论文部分内容阅读
针对一类单输入单输出高阶非线性控制系统,提出一种基于滑模思想和Elman网络的操作条件反射(OCR)学习控制方法.该方法采用Elman网络构造滑模面-行为对的评价函数,通过滑模面的变化设计奖赏函数,根据奖赏信号更新评价函数,实现行为选择概率的更新.通过每轮次熵的定义,定量分析了所学知识的变化量.针对行走倒立摆系统的仿真实验结果表明,采用该仿生的OCR学习控制方法,可实现行走倒立摆的平衡控制.
Aiming at a class of single input single output high order nonlinear control systems, an operating condition reflection (OCR) learning control method based on sliding mode thought and Elman network is proposed. The method uses Elman network to construct the evaluation function of sliding surface plane-behavior pairs , The rewards function is designed based on the change of sliding surface and the evaluation function is updated according to reward signals to update the probability of behavior selection.The quantitative change of knowledge is analyzed by the definition of rounds of entropy.Aim at the simulation of walking inverted pendulum system The experimental results show that the bionic OCR learning control method can realize the balance control of walking inverted pendulum.