This study is devoted to establishing the existence of optimal feedback controls for a class of non-autonomous second-order, damped stochastic neutral impulsive evolution systems formulated within Hilbert spaces. Employing the analytical framework of strongly continuous cosine families in conjunction with Schauder’s fixed-point theorem, stochastic analysis techniques, and the theory of evolution operators, we derive a comprehensive set of sufficient conditions ensuring the existence of mild solutions to the considered class of systems. Building upon these foundational results, we further develop a novel set of structural hypotheses-integrating the Cesari property and Filippov’s theorem to secure the existence of admissible control-state pairs corresponding to the associated Lagrange-type optimal control problem. Finally, a constructed example is provided to demonstrate the practical applicability and robustness of the theoretical findings.
This paper describes a unique approach to the design of a functional H_∞ filter for a class of nonlinear descriptor systems subject to disturbances. To create a broad and applicable framework that encompasses a wide range of systems, this work includes nonlinearities and disturbances in both the state and output equations. Deviating from traditional system regularity assumptions, we adopt a more inclusive approach by considering general descriptor systems that satisfy a rank condition on their coefficient matrices. Within this rank condition framework, we establish sufficient conditions for designing the filter, initially formulated as a bilinear matrix inequality (BMI) and subsequently transformed into a linear matrix inequality (LMI). The resulting filter ensures the asymptotic stability of the estimation error and constrains the ℒ_2 gain from disturbances to errors to a predetermined level. Moreover, the filter is realized in a standard state-space form and can be initialized with arbitrary conditions, significantly simplifying implementation compared with prior descriptor-form filters. Numerical examples are provided to demonstrate the effectiveness of the theoretical results obtained.
In this paper, we consider the formation control of multi-agent Lagrangian systems and propose a distributed structured deep neural network (DNN)-based controller using backstepping techniques. The controller is designed using DNNs with specific structures that guarantee unconditional stability for any set of DNN parameters. We explicitly provide an upper bound on the formation error in the presence of disturbances, and this bound can be adjusted by tuning the DNN parameters. Furthermore, an adaptive controller is developed to address scenarios with unknown model information using radial basis function neural networks (RBFNNs). By customizing the parameter update rate of the RBFNNs, we provide the same stability and performance guarantees for the adaptive controller. In addition, the controller’s performance can be enhanced by optimizing the parameters of the DNNs. The effectiveness of the proposed controllers is validated through several simulations on manipulators.
In this work, we propose a new adaptive iterative learning radial basis function (RBF) neural network control for uncertain nonlinear systems where the global Lipschitz continuity condition is required only for the input function. In addition, a priori knowledge of the control direction is not required. For this purpose, we employ RBF neural networks to estimate the unknown nonlinear functions in such a way that the weights are adjusted using a proposed adaptive law. Furthermore, it is well known that the Nussbaum function technique is an appropriate choice to deal with the unknown control direction. In our paper, this approach is not adopted and the unknown input function is adjusted using a new algorithm. A further advantage of the proposed control is that there is no restriction on nonlinearities. Using Lyapunov theory, the stability analysis of the closed-loop learning system is guaranteed. Finally, simulation results on perturbed nonlinear system are provided to illustrate the effectiveness of the proposed method.
This paper investigates the problem of fault-tolerant fuzzy iterative learning control (ILC) for hybrid-order nonlinear multi-agent systems (MASs) subject to actuator faults. The considered MASs comprise a combination of first- and second-order agents, incorporating nonlinear unmodeled dynamics as well as actuator bias and gain faults. To address these challenges, a novel fault-tolerant fuzzy ILC scheme is proposed. Fuzzy logic systems (FLSs) are employed to approximate unknown nonlinear dynamics, while Nussbaum gain functions and an adaptive iterative learning strategy are utilized to compensate for unknown fuzzy weight parameters and actuator faults. A Lyapunov-like function is designed to rigorously analyze the convergence of tracking errors. Finally, simulation results are provided to validate the effectiveness of the proposed control approach.
An extremum seeking control (ESC) scheme based on switching between adaptive amplitude laws is presented in this paper to enhance both dynamic and steady-state performance. Two novel amplitude adaptation laws are developed using sliding-window dynamic gradient information, enabling more accurate prediction of output trends and ensuring that the excitation amplitude decays to zero upon crossing local extrema to reach the global maximum, thereby eliminating steady-state oscillations. To greatly improve dynamic performance without sacrificing steady-state accuracy, a Lyapunov-based switching strategy is employed to transition between the adaptive laws, allowing the system output to converge rapidly to the global maximum without oscillations. A rigorous stability proof is provided, and numerical simulations are conducted to demonstrate the effectiveness of the proposed scheme. The method is also applied to an anti-lock braking system (ABS) to verify its practical applicability.
This paper investigates the adaptive hierarchical tracking control problem for heterogeneous vehicles platoon subject to model uncertainties and input delay. A hierarchical control architecture is constructed, with separate controllers designed for the leader and followers to enhance the robustness of the system. Based on this framework, a modified constant time headway policy is proposed to address excessive transient response caused by nonzero initial spacing between vehicles. Specifically, by introducing a compensation signal, this policy eliminates the zero-velocity assumption and initial spacing issues while reducing inter-vehicle spacing, ensuring rapid convergence to a stable platoon even in the presence of initial errors. Moreover, an extended state observer (ESO) is employed to approximate and compensate the uncertainties for the system, and Pade approximation approach is utilized to avoid input delay. Lyapunov stability analysis demonstrates the internal stability and string stability of the closed-loop system. Simulation examples and comparative results validate the reliability of the control scheme.
An adaptive output feedback tracking control problem is investigated based on an improved reduced-order observer for nonlinear multiagent systems under event-triggered mechanism. Firstly, an improved reduced-order observer, which only depends on the relative output information of each follower agent, is proposed to address the problem of unmeasurable states. Different from other observers, it can reduce the computational burden while fully utilizing distributed information. Next, a dynamic event-triggered mechanism is constructed that considers tracking errors in the threshold design. Its advantage is that the trigger frequency can be adjusted according to tracking errors of the system. By using Lyapunov stability theory, the signal of the whole closed-loop system is guaranteed to be stable, and the output of all followers finally tracks the reference signal synchronously. In addition, Zeno behavior is also avoided. Finally, the viability of the designed control method is confirmed by a simulation example.
This paper investigates the cooperative circumnavigation problem for multi-agent systems in various environments. Unlike existing studies that primarily focus on continuous-time systems in two-dimensional (2D) obstacle-free scenarios, this work addresses a more general case in three-dimensional (3D) environments within a discrete-time framework, considering that the physical agent and its digital controller inherently constitute a sampled-data system in practice. First, a convex model is introduced to enclose and represent real-world obstacles. Subsequently, a distributed discrete-time controller is developed, and sufficient conditions for system stability are derived via Lyapunov analysis in the discrete-time domain. These constraints theoretically define the feasible range of controller parameters, thereby establishing a foundation for their selection. Afterward, an obstacle avoidance strategy with a dynamic weighting mechanism is developed for the agents, enabling proactive collision avoidance in 3D environments. Consequently, the multi-agent system safely avoids obstacles while performing circumnavigation tasks following predefined configurations. Finally, numerical experiments are conducted to validate the effectiveness of the proposed method and demonstrate its performance.
This paper addresses the distributed generalized Nash equilibrium (GNE) problem of multi-cluster games with coupling constraints, where only partial information is available. In this multi-cluster game, agents within each cluster cooperate to minimize the cluster’s objective function while ensuring that the strategies satisfy the coupling constraints. In our setup, each agent can only communicate with its neighboring agents via an undirected connected graph. To solve this GNE problem, we propose a distributed algorithm, which aims to achieve a variational GNE by appropriately selecting fixed step sizes. In the proposed algorithm, agents directly estimate the consensus strategy of each cluster without estimating the strategies of all other agents. It also eliminates the requirement for a centralized coordinator to collect and broadcast the total gradient of the cluster. With the help of operator theory, the proposed algorithm is expressed as a forward–backward iterative form represented by a pair of preconditioned operators, and its convergence is rigorously analyzed. Finally, the effectiveness and practical applicability of the designed algorithm are verified through an example in an energy system.
In unknown road environments, the path chosen by autonomous vehicles significantly impacts suspension vibrations, which directly affects ride smoothness. Active suspension control systems improve vehicle maneuverability and extend the feasible driving range. However, current methods for path planning and suspension control heavily rely on offline models or historical data, limiting adaptability to real-time conditions. To mitigate these limitations, this study utilizes local road surface data captured by onboard optical sensors. These data provide insights into variations in Markov transition probabilities between different environmental states, termed “spatial side information” in this context. By integrating this spatial side information into existing model-based reinforcement learning algorithms, a novel rapid online learning algorithm is developed. This approach enables vehicles to effectively learn and adapt without the need for direct traversal of every state. Furthermore, this research designs optimal control strategies tailored to unfamiliar environments, striking a balance between exploration and exploitation. Safety measures are also integrated to ensure secure exploration by autonomous vehicles. Experimental validation using real-world road data in conjunction with the active suspension test platform confirms the algorithm’s robust performance in learning and control, demonstrating its capability for safe navigation and exploration.
Internal self-collision between the bucket connecting barrel and the arm poses serious risks in mining excavators, especially under high-load and high-frequency operations. Traditional control methods cannot simultaneously guarantee safety, adaptability, and real-time performance in such strongly coupled nonlinear systems. To address these limitations, this paper proposes a hierarchical event-triggered reinforcement learning model predictive control (RL-MPC) framework that achieves structure-level adaptive coordination between learning and prediction. A neural network dynamic model embedding actuator constraints and physical priors is developed to accurately predict future motion states. Based on this model, the event-triggered MPC layer performs safety-constrained optimization only when collision risks occur, forming a selective activation mechanism that enhances computational efficiency while maintaining safety margins. On this basis, the reinforcement learning layer adaptively adjusts the MPC structural parameters including weighting matrices and horizon lengths, according to the system’s operating condition, enabling online adaptation to dynamic environments. This integrated design builds a closed-loop hierarchical optimization mechanism, where reinforcement learning provides adaptive parameter tuning and MPC ensures constraint satisfaction and stability under event-triggered coordination. Simulation studies under representative risk scenarios demonstrate that the proposed framework achieves superior safety assurance, tracking accuracy, and computational efficiency compared with traditional proportional-integral-derivative and fixed-parameter MPC approaches. The results establish a new paradigm for safe adaptive control in heavy machinery by unifying learning-based adaptability and model-based predictive safety control within a single architecture.
Real-world data tend to follow a long-tailed distribution. Deep models trained on such datasets often exhibit bias toward head classes and perform poorly on tail classes. Existing methods, such as reweighting, have significantly improved the performance of the long-tailed recognition. However, they fail to address the problem of the great uncertainty in model prediction, which results in the weak generalization ability of deep networks. To tackle the issue, we propose a novel approach called consistency self-distillation with augmented mixture (CSAM), which consists of two core components: augmented mixture (AM) and consistency self-distillation (CSD). AM generates two different types of augmented samples to enrich the data, thus improving generalization and reducing the uncertainty in model predictions. Specifically, it employs a weak–strong augmentation strategy and a global–local mixture method to generate weakly augmented global and strongly augmented local samples. CSD further reduces the prediction uncertainty by distilling knowledge from predictions of the weakly augmented global data to regularize strongly augmented local images. Moreover, we propose a hybrid rebalancing strategy that integrates resampling and logit adjustment methods to handle the worse tail class compression problem in our method. Our approach achieves outstanding performance on CIFAR10-LT, CIFAR100-LT, ImageNet-LT, and Places-LT, demonstrating the effectiveness and superiority of our proposed CSAM.
Developing a high-performance elevator group control system (EGCS) remains a challenging problem in the elevator industry. Aiming to achieve optimal transportation efficiency and reduced passenger waiting time, this paper proposes an optimized scheduling strategy for elevator groups that integrates evolutionary computation with deep neural networks (DNNs). First, the overall dispatching time (ODT) is introduced as a method to determine the optimal scheduling strategy for destination-oriented elevators with reservation calls. The ODT is defined as the total time required for an elevator to service a passenger call, from its current position to the call floor and then to the destination floor. Second, a DNN is designed and trained to approximate the desired optimal ODT-based cost function, thus serving as a predictive scheduling model. Furthermore, evolutionary computation techniques—including particle swarm optimization, genetic algorithm, and covariance matrix adaptation evolution strategy—are employed to optimize the DNN parameters, leading to improved scheduling decisions that reduce passenger waiting time. Finally, experimental results verify that the optimized DNN consistently outperforms both the traditional nearest car method and the ODT method across diverse passenger flow patterns, demonstrating its significant advantage in improving EGCS performance.
Hopf bifurcation and bifurcation control of a generalized delay fractional-order prey–predator–scavenger system (GFPS) are analyzed in this paper. Firstly, the conditions of existence and uniqueness of solutions of GFPS are discussed. Secondly, local stability conditions of the equilibria of the non-delay GFPS are derived by using the stability theory of the fractional-order system. Thirdly, choosing delay as a bifurcation parameter, the stable region and conditions of emergence of Hopf bifurcation at the coexistence equilibrium of the GFPS are deduced. A hybrid control method consisting of a linear delay feedback controller and a fractional-order proportional-derivative controller is introduced successfully to control the Hopf bifurcation of the GFPS. Finally, numerical simulations are implemented to validate the correctness of the theoretical analysis and the effectiveness of hybrid controller. Furthermore, numerical simulations show that the hybrid controller has a better control effect than that of a single controller.
This paper addresses the attitude tracking control problem for a 3-degree-of-freedom (DOF) helicopter subjected to both external and state-dependent internal disturbances. To counteract these disturbances, we propose an output feedback controller based on an extended state observer (ESO) that estimates state-dependent system uncertainties and attitude angle velocities. Additionally, we impose constraints on the control input and system state to enhance the safety of the 3-DOF helicopter system. Experimental results validate the effectiveness of the proposed control approach.
In this paper, we propose a novel control scheme that integrates a disturbance observer with a nonlinear anti-swing control approach for underactuated overhead crane systems. Unlike existing methods, this strategy accounts for double-pendulum dynamics, unknown disturbances, and system constraints. First, a high-gain observer is designed based on the system’s dynamic equations to estimate the unknown disturbances. Then, a Lyapunov-based model predictive controller is developed, using the observer’s information along with a second-order sliding mode controller, which is used for tracking while adhering to system constraints. Through comparative simulations with conventional controllers, the proposed method is validated, demonstrating improved performance and robustness.
This paper proposes the design of an observer-based adaptive integral sliding mode control strategy for a class of nonlinear uncertain switched systems with unmeasurable states. During the controller design process, the information of model uncertainties, as well as the bounds of the nonlinear and external disturbance, is assumed to be unknown. By using the estimated system states, a novel integral-type sliding surface is constructed, and a corresponding adaptive sliding mode controller is developed. The proposed controller not only mitigates the chattering induced by subsystem switching but also guarantees finite-time reachability of the sliding surface. A switching signal, designed based on both time and state, is incorporated to optimize the switching logic. By employing Lyapunov theory and the linear matrix inequality (LMI) technique, sufficient conditions are established to ensure the stability of the overall closed-loop system under the proposed control scheme. Numerical simulations validate the effectiveness and feasibility of the proposed method in suppressing uncertainties and external disturbances.
This paper introduces spatio-temporal collaboration (STC), a novel formalism for coordinating multi-agent systems under signal temporal logic (STL). STC defines critical inter-agent dependencies that may be violated by the cascading delays resulting from temporal relaxation, a common method for resolving local task conflicts. To address this issue, we first analyze the propagation of task delays through dependent tasks. A time interval refinement strategy is then proposed to maintain the required collaborations. This strategy is integrated into a distributed predictive control algorithm, ensuring simultaneous satisfaction of both STL specifications and STC relations while preserving recursive feasibility and closed-loop stability. Validation via a case study demonstrates the effectiveness of proposed strategy in preventing collaboration failures.
This study introduces a hybrid integral recurrent neural network (HIRNN) developed to tackle time-varying nonlinear optimization challenges characterized by multiple constraints. The proposed HIRNN architecture incorporates dynamic constraints directly into the optimization process, enabling effective performance under various types of noise, including constant, linear, and quadratic disturbances. Numerical simulations and physical experiments on a manipulator trajectory tracking task demonstrate that the proposed model achieves rapid convergence, consistent performance in the presence of disturbances, and a notable reduction in residual levels compared to existing approaches. These results highlight the enhanced accuracy and noise robustness of the HIRNN, underscoring its potential as a practical and reliable solution for time-varying constrained optimization in complex engineering applications.
