This paper investigates the reinforcement-learning-based structured optimal control for linear stochastic systems with an unknown state matrix under communication topology constraints. A data-driven structured policy iteration algorithm is developed within the framework of adaptive dynamic programming (ADP) to learn a structured feedback policy directly from system trajectory data. By reconstructing the policy evaluation equation from data and updating the feedback gain matrix through analytical orthogonal projection, the proposed method avoids explicit dependence on the exact system dynamics while preserving the desired sparsity pattern. The convergence of the proposed algorithm is established theoretically. Sufficient conditions for the mean-square stability of the closed-loop system are derived, and suboptimal performance bounds induced by topology constraints are characterized. Simulation results show that, even when specific communication links are removed, the learned structured policy yields a steady-state control cost close to that of the unconstrained optimal controller. These results verify the effectiveness of the proposed reinforcement-learning-based structured control method.
This paper proposes a prescribed-time bipartite consensus control of multi-agent systems subject to limited communication resources. Specifically, a tunnel-type prescribed performance method is constructed using prescribed-time functions to enhance transient performance. Notably, unlike conventional funnel-type approaches, this method achieves tighter error bounds. Moreover, an improved coordinate transformation is further designed to properly characterize the tracking error when the performance functions lie on the same side of the axis. Additionally, a prescribed-time controller is developed to achieve user-specified convergence, wherein a composite learning-based estimation strategy is embedded to handle unknown dynamics, significantly improving estimation efficiency while preserving accuracy. Furthermore, a performance-driven dynamic event-triggered mechanism with an adaptive threshold is developed to reduce communication resources while without compromising control accuracy. Finally, it is proven that the closed-loop system achieves the prescribed-time stability, and simulation results verify the effectiveness of the proposed method.
This paper proposes a fuzzy adaptive control prescribed performance coded event-triggered method for switched nonlinear systems. The objective is to achieve prescribed-time convergence under unknown system dynamics while reducing communication load. A coded event-triggered method is introduced, in which only encoded bit strings are transmitted between the controller and the actuator, thereby significantly lowering communication frequency compared with conventional continuous transmission schemes. To handle arbitrary initial conditions, an adaptive prescribed performance function is designed that dynamically adjusts constraint boundaries and convergence time, overcoming the limitations of traditional prescribed performance control, which relies on predefined performance bounds. Simultaneously, based on fuzzy logic system, the system estimates the uncertainty function within the system. Through Lyapunov stability analysis, it is shown that the system state can converge to the dynamically adjusted performance boundary and remain stable. Simulation results confirm that the proposed strategy achieves precise tracking performance while markedly reducing communication updates.
This paper investigates the distributed cluster-bipartite output consensus problem for singular heterogeneous linear multi-agent systems under cooperative-competitive communication networks. Different from conventional state consensus tracking, a consensus control framework that consists of multiple clusters and multiple leaders modeled as exosystems is introduced to accomplish more complex collaborative tasks. However, when agents exhibit heterogeneous and singular dynamics, the cooperative-competitive interactions both within and between subgroups pose a challenge to the design of distributed control protocols. To address this issue, a new cluster-bipartite output consensus control protocol based on output regulation theory is proposed, which enables agents in different subgroups to respectively achieve bipartite consensus. For bipartite tracking of leaders within their own subgroups, a distributed observer relying on Riccati inequality and a topological matrix is developed. By solving the singular regulation equations, this paper presents both state-feedback and output-feedback cluster-bipartite control protocols. Finally, simulation results validate the effectiveness of the proposed algorithm with the designed controllers.
This paper studies the distributed-optimization problems of general linear multi-agent systems (MASs) with unknown dynamics. The goal is to collaboratively optimize the global objective function composed of the sum of convex objective functions. Firstly, a noise-free data-driven adaptive distributed optimization protocol based on edges is designed for MASs, which is able to achieve asymptotical consensus tracking by adjusting the weight of each edge online, while minimizing the global objective function. Then, an improved data-driven adaptive distributed optimization protocol is established for the scenario with noisy data, which is able to achieve asymptotic consensus by updating weights solely based on state-errors under bounded noise, while minimizing the global objective function. Importantly, compared with existing distributed optimization protocols, the two proposed distributed optimization protocols do not rely on system model knowledge. Finally, two examples are provided in the paper to verify the effectiveness of the proposed noise-free and noisy data-driven adaptive distributed optimization protocols, respectively.
To address the issues that single-sensor Unmanned Vehicles are prone to environmental constraints in navigation and obstacle avoidance tasks, and that reinforcement learning-based navigation and obstacle avoidance models suffer from low training efficiency and weak reward guidance, this paper constructs a multimodal perception framework centered on LiDAR and RGB cameras, introduces a curriculum learning strategy into the reinforcement learning training of unmanned vehicle navigation and obstacle avoidance models, and incorporates the core idea of the artificial potential field method into the reward function design. Experimental results demonstrate that the proposed method yields positive optimization for training stability; the multimodal approach remedies the defect of the frontal perception blind spot existing in 2D LiDAR, and the full sim-to-real workflow deployment of the model is accomplished, which verifies the effectiveness of the proposed method.
Precise and real-time localization of surgical instruments is essential for perception-driven automation in robot-assisted minimally invasive surgery. This work presents a unified monocular 3D localization framework that combines a Transformer-augmented detector (YOLOX-STrans), geometry-based multiview triangulation, and temporal fusion via an Extended Kalman Filter (EKF). Unlike stereo or marker-based methods, our system operates entirely on monocular endoscopic imagery, ensuring robustness under occlusion, motion blur, and illumination degradation. We have included extensive evaluations against state-of-the-art Swin-YOLOX variants and geometry-based localization frameworks. Experimental results demonstrate that the proposed Transformer-Kalman fusion improves spatial accuracy by up to 40% Root Mean Square Error (RMSE) reduction and temporal consistency over recent baselines such as ST-YOLOA, BiPAFPN-YOLOX, and STrans-YOLOX. The proposed framework offers a practical, markerless, and real-time solution for perception-driven surgical robotics.
Spiking Neural Networks (SNNs) offer a biologically grounded and energy-efficient alternative to conventional deep learning, yet have historically struggled to match the representational power of Artificial Neural Networks (ANNs) on complex perceptual tasks. Attention mechanisms present a promising route to closing this gap, but integrating them with spike-based computation raises fundamental challenges spanning representational incompatibility, temporal credit assignment, and energy preservation. This survey provides a systematic analysis of attention mechanisms in SNNs, tracing the full pipeline from neural encoding strategies through algorithmic design to neuromorphic hardware implementation. We categorize spiking attention into four types: rate-based, temporal, spatial–temporal, and spike-based, and analyze how each interacts with different encoding schemes and hardware constraints. Applications spanning computer vision, speech recognition, robotics, and multimodal sensing suggest that attention-enhanced SNNs can narrow the accuracy gap with ANN baselines while retaining the energy advantages of event-driven computation. We further examine the field through the lenses of theory, algorithms, hardware co-design, and community infrastructure, identifying the open challenges that must be resolved for attention-enhanced SNNs to mature into deployable neuromorphic systems, and articulating promising research directions toward that goal.
With the ongoing technological advancements, mobile robots with biomimetic structures have garnered significant attention due to their dynamic performance and adaptability to complex environments. Among these, quadruped robots, characterized by their compact physique and legged morphology, exhibit remarkable agility and are capable of overcoming challenging terrains. This makes them promising candidates for performing complex tasks in unstructured real-world scenarios. Currently, quadruped robots have found preliminary applications in domains such as search and rescue, infrastructure inspection, defense, and space exploration. In this paper, we provide a comprehensive review of quadruped robots, beginning with their developmental progress from hardware and structural evolution to control strategies. We examine motion control approaches based on both optimization and learning, with a particular focus on legged locomotion and mobile manipulation, as well as sim-to-real transfer techniques. Moreover, we discuss the role of vision perception and environmental sensing in enabling robust interaction with complex terrains and objects. Finally, we summarize representative applications, identify key challenges, and highlight future opportunities for advancing quadruped robot research.
This survey presents a comprehensive overview of networked control systems (NCSs), focusing on stability and performance guarantees under network-induced effects and constraints, mainly emphasizing on signal quantization, sampling, transmission delays, data losses and security. We first review fundamental concepts of NCSs, including modeling aspects of the control framework and common network-related assumptions. Subsequently, we examine stability-oriented results and their limitation with respect to performance issues. The survey then shifts to performance-oriented control of NCSs and discusses the trade-offs between achievable performance metrics and network limitations. Finally, we identify open issues and future research directions, including challenges associated with complexity issues, uncertainties on network conditions, practical implementation constraints and computational demands. This structured perspective provides a clear road-map for both understanding current NCS research and guiding future investigations.
In robot-assisted minimally invasive surgery (MIS), precise automatic control of the endoscope is crucial for maintaining optimal visualization. Regarding the additional burden imposed on surgeons due to manual adjustment of the field of vision, as well as the unstable perception caused by environmental noise during the surgical process, this paper proposes a Perception-Enhanced Hybrid Visual Servo (P-HVS) framework. Its core contribution lies in a unique architectural integration of perception and control at the system level, specifically designed to suppress measurement noise and provide a more stable and smoother perception data stream compared to the raw visual feedback during surgery. At the perception layer, the framework combines a deep learning model (YOLOX-STrans) with an Extended Kalman Filter (EKF) to form a processing pipeline that mitigates fluctuations in raw depth measurements from stereo vision, provide a more stable three-dimensional input for the control layer. The control layer then implements a speed-proportional blend of IBVS and PBVS based on this unified state. From an engineering perspective, this integration effectively bridges the gap between image-plane sensitivity and workspace accuracy while avoiding the command interruptions inherent in traditional discrete switching schemes. Quantitative simulation results on the dVRK platform demonstrate that the proposed hybrid servo framework excels in trajectory smoothness, tracking accuracy, and joint velocity stability, laying the foundation for autonomous control in the field of surgical robotics.
In this paper, a dual-channel event-triggered quantized model-free adaptive control method for a three-degree-of-freedom helicopter attitude control system with limited communication resources is presented. First, the formulated dual-event triggering strategy integrates an input-triggered scheme and an output-triggered scheme to achieve semi-independent event-triggered control for the sensor-to-controller and controller-to-actuator channels, thereby further reducing communication frequencies. Second, a quantization mechanism is proposed to compress the control signals, thereby reducing the required communication bandwidth. Then, a dual-channel event-triggered quantized model-free adaptive control method is formulated, in which the accurate dynamics of the controlled helicopter are no longer needed. A stability analysis demonstrates that the control error of the controlled system is bounded by the designed method. Finally, the effectiveness of the proposed method is verified through simulation studies.
This paper proposes a dual-channel dynamic event-triggered (DET) constrained control strategy with packet loss compensation to regulate power in battery energy storage systems (BESSs) under packet loss conditions and input saturation constraints. The compact form dynamic linearization (CFDL) method is first employed to simplify the modeling of BESSs. Based on the linearized model, output data packet loss is compensated, while an input data packet loss compensation mechanism is developed using the latest control increment. A dynamic antiwindup compensator is designed to mitigate the effects of constraints on the magnitude and rate of control input. Furthermore, semi-independent DET mechanisms are established for control input and measurement output channels to reduce communication and computation resource consumption. Theoretical convergence analysis is given by employing the contraction mapping principle and indicates that the tracking error is uniformly bounded. Finally, numerical simulations verify the efficacy of the proposed approach.
This paper is concerned with the tracking control problem of a quadrotor for the motion trajectory of wind turbine blades subject to external disturbances. Due to the unpredictability and complexity of the existing disturbances such as gusts and turbulence, a dual closed-loop active disturbance rejection control (ADRC) method, including a position loop and an attitude loop, is proposed to mitigate the effect of external disturbances on the quadrotor’s flight. In particular, an extended state observer (ESO) is first employed to estimate the real-time disturbances. Then, a control law is further designed so as to suppress the impact of disturbances on the wind turbine blades’ trajectories. The convergence analysis of the extended state observer and the stability proof of the system are also presented. Experimental simulations are conducted to show the superiority of the proposed control strategy. In addition, in comparison with the traditional PID-based control method, the proposed dual closed-loop ADRC tracking control scheme demonstrates superior control performance in the presence of external disturbances with high intensity, which confirms the effectiveness of the proposed control architecture.
This paper addresses sliding-mode control for bridge crane systems within a finite time horizon, accounting for uncertain dynamics and sensor saturation constraints. Firstly, partial feedback linearization is applied to the nonlinear model of the bridge crane system to streamline the control design process. Then, a terminal sliding-mode surface is constructed, and a robust sliding-mode controller is designed to ensure system performance in the presence of uncertainty. Moreover, a uniform quantization scheme is developed to limit the data bit length during communication, thereby optimizing resource utilization. Finally, through rigorous theoretical analysis, we prove finite-time convergence of the proposed control method, and extensive simulations validate its effectiveness and superiority.
In this work, we delve into the asymptotic attitude tracking control problem of a class of uncertain quadrotors, considering multiple kinds of uncertainties including unknown system parameters, unmodeled attitude dynamics, unforeseen external disturbances, and undetected actuator faults. By leveraging the intrinsic structural properties of quadrotors, we establish an enhanced controllability condition to reduce the conservativeness of current ones, upon which we design a robust adaptive tracking control scheme by using some easily computable “deep-rooted information” derived from the quadrotor model, which not only lets the attitude tracking errors reduce to zero asymptotically, but also remains the structural complexity at the same level compared with the results using traditional controllability conditions. Moreover, all the signals in quadrotor closed-loop prove to be ultimately uniformly bounded (UUB) and the effectiveness of our proposed approach is verified via numerical simulations.
Pulmonary nodule detection in CT images plays a crucial role in the early diagnosis of lung cancer. However, the development of efficient detection systems, especially in reliable detection of small nodules with low false positive rates, remains challenging as small nodules often occupy minimal pixel regions and are morphologically confusable with vascular and surrounding anatomical structures. In this paper, we propose an alternate nodule detection framework to solve such tricky issue, which exhibits several salient features. Firstly, we propose a Progressive Multi-Scale Network (PMSN), a feature extraction module with a four-branch interactive architecture, specifically designed to enhance small nodule detection. By integrating this module into the YOLO object detection framework, we develop a PMSN-YOLOX network that effectively generates small-nodule-sensitive 2D candidates. Based on this, we develop a Geometric Constraint False Positive Reduction (Geo-FPR) algorithm via spatial clustering and confidence scoring to eliminate inconsistent detections while preserving true nodules. Additionally, we employ a self-supervised pre-training framework and a dynamic segmentation enhancement strategy to further improve model robustness. Experimental results on the LUNA16 dataset demonstrate that the proposed method exhibits superior performance in small nodule detection and false positive reduction, achieving 100% sensitivity with 2 to 8 false positives per scan. Even at the strict 1/8 false positives per scan threshold, sensitivity remains 98.3%. In addition, only 1.2 s is required to process each case. In summary, the proposed approach significantly outperforms existing SOTA techniques in terms of both performance and efficiency.
In this paper, we study the robust control of discrete event systems modeled by Petri nets with uncontrollable and unobservable transitions under replacement attacks. The existing supervisory control methods for discrete event systems modeled as Petri nets under replacement attack can only deal with the systems with observable and controllable transitions, and cannot deal with the systems with unobservable and uncontrollable transitions. Three methods are proposed in this paper to prevent the system from reaching illegal markings by disabling some controllable transitions according to a generalized mutual exclusion constraint. These methods can deal with Petri nets containing transitions that are not observed or controlled. The proposed framework provides a versatile range of strategies, from optimal control policies for high-precision requirements to efficient structural algorithms suitable for large-scale systems. It effectively mitigates the impact of replacement attacks while ensuring rigorous system safety. By extending Petri net supervisors to compromised environments with incomplete information, this work establishes a robust theoretical foundation for securing complex industrial discrete event systems.
Predicting wind power is essential for the stable and economical operation of power grids. Traditional point forecasts often struggle to capture volatility, while wind power interval prediction (WPIP) effectively quantifies associated uncertainties, providing comprehensive information for power dispatch. This paper proposes a novel deep-broad WPIP ensemble framework. First, considering the relationships among variables collected by the wind turbine’s supervisory control and data acquisition system, an operational mechanism-based method is designed to identify the outliers, while a novel multiple imputation approach is proposed to fill in the missing values and rectify the detected outliers to guarantee good data quality. Then, based on fuzzy and rough set principles, rolling fuzzy information granulation is used to capture the wind power fluctuation and construct high-quality training labels. Subsequently, the graph attention network with dynamic attention and gated recurrent unit, and the stacked broad learning system are developed separately as wind power interval predictors. These two models can extract and fuse spatial–temporal features selected from the training data thoroughly and efficiently to enhance the WPIP accuracy. Moreover, the human evolutionary optimization algorithm is used to dynamically optimize the combined weights online and integrate the prediction results of these two predictors to improve the interval prediction performance effectively. The comparative and ablation experiments demonstrate that our model outperforms the benchmark methods.
