IEEE Transactions on Cybernetics最新文献

Machine learning-based optimization frameworks have attracted increasing attention for accelerating the solution of large-scale quadratically constrained quadratic programs (QCQPs) by exploiting shared problem structure across instances. However, existing machine learning (ML) frameworks often rely on the assumption of parametric models and large-scale solvers. This article introduces HEQP, a hypergraph neural network-based evolutionary optimization framework for large-scale QCQPs. This framework features two main components: 1) hypergraph-based neural prediction, which predicts optimal solutions for QCQPs without assumptions of models; and 2) evolutionary large neighborhood search (Evo-LNS), which employs a McCormick relaxation-based repair strategy to search and apply crossover on neighborhood solutions using a small-scale solver. We further show that our framework is equivalent to the interior-point method (IPM), a polynomial-time algorithm, for quadratic programming. Experiments on two types of benchmark problems and 13 large-scale real-world instances from the QPLIB illustrate that our framework outperforms state-of-the-art solvers (including Gurobi, SCIP, and SHOT) in both solution quality and time efficiency, highlighting the efficiency of ML-based optimization frameworks for QCQPs.

Traditional evolutionary game theory (EGT) typically assumes fixed, well-mixed populations, neglecting the fact that agents in many real-world systems can strategically form or eliminate connections based on individual incentives. This article introduces a novel framework that integrates EGT with strategic network formation to model the co-evolution of strategies and intercommunity links. We consider populations in which agents interact primarily within their own communities but may establish external connections when such interactions yield higher payoffs. To formalize this, we propose a strategic connections model (SCM) based on the concept of pairwise stability that determines which subsets of agents from different communities form links, and how these links influence evolutionary dynamics. The SCM operates as a two-stage optimization process that accounts for mutual incentives and payoff improvements. Applying our framework to classical evolutionary games, we show that intercommunity connections reshape population-level outcomes. In the prisoner's dilemma, cooperation-typically unstable in well-mixed settings-can emerge and persist under a simple benefit-to-cost condition. In the rock-paper-scissors ( $RPS$ ) game, intercommunity links can alter or suppress the characteristic cycles observed, for instance, under replicator dynamics, affecting the long-term coexistence of strategies. In congestion games, our framework improves infrastructure usage and resource allocation, outperforming intuitive but nonstrategic connection choices. These results highlight the critical role of strategic intercommunity links in shaping collective behavior, offering new insights into the interplay between intercommunity and intracommunity dynamics in biological, social, and engineered systems, where such links can represent interactions among species, individuals, and/or machines within cybernetic environments.

Transferability estimation is a fundamental problem in transfer learning, which aims to predict whether transferring knowledge from a source domain will improve performance on a target task. Existing research focuses on classification and neglects domain/task differences, as well as only very limited research for regression. Most importantly, there is a lack of research to determine whether to transfer or not. To address these gaps, we propose Wasserstein distance-based joint estimation (WDJE), a unified transferability metric for both classification and regression under domain and task differences. WDJE facilitates decision-making on whether to transfer by comparing the target risk with and without transfer. To enable this comparison, we estimate the unobservable post-transfer risk using a nonsymmetric, interpretable, and easy-to-calculate upper bound that remains applicable even with limited target labels. The proposed bound relates the target transfer risk to source model performance, domain, and task differences based on the Wasserstein distance. We further extend the proposed bound to the unsupervised setting and establish a generalization bound from finite empirical samples. We evaluate WDJE and the proposed risk bound across 42 transfer scenarios, including CIFAR-100 (CF100) and Office-Home image classification and C-MAPSS remaining-useful-life regression prediction. WDJE achieves a perfect consistency index ( $CI$ ) of 1 in 25 cases and an overall mean $CI$ of 0.89, accurately suggesting when transfer should (or should not) be performed. The proposed bound achieves the average Pearson correlations of 0.99 on CF100, 0.72 on Office-Home, and 0.96 on C-MAPSS, illustrating state-of-the-art performance in approximating the true post-transfer risk.

In this article, the stability and stabilization for T-S fuzzy systems with periodically varying delays (PVDs) are explored. First, according to the characteristics of PVDs, an augmented Lyapunov-Krasovskii functional (LKF) related to fuzzy rules is established. Second, enhanced stability criteria are developed based on fuzzy augmented LKF together with a generalized delay derivative-dependent reciprocally convex matrix inequality (RCMI). It successfully overcomes the limitations of estimation for only two reciprocally convex terms (RCTs), meanwhile providing a tighter bound by introducing more free variables. Under the parallel distributed compensation (PDC) scheme, a fuzzy memory controller ensuring system stabilization is investigated. Finally, the enhancement of stability criteria and the effectiveness for controller design are verified through numerical simulation.

In this article, the distributed filtering issue is explored for time-varying state-saturated systems affected by Byzantine attacks over sensor networks. To regulate data transmission, a token bucket protocol (TBP) is utilized, in which the stochastic nature of token consumption arises from variations in packet sizes. Particularly, the measurements are transmitted to the filter only when the available tokens suffice to meet the required consumption. A Byzantine attack model is formulated in which Byzantine nodes arbitrarily alter the measurement signals transmitted to neighboring nodes. The primary objective is to construct an upper bound of the filtering error covariance (FEC) and to compute suitable filter gains by minimizing this bound. Furthermore, the boundedness of the proposed filtering error dynamics is rigorously analyzed via matrix-based theoretical analysis. Finally, numerical simulations are conducted to verify the effectiveness of the proposed algorithm.

This research endeavors to design the perching maneuver and control in ornithopter robots. By analyzing the dynamic interplay between the robot's flight dynamics, feedback loops, and the environmental constraints, we aim to advance our understanding of the perching maneuver, drawing parallels to biological systems. Inspired by the elegant control strategies observed in avian flight, we develop an optimal maneuver and a corresponding controller to achieve stable perching. The maneuver consists of a deceleration and a rapid pitch-up (vertical turn), which arises from analytically solving the optimization problem of minimal velocity at perch, subject to kinematic and dynamic constraints. The controller for the flapping frequency and tail symmetric deflection is nonlinear and adaptive, ensuring robustly stable perching. Indeed, such adaptive behavior in a sense incorporates homeostatic principles of cybernetics into the control system, enhancing the robot's ability to adapt to unexpected disturbances and maintain a stable posture during the perching maneuver. The resulting autonomous perching maneuvers-closed-loop descent and turn-have been verified and validated, demonstrating excellent agreement with real bird perching trajectories reported in the literature. These findings lay the theoretical groundwork for the development of future prototypes that better imitate the skillful perching maneuvers of birds.

This article presents a prescribed-performance-controller-based equivalent-input-disturbance approach (PEID) that enhances the disturbance-rejection performance of a Stewart platform. The PEID approach ensures that the disturbance-estimation error converges to a bounded region in a prescribed time. The bound of disturbance-estimation error can be chosen arbitrarily small, and the convergence time of disturbance-estimation error can be prescribed in advance. Stability conditions are derived by dividing the PEID-based control system into three subsystems, and a barrier Lyapunov function is used to design a prescribed-performance-controller-based compensator. An algorithm for designing system parameters devises state feedback and observer gains. The experiments conducted on a Stewart platform validate the effectiveness and superiority of the PEID approach.

This article presents a novel reinforcement learning (RL)-based control scheme for trajectory tracking of robot manipulators, incorporating an uncertainty estimator and an optimal tracking controller. A momentum observer (MO) is implemented in conjunction with an integral sliding mode control (ISMC) to facilitate uncertainty estimation for compensation. A proportional-derivative-like (PD-like) control plus an actor-critic neural network (NN)-based feedforward control is utilized for trajectory tracking. An NN parameter selection scheme is adopted to circumvent the time-consuming adjustments of its activation functions and initial weights, thereby ensuring the admissibility of the initial control policy and improving its efficiency. Lyapunov function analysis demonstrates stability of the closed-loop system, in which all error signals are shown to remain bounded and eventually settle into a small residual set. The proposed control scheme is compared with a conventional PD approach consisting of a feedforward controller and an NN controller with adaptive radial basis functions (RBFs). Comparison is also made with a recent approach featuring a feedforward super-twisting sliding mode control (FSTSMC). Simulation and experimental results show superior tracking performance of the presented control scheme against the comparative counterparts, validating the new approach and further supporting its effectiveness and feasibility.

Robust multiobjective evolutionary algorithms (RMOEAs) aim to obtain robust optimal solutions. However, traditional RMOEAs typically require evaluating a large number of sampling points, which is often impractical in real-world applications due to the high computational cost. In this article, we propose a robust multiobjective evolutionary algorithm based on surrogate-assisted (RMOEA-SA), which incorporates a radial basis function (RBF) surrogate model and a novel robust distance metric (RDM). The proposed algorithm employs the RBF surrogate model to approximate the fitness values of sampling points, thereby significantly reducing the number of function evaluations during the robust optimization process. Furthermore, an RDM assisted by the RBF surrogate model is introduced to measure the robustness of solutions. Besides, the RDM value of each solution is treated as an additional objective, expanding the original objective space, and selection is conducted in this augmented space to achieve a desirable trade-off between robustness and optimality. The experimental results on standard benchmark functions and two real-world application problems demonstrate the superior feasibility and effectiveness of the proposed method compared with several existing algorithms.

This article addresses the lateral dynamics control problem for autonomous vehicle systems under randomly perturbed sampling (RPS) periods and the FlexRay communication protocol. To capture vehicle nonlinearities under variable-velocity conditions, a T-S fuzzy model is constructed using longitudinal velocity as the premise variable. The random sampling behavior caused by hardware aging and environmental disturbances is modeled as a Markovian process. Then, measured outputs are transmitted under the FlexRay protocol (FRP) that integrates both time-driven (static) and event-driven (dynamic) scheduling characteristics. By fully analyzing the situation of static and dynamic scheduling, a unified compensation strategy is employed to build a new switching output model reflecting the impact of the FRP on the measured outputs. Based on this output model, a sampling-mode-dependent fuzzy controller is designed to handle random sampling and hybrid scheduling issues, which results in a membership asynchronous phenomenon between the autonomous vehicle model and controller. By using the asynchronous constraint technique, sufficient conditions with low conservatism are derived to guarantee stochastic stability and $H_{infty }$ performance of the closed-loop system. Furthermore, a comprehensive optimization problem (OP) is established, and a corresponding genetic algorithm (GA) is presented to provide a solution-solving scheme. Simulation results confirm the effectiveness and superiority of the proposed control strategy under complex communication environments.