复变非线性系统自学习最优控制理论与方法研究

项目名称： 复变非线性系统自学习最优控制理论与方法研究

项目编号： No.61304079

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 宋睿卓

作者单位： 北京科技大学

项目金额： 26万元

中文摘要： 本项目将研究基于自适应动态规划的复变非线性系统最优控制问题。拟建立一套全新的复变非线性系统最优控制理论与实现方案，系统地解决复变非线性系统最优控制问题研究的难点，开辟对复变非线性系统控制与优化的新途径。主要研究内容包括：1.研究基于自适应动态规划的连续时间复变非线性系统最优控制器设计方案和稳定性分析方法；2.设计连续时间复变非线性系统鲁棒控制器，用来补偿神经网络近似误差，进而得到连续时间复变非线性系统鲁棒最优控制器；3.针对控制输入信号实部与虚部相互博弈的情况，研究连续时间复变非线性系统二人零和微分对策问题，获得最优控制器的表达式和求解方法；4.研究离散时间复变非线性系统最优控制问题，并证明闭环系统的渐近稳定性；5.开发本项目研究成果的软件仿真平台及硬件实验平台。上述研究成果将为自适应动态规划理论和复变非线性系统控制理论的发展提供新思路，丰富其研究内容，同时推动我国自动化技术的深入发展。

中文关键词： 自适应动态规划；复变非线性系统；神经网络；最优控制；博弈

英文摘要： The optimal control problems for complex-valued nonlinear systems are discussed based on adaptive dynamic programming in this project. An entirely novel optimal control theory and implementation methods for complex-valued nonlinear systems are established. The difficulties of optimal control for complex-valued nonlinear systems are solved systematically. The main works include: 1. The optimal controller design scheme and the stability analysis method for continuous time complex-valued nonlinear systems are proposed. 2.The robust controller is designed for complex-valued nonlinear systems, which is used to compensate neural network approximation errors. So the robust optimal control for continuous time complex-valued nonlinear systems is obtained. 3. According to the games of the real part and the imaginary part of the control input, the continuous time complex-valued nonlinear zero-sum games problem is discussed, and the optimal controller is obtained. 4. The optimal control problem of discrete time complex-valued nonlinear system is studied, and the asymptotic stability is proved. 5. Construst the soft simulation system and the hardware experimental system. The research achievements in this project, provide new ideas for the development of adaptive dynamic programming and complex-valued nonlinear systems, enric

英文关键词： adaptive dynamic programming；complex-valued nonlinear systems；neural networks；optimal control；games