项目名称: 最优控制的快速算法
项目编号: No.61473326
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 自动化技术、计算机技术
项目作者: 吴昌质
作者单位: 重庆师范大学
项目金额: 62万元
中文摘要: 由于反馈控制的鲁棒性,最优反馈控制在工程应用中极为重要。然而,对于带约束的非线性系统的最优反馈控制问题,现在仍缺少直接有效的方法。目前,对于该类问题,求其最优反馈控制通用技术是模型预测控制。该方法通过在线反复优化计算并滚动实施控制作用从而达到模型误差的反馈校正。其中,在线动态优化问题的快速求解是该技术应用成败的关键。对于线性模型预测控制,现在已有很多的快速算法可以满足在线计算的需要。然而,对于非线性模型,相关成果甚少。为此,本项目将针对非线性约束最优控制问题,开发出新型快速计算方法,以便满足模型预测控制在线求解的需要。具体的,我们将从以下两个方面展开研究:1)、利用多步长信息更新状态以便加速常规算法的收敛性;2)开发出新的分布式并行算法,从而将原来的大规模问题分解成小规模并行子问题求解。
中文关键词: 最优控制;加速算法;分布式算法;模型预测控制;非线性系统
英文摘要: Optimal feedback controls are of critical importance in engineering applications due to their robustness against disturbances and uncertainties. However, no practical methods for constructing optimal feedback controls are currently available for nonlinear constrained optimal control problems. In the current, one of the most popular way to construct an optimal feedback control is through model predictive control (MPC). The availability of this method is online solution of a dynamic optimizaiton problem. For linear MPC, there are a lot of methods available which can be used to satisfy the online computation. However, this is not the case for nonlinear MPC. This project aims to ?ll this gap through the development of new, novel and fast computational methods for dynamic optimization problem. More specifically, we will develop fast computational method through the following two aspects: 1). Utilizing multi-step information to accelerate the convergence rate of the traditional algorithm; 2)Developing new technique to decompose the original large-scaled dynamic optimization into parallel small-scaled dynamic optimizaiton problem such that the computational method can be realized in parallel.
英文关键词: optimal control;accelerated algorithm;distributed algorithm;model predictive control;nonlinear dynamical system