摘要
针对带有不确定项和部分输入受限的非线性系统多玩家合作博弈问题,设计了基于粒子群优化神经网络的事件触发鲁棒合作控制方法。首先,设计改进的值函数,将带有输入受限的鲁棒控制问题转化为针对标称系统的合作博弈问题。通过引入近似动态规划方法,利用粒子群优化的评判神经网络求解哈密尔顿-雅可比-贝尔曼方程,无需手动设定合格的初始权重。然后,设计事件触发条件以减少控制策略的更新频率,从而节省计算资源。此外,通过李雅普诺夫稳定性理论,证明闭环系统是稳定的。最后,通过仿真算例验证了所设计控制方法的有效性。
An event-triggered robust cooperative control method is developed based on particle swarm optimized neural networks for multi-player fully cooperative games of nonlinear uncertain systems with partial input constraints.A modified value function is designed,thus the input-constrained robust control problem is transformed into the fully cooperative game problem of the nominal system.Approximate dynamic programming is introduced to solve the Hamilton-Jacobi-Bellman equation via particle swarm optimized critic neural network,whose feasible initial weight vector is not required to be designed manually.Then,an event-triggering condition is designed to decrease the updating frequency of the control policies,which saves the computational resources.Moreover,the closed-loop system is proven to be stable via the Lyapunov stability theory.A simulation example is provided to demonstrate the effectiveness of the designed control method.
作者
吴球业
张勇威
张顺超
WU Qiuye;ZHANG Yongwei;ZHANG Shunchao(School of Public Security and Traffic Management,Guangdong Police College,Guangzhou 510000,China;College of Mathematics and Informatics,South China Agricultural University,Guangzhou 510000,China;School of Internet Finance and Information Engineering,Guangdong University of Finance,Guangzhou 510000,China)
出处
《电光与控制》
北大核心
2025年第9期86-91,103,共7页
Electronics Optics & Control
基金
国家自然科学基金(62303122)
广东省基础与应用基础研究基金联合基金(2021A1515110022)
广州市科技计划项目(2024A04 J3363)。
关键词
近似动态规划
合作博弈
事件触发控制
鲁棒控制
神经网络
粒子群优化算法
approximate dynamic programming
cooperative game
event-triggered control
robust control
neural network
particle swarm optimization
