求解最优控制问题的混合变量变分方法及其航天控制应用  

A Mixed Variable Variational Method for Optimal Control Problems with Applications in Aerospace Control

作　　者：彭海军[1] 高强[1] 吴志刚[2] 钟万勰[1] 

机构地区：[1]大连理工大学工程力学系工业装备结构分析国家重点实验室,大连116024 [2]大连理工大学航空航天学院工业装备结构分析国家重点实验室,大连116024

出　　处：《自动化学报》2011年第10期1248-1255,共8页Acta Automatica Sinica

基　　金：国家重点基础研究发展计划(973计划)(2010CB832704;2009CB918501);国家高技术研究发展计划(863计划)(2009AA044501);国家自然科学基金(10902020;11072044);中央高校基本科研业务费(DUT11ZD(G)02);辽宁省博士启动基金(20081091);辽宁省重点实验室项目(2009S018)资助~~

摘　　要：针对非线性最优控制导出的Hamiltonian系统两点边值问题,提出一种以离散区段右端状态和左端协态为混合独立变量的数值求解方法,将非线性Hamiltonian系统两点边值问题的求解通过混合独立变量变分原理转化为非线性方程组求解.所提出的算法综合了求解最优控制的"直接法"和"间接法"的特征,既满足最优控制理论的一阶必要条件,又不需要对协态初值的准确猜测,避免了求解大规模非线性规划问题.通过两个航天控制算例讨论了本文算法的精度和效率等问题.与近年来在航空航天控制中备受关注的高斯伪谱方法相比较,本文算法无论是在精度还是效率上都具有明显的优势.The nonlinear optimal control problem is transformed into the Hamiltonian two point boundary value problem（TPBVP）,and a numerical method is proposed based on the dual variational principle.The nonlinear Hamiltonian TPBVP is transformed into a system of nonlinear equations by using the dual variational principle and by taking the left costate and the right state as independent variables.The proposed algorithm has the feature of both ＂direct method＂ and ＂indirect method＂ for solving optimal control problem,i.e.,it satisfies the first-order necessary condition of optimal control theory,and it needs no precise initial guess for costate variables and avoids the solving of large scale nonlinear programming problem.The accuracy and efficiency of the proposed method are discussed by numerical simulations in aerospace control.Comparison between the proposed algorithm and the famous Gauss pseudo-spectral method in aeronautics and astronautics control shows that the proposed algorithm has obvious advantages on accuracy and efficiency.

关 键 词：非线性最优控制 两点边值 对偶变量 变分原理 HAMILTONIAN系统 

分 类 号：V448.2[航空宇航科学与技术—飞行器设计]