基于数据的自学习优化控制:研究进展与展望  

Data-based Self-learning Optimal Control: Research Progress and Prospects

作　　者：刘德荣[1] 李宏亮[1] 王鼎[1] 

机构地区：[1]中国科学院自动化研究所复杂系统管理与控制国家重点实验室,北京100190

出　　处：《自动化学报》2013年第11期1858-1870,共13页Acta Automatica Sinica

基　　金：国家自然科学基金(61034002;61233001;61273140)资助~~

摘　　要：自适应动态规划(Adaptive dynamic programming,ADP)方法可以解决传统动态规划中的"维数灾"问题,已经成为控制理论和计算智能领域最新的研究热点.ADP方法采用函数近似结构来估计系统性能指标函数,然后依据最优性原理来获得近优的控制策略.ADP是一种具有学习和优化能力的智能控制方法,在求解复杂非线性系统的最优控制问题中具有极大的潜力.本文对ADP的理论研究、算法实现、相关应用等方面进行了全面的梳理,涵盖了最新的研究进展,并对ADP的未来发展趋势进行了分析和展望.Adaptive dynamic programming （ADP） method can solve the problem of ＂curse of dimensionality＂ in the traditional dynamic programming, and has recently become a hot topic in the field of control theory and computational intelligence. For ADP method, a function approximation structure is used to estimate the performance index function, and then the approximate optimal control policy can be obtained based on the principle of optimality. As a kind of intelligent control methods with learning and optimization capabilities, ADP has great potential in solving the optimal control problem of complex nonlinear systems. This paper presents a comprehensive survey on the theoretical research, algorithm development, and related applications of ADP, which covers the latest research progress. It also analyzes and predicts the future development trend of ADP.

关 键 词：自适应动态规划 近似动态规划 强化学习 神经网络 智能控制 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]