【目的】构建基于自适应动态规划的多驱动负载系统跟踪控制器。【方法】应用神经网络学习多驱动负载伺服系统的未知动态;给定关于跟踪误差和能耗的性能指标,应用逼近的非线性特性,设计系统基于近似动态规划的跟踪控制器。可以分解为稳态控制和最优反馈控制两部分,稳态控制器可根据跟踪信号和系统动态直接获得;最优反馈控制应用近似动态规划方法求得,多个最优反馈控制达到纳什均衡,最小化性能指标函数,优化系统运行性能。应用李雅普诺夫方法分析最优多输入情况下学习权值的收敛性以及负载系统的稳定性。【结果】仿真结果证明所提方法能够优化多驱动负载的跟踪性能。

【Purposes】 The unknown multi-motor driven load systems are addressed in this pa‐ per with game theory and adaptive dynamic programming. 【Methods】 The neural network (NN) is  used to approximate the unknown dynamics. On the basis of the approximated dynamics, the optimal  performance index is defined and the optimal tracking controls are designed. The adaptive dynamic  programming based tracking controls can be divided into steady-state controls and optimal feedback  controls. Steady-state controls can be directly obtained with the tracking commands and load dynam‐ ics. The optimal feedback controls are obtained with an approximate dynamic programming algorithm,  which can be used to research the saddle point and minimized the performance index. The convergence  of the NN weights is analyzed and the stability of the load system is proved. 【Findings】 A simulation  is provided to illustrate the effectiveness of the methods, which can optimize the load performance.