies, the PAETC mechanism further reduces resource usage by transmitting measurements only when some dynamic thresholds are exceeded. Interestingly, dual-rate sampling is also integrated to address some sensor limitations that lead to a slower working period, resulting in a further decrease in the number of transmissions. The NUEKF is able to provide the PPC with the estimates required to generate faster control actions and achieve satisfactory control performance. To accommodate a slower communication rate, the control signal is sent using a packet-based control mechanism. A complex, realistic model of the RC car is implemented in Simscape Multibody. The model has been experimentally validated using a high-fidelity platform such as Optitrack. The comprehensive approach leads to an excellent trade-off between resource savings and path-following behaviour, outperforming traditional time-triggered and periodic event-triggered mechanisms.
The low inertia, nonlinear dynamics, and high penetration of renewable energy sources (RES) in islanded microgrids (MGs), combined with continuous load variations and inherent parameter uncertainties, pose significant challenges to maintaining frequency stability. Conventional proportional-integral-derivative (PID) controllers and single-stage intelligent control schemes exhibit limited robustness against renewable intermittency, rapid operating condition changes, and modelling inaccuracies, thereby necessitating advanced adaptive load-frequency control (LFC) strategies. This paper proposes a novel hybrid grey wolf-ant colony optimization-tuned interval type-II fuzzy logic multi-stage PID (hGWO-ACO-T2F-MPID) controller for robust frequency regulation in isolated MG integrating photovoltaic (PV) systems, wind turbine generators (WTG), diesel energy generators (DEG), and plug-in hybrid electric vehicles (PHEV). The proposed framework synergistically combines a MPID architecture to enhance transient response and steady-state accuracy, an interval T2F inference mechanism to effectively manage nonlinearities, measurement noise, and parametric uncertainties through footprint-of-uncertainty modelling, and a hybrid hGWO-ACO metaheuristic algorithm that integrates the global exploration capability of grey wolf optimization (GWO) with the strong local exploitation ability of ACO to ensure rapid convergence and avoidance of local optima during controller tuning. Comprehensive MATLAB/Simulink simulations are conducted under diverse and realistic operating scenarios, including step load disturbances, solar and wind power fluctuations, simultaneous multi-source disturbances, and severe parametric variations representing component aging and system uncertainties. Comparative analysis against conventional PID, fractional-order PID (FOPID), MPID, and type-1 fuzzy controllers optimized using standalone metaheuristic techniques demonstrates the superior dynamic performance and robustness of the proposed approach. Under step load disturbance, the integral time-weighted absolute error (ITAE) is reduced from 3.758 & times; 10- 4 to 1.197 & times; 10- 4, corresponding to a 68.2% improvement over the hGWO-ACO-MPID controller; during parameter uncertainty tests, the maximum frequency deviation decreases from 0.21-0.07 Hz (66.7% reduction), while the settling time shortens from 6.3-3.8 s (39.7% improvement). Furthermore, under combined wind, solar, and load disturbances, the proposed controller achieves the lowest oscillation amplitude, fastest recovery, and highest damping performance among all benchmarked methods. These findings confirm that the integration of interval type-II fuzzy inference, MPID control, and hybrid hGWO-ACO optimization establishes a computationally efficient, highly robust, and adaptive frequency regulation framework suitable for renewable-dominated isolated MG operating under severe uncertainty and dynamic disturbances.
This paper explores the Polyak-& Lstrok;ojasiewicz inequality and presents an associated convergence analysis of the gradient method with the backtracking line-search in the iterative feedback tuning (IFT). In the conventional IFT, because stepsize design requires the information of a target unknown model, theoretical convergence guarantee is unavailable in practical applications, and the stepsize design has been developed in terms of numerical experiments in existing studies. Even if the aforementioned stepsize, depending on a part of the model information, can be ideally available, convergence rate has hardly been studied. Under some assumptions, this paper derives a P & Lstrok; inequality, and the linear convergence is established by exploiting the gradient descent condition using the backtracking line-search, which leads to an automated stepsize search mechanism with the theoretical convergence guarantee.
This paper presents a trajectory generation method that unifies and effectively links key ideas from input shaping and optimal control for vibration-sensitive motion systems. The method constructs trajectories directly from time-optimal motion primitives, enabling fast transition times while maintaining insensitivity to parameter uncertainty. By explicitly incorporating system damping and stiffness into the motion-primitive design, the resulting trajectories account for oscillatory dynamics, thereby reducing residual vibrations compared to approaches that neglect these effects. The approach avoids non-linear optimization and instead relies on efficiently computed motion primitives, allowing real-time execution on industrial hardware (PLC). An extensive measurement study on a laboratory system, together with comparisons to established and recently proposed methods, demonstrates the advantages of the approach in terms of transition time, residual oscillations, and parameter sensitivity. The results highlight the practical relevance of combining dynamic awareness, structural simplicity, and computational efficiency within a unified trajectory planning scheme.
Modified active disturbance rejection control (MADRC) is commonly designed for high-order inertial processes. This paper focuses on the single parameter tuning rule of the system with MADRC based on the desired maximum sensitivity. It is observed that an asymptote exists in the Nyquist curve of the open-loop transfer function of the high-order system with MADRC. Based on this asymptote and the maximum sensitivity constraint, the mathematical relationship between the maximum sensitivity and different parameters is derived, and consequently a novel single parameter tuning rule is established. The core advantage is that parameter tuning can be completed by adjusting only one desired maximum sensitivity parameter, while allowing the actual maximum sensitivity to closely match the set value. Comparisons with other control methods verify the effectiveness of MADRC under the proposed tuning rule. Finally, the field application of MADRC to the superheated steam temperature system further validates its advantages. Compared with conventional PI-PI control, its overall performance is improved by 15%-30%. This indicates its broad application potential in other large inertial processes in industry.
In various industrial domains, such as intelligent transportation systems, maintaining accurate tracking performance in networked control systems (NCS) operating over fading wireless channels remains a critical challenge. The fundamental difficulty arises from the stochastic and temporally correlated nature of wireless links, which limits the effectiveness of existing control strategies. To address this problem, we introduce a transition-aware Q-learning (TA-QL) framework that enables model-free robust tracking control of NCS subject to fading-induced uncertainties. The proposed approach learns optimal policies directly from networked data while preserving the Markovian dependencies among network states, without requiring explicit models or solving coupled algebraic Riccati equations. We rigorously prove that the learned policies ensure mean-square stability and satisfy the disturbance attenuation criterion. Extensive simulations on a leader-follower vehicle spacing control scenario over realistic 5G fading channels validate the effectiveness of TA-QL. Compared with baseline schemes, including subsystem transformation-based approach, TA-QL improves convergence rate and data efficiency by approximately 54%. Overall, the proposed method bridges robust control and data-driven learning, offering a practical solution for industrial applications of NCS.
In recent years, autonomous underwater vehicles (AUVs) have played an important role in underwater exploration. Researchers are investigating the coordination of multiple AUVs to enhance the efficiency of collaborative tasks. The formation control problem represents a central challenge in this domain. However, current solutions couple the AUVs' formation characteristics and individual control tasks, significantly complicating the controller design while reducing flexibility and practicality. We propose a layered framework for formation control of multiple underactuated AUVs. This framework decouples AUVs' formation requirements and individual control challenges into two distinct layers: a swarm coordination (SC) layer and a precise trajectory tracking (PTT) layer. Specifically, the first layer acts as virtual AUVs, and a consensus-based controller is proposed to generate desired trajectories for AUVs that meet the requirements of complex formation tasks. In the second layer, adaptive controllers are designed based on the kinematic and dynamic models of underactuated AUVs to track the desired trajectories. To increase the convergence speed, both layers can converge in fixed-time under designed controllers. The effectiveness of the proposed layered framework is validated through simulations involving six AUVs. The results show that the novel layered framework makes multi-AUVs complete formation tasks well and significantly improves the flexibility and practicality of the systems.
The aim of this study is to examine the disturbance rejection and stabilization problem for a class of periodic piecewise polynomial time-varying systems with time-delays, uncertainties and multiple disturbances. Specifically, the matching disturbances induced by the external systems are initially addressed by building an observer of disturbances. Meanwhile, the mixed and passivity performance handles the mismatched portion. The control law is created by combining the disturbance observer's output with a composite control rule. Precisely, periodic piecewise time-varying systems are formed by dividing the fundamental period of periodic systems into a limited number of subintervals. Also, by combining the periodic piecewise Lyapunov-Krasovskii functional with a matrix polynomial lemma, a series of delay-dependent adequate conditions for linear matrix inequalities is obtained. These developed conditions confirm that the system under consideration is asymptotically stable. According to these conditions, periodic gain matrices are computed for both the controller and observer. The final step in our work is to demonstrate the validity of theoretical results as well as control schemes using a numerical example.
Brushless DC (BLDC) motors are widely used in applications that are highly-efficient, reliable, and compact, such as electric vehicles, robotics, and medical devices. However, the inherent nonlinearities and load sensitivity of BLDC motors require a robust and adaptive control strategy to ensure satisfactory performance under various operating conditions. Sliding mode control (SMC) has been widely used for the BLDC drives. However, because of its simplicity and robustness, the control effectiveness of the control is limited by the sensitivity to the disturbances and the chattering phenomenon. To remedy this, super-twisting (ST) technique has been proposed to achieve smoother response and better robustness to overcome the obvious problem while requiring a large amount of filtering. The incorporation of fractional-order (FO) dynamics adds flexibility, noise resilience and adaptability but FO controllers are still based on offline gain tuning and cannot be used in practice. In order to overcome all these challenges, a novel transfer learning-based fractional-order super-twisting sliding mode controller (TL-FO-ST-SMC) is proposed in this paper. Transfer learning is applied to improve the speed of adaptation on different operating domains by reusing the previous knowledge, thus decreasing the need of repeated offline tuning. Nevertheless, TL itself requires good initial fine-tuning. To this end, antlion optimization (ALO) algorithm is used to find the optimal initial gains which helps in stable learning and robust convergence. A Lyapunov-based stability analysis is given and the proposed method is validated by a typhoon hardware-in-the-loop (HIL) real-time platform. Results show that the TL-FO-ST-SMC has better speed tracking capability, less chattering, disturbance rejection, and faster adaptation than conventional controllers and is suitable for advanced electric vehicle and motor drive applications.
Model predictive control (MPC) is a powerful control strategy that delivers optimal performance while ensuring the satisfaction of safety and operational constraints at all times. However, its reliance on online optimization significantly increases computational demand, limiting its deployment in safety-critical systems with limited onboard computational resources. A promising solution is to replace the MPC optimizer with a neural network trained to approximate the MPC policy. However, the effectiveness of this neural MPC framework relies heavily on both the quality and efficiency of the training process. Supervised learning requires extensive data collection, whereas unsupervised methods often suffer from convergence issues and instability. To address these challenges, this paper introduces a hybrid training method that leverages the strengths of both supervised and unsupervised learning approaches to accelerate training and improve the performance of the resulting neural MPC framework. Extensive numerical experiments show that the proposed training method is up to 147 times faster than conventional approaches, while yielding neural MPC frameworks that improve safety and tracking performance by up to 99.37% and 84.7%, respectively. The practicality and applicability of the hybrid training approach are further validated through its deployment in neural MPC designs for drone hovering and thermal regulation tasks.
Systems containing ambiguous components demonstrate intricate behaviours crucial for designing control systems. Handling twist oscillations and stick-slip fluctuations in a drill-string system presents a significant engineering challenge in oil drilling, due to their detrimental and expensive effects. This control system regulates the rotational speeds of drill-string components to achieve predetermined targets. In this study, we utilise a lumped-parameter model to analyse the twisting dynamics of the drill-string with four degrees of freedom. To alleviate stick-slip oscillations, reduce bit sticking in vertical oil-well drill strings and diminish saturation effects on control performance, a composite nonlinear feedback is used with a supertwist integral sliding mode control technique, including an anti-windup mechanism. By identifying a suitable Lyapunov function, mathematical calculations derive a condition based on actuator saturation areas for exponential stabilisation, expressed as an LMI. Additionally, such oscillatory phenomena are not confined to vertical oil-well drill strings; analogous dynamic instabilities are observed in both offshore and onshore derrick systems, emphasising the broader applicability of the proposed control strategy. The efficacy of the approach is confirmed by simulation results.
We propose a computationally efficient model predictive control (MPC) approach for optimal electrical and thermal energy flows in non-residential buildings, where components such as heat pumps and heating rods introduce discrete variables, necessitating the solution of mixed integer quadratic programming (MIQP) problems. To reduce the computational effort of solving these problems, we relax the discrete dynamics along the horizon and incorporate the resulting error as a bounded additive disturbance. This compensates for suboptimality and potential constraint violations of the relaxed approach in closed-loop operation. We employ multi-stage MPC as a robust MPC framework to handle these disturbances. The recursive feasibility of the proposed multi-stage MPC formulation is established using a robust control invariant terminal set. We evaluate the performance of the proposed multi-stage MPC in simulations for a real office building. The results demonstrate that the proposed computationally efficient robust formulation achieves performance comparable to the optimal solution obtained by solving the MIQP problems for the entire horizon while ensuring constraint satisfaction in closed-loop simulations.
The -winner-take-all (-WTA) operation is a fundamental neural computation that models competitive selection among multiple agents. While existing -WTA networks have been extended to handle noise, unbalanced topologies and time delays, most approaches still suffer from high communication costs and limited ability to guarantee global selection optimality in distributed settings. To address these challenges, this paper develops a distributed -WTA neural selection-control model that integrates nonlinear winner-selection dynamics with communication-efficient protocols. The proposed model ensures the convergence to globally optimal winners under local neighbour communication and provides rigorous stability guarantees. Building upon the selected winners, a cooperative formation control strategy with a double closed-loop structure is designed, enabling fast convergence, robustness against disturbances and reduced communication burden. The effectiveness of the framework is validated through theoretical analysis, numerical simulations under both static and dynamic conditions, and a representative application to multi-UAV cooperative competition and formation control. The results demonstrate that the proposed approach achieves global selection consistency, high-precision tracking performance and strong robustness, thereby highlighting its potential as a general distributed neural framework for cooperation-competition systems.
A new adaptive configuration of fuzzy logic controller (FLCh) based on the hyperbolic function is introduced to improve the performance of control systems. The presented configuration is named adaptive because its input scaling factors (SFs) are dynamically changed while the controller is servicing. This is accomplished by employing two hyperbolic functions with tunable slope and magnitude parameters to act on the input error signal and its derivative nonlinearly. The efficacy of our proposal is explored on a classical PID type FLC (C-PID-FLCh) and the controller parameters are simultaneously tuned using the stochastic fractal search (SFS) algorithm for proper functioning. Extensive simulations are conducted to assess the performance of the presented control scheme. A comparison study is also realised against the existing solutions to prove the true contribution of the work. The results show that thanks to the transient change in input SFs, the C-PID-FLCh exhibits faster yet nonoscillatory behaviour with reference to the reported schemes. The presented configuration does not cause major changes as far as the structure of the classical FLC is concerned, evolving C-PID-FLCh as a potential contender in control systems design.
Automatic voltage regulation (AVR) is an essential component of the stability and reliability of the electrical grid. It is vital that the controller, which has been designed for the purpose, functions in the most optimal manner. This is particularly important with regard to keeping the voltage amplitude within specified limits. In view of the aforementioned factors, the present study proposes a novel controller, designated the adaptive logistic proportional-integral-derivative (A-LogPID) controller, whose purpose is to address the control challenge in AVR systems. In order to ascertain the most effective controller parameters of the A-LogPID, the jellyfish search (JS) optimization methodology is utilized, with the objective being to minimize Zwee-Lee Gaing's (ZLG) time-domain performance function. An evaluation of the proposed A-LogPID controller is conducted through a comparative analysis of extant controller structures in the literature. A comprehensive and detailed set of tests is considered, including time-domain analysis, uncertainty in system parameters, and attack performance of the proposed controllers. In performance tests, two distinct cyberattack models involving false data injection and denial-of-service attacks are used for the first time in an AVR system. The study also introduces a cyberattack detection index for the controllers, complete with a statistical analysis. To support reproducibility and future research, the dataset and simulation models used for cyberattack-based controller evaluation are made available in a public repository. The performance assessments clearly show that the A-LogPID controller has a better capacity for dynamic response and resilience in specified scenarios than alternative PID controller methodologies.
A proportional-integral (PI) controller is still the workhorse in the industry due to its ease of commissioning and reliability. Therefore, this paper introduces an exponential PI (EXP-PI) controller as a potential alternative. In the proposed scheme, two tuneable EXP functions acting nonlinearly on the error and the rate of change of the error are incorporated in cascade with the PI control architecture. This controller is implemented as a speed controller on a permanent magnet DC motor drive system. Five recent intelligent algorithms, namely stochastic fractal search (SFS), snake optimizer (SO), dragonfly search algorithm (DSA), symbiotic organisms search (SOS) and reptile search algorithm (RSA) are employed to identify the best performer for calibrating the controller gains. According to the statistical results, SFS is found to provide controller gains of higher quality, reducing the designed cost function value to 48.19. The superiority of SFS is verified by the nonparametric Wilcoxon rank-sum test. Several experimental results with SFS-calibrated EXP-PI controller and other existing control schemes are presented using the DSP of TMS320F28335. The results show that our proposal performs better than its competing opponents in terms of various performance metrics, including integral-based error criteria, stability margin, overshoot and settling time for plants with and without dead time.
