引用本文: | 林小峰,黄元君,宋春宁.带ε误差限的近似最优控制[J].控制理论与应用,2012,29(1):104~108.[点击复制] |
LIN Xiao-feng,HUANG Yuan-jun,SONG Chun-ning.Approximate optimal control with ε-error bound[J].Control Theory and Technology,2012,29(1):104~108.[点击复制] |
|
带ε误差限的近似最优控制 |
Approximate optimal control with ε-error bound |
摘要点击 2793 全文点击 2038 投稿时间:2011-01-10 修订日期:2011-04-18 |
查看全文 查看/发表评论 下载PDF阅读器 |
DOI编号 |
2012,29(1):104-108 |
中文关键词 ε误差限 非线性系统 近似动态规划 最优控制 |
英文关键词 ε-error bound nonlinear systems approximate dynamic programming optimal control |
基金项目 国家自然科学基金资助项目(60964002); 广西理工科学实验中心重点资助项目(LGZX201106); 广西研究生教育创新计划资助项目(105931003009). |
|
中文摘要 |
近似动态规划方法求解非线性系统最优控制, 需要迭代无限步才能得到最优控制律. 本文提出了一种ε–近似最优控制算法, 选择ε误差限, 通过自适应迭代不断逼近哈密顿– 雅可比– 贝尔曼(HJB)方程的解, 应用神经网络实现在有限步迭代后得到带ε误差限的近似最优控制律. 计算机仿真结果表明了该算法的有效性. |
英文摘要 |
In applying the approximate dynamic programming algorithm to determine the optimal control law for nonlinear systems, we need an infinite number of iterations to obtain the result. To deal with this problem, we propose an algorithm for obtaining an approximate optimal control law. Based on the given value of the ε-error bound, the approximation optimal control law will approach the solution of Hamilton-Jacobi-Bellman(HJB) equation through self-adaptive iteration. By applying neural networks, we can obtain the ε-approximation control law in a finite number of iterations. Simulation validates the efficacy of the above algorithm. |