基于近似动态规划方法的未知系统的最优跟踪控制  

Optimal Tracking Control Based on Approximate Dynamic Programming for Unknown System

作　　者：崔小红[1] 王缔[1] 刘素兵[2] 柴宝杰[3] 

机构地区：[1]牡丹江师范学院理学院,黑龙江牡丹江157012 [2]第二炮兵工程大学,陕西西安710025 [3]牡丹江师范学院工学院,黑龙江牡丹江157012

出　　处：《数学的实践与认识》2015年第11期266-274,共9页Mathematics in Practice and Theory

基　　金：黑龙江省教育厅教改项目(JG2014011057);牡丹江师范学院省级重点创新预研项目(SY2014005;SY2014006)

摘　　要：针对不确定的非线性连续系统,通过神经网络对系统进行辨识.基于辨识后的确定系统,利用执行网-评价网双网结构进行同步调节解决最优跟踪问题.应用李雅谱诺夫方法进行辨识分析和系统稳定性分析,定理结论表明辨识系统为渐近辨识的,同时系统的跟踪误差和权重误差一致最终有界,倒立摆仿真例子验证了算法的有效性.In this paper, for the unknown continuous nonlinear system, Neural Networks identifier model is applied. Based on the identified model, Approximate Dynamic Program- ming（ADP） is used and actor-critic dual networks simultaneously are tuned in order to solve the optimal tracking problem. The identified result and the stability analysis based on the Lyapunov method are given. Theorems shows the identification is asymptotic and all the tracking error and the estimate weights are UUB, the inverted pendulum system simulation verified the effectiveness of algorithm.

关 键 词：近似动态规划 神经网路 最优控制 跟踪问题 

分 类 号：O232[理学—运筹学与控制论]