自适应评价设计的执行依赖方法

Action-dependent Methods of Adaptive Critic Design

自适应评价设计(ACD)是一种适用于非线性系统的近似最优控制方法。介绍了自适应评价设计的执行依赖启发式动态规划(ADHDP)和执行依赖双启发式动态规划(ADDHP)方法,该方法可以解决由对象非线性或者系统建模不良所造成的不确定性问题,适于处理时变的复杂系统和动态变化的复杂任务。阐述了两种方法的结构、计算和评价网络输出上的不同,并通过仿真分析了两种方法各自的学习能力、控制效果。

Adaptive critic design（ACD）is a mothed used to approximate optimal control in nonlinear systems.Action-dependent heuristic dynamic programming and the action-dependent dual heuristic programming are introduced to solve the uncertainty problem created by non-linearity of plant or worse system modeling,and being appropriate to deal with the time-varying complex system and the dynamic variation complex task.The differences between the two action-dependent ACDs in structure,evaluation and critic＇s output are discussed,and the learning capability and control effect of the two methods are analyzed respectively.