一类非线性系统基于SVR的鲁棒自适应控制  

Robust Adaptive Control for a Class of Nonlinear Systems Based on Support Vector Regression

作　　者：刘毅男 张超 

机构地区：[1]第二炮兵工程大学自动控制工程系,陕西西安710025

出　　处：《控制工程》2013年第6期1055-1059,共5页Control Engineering of China

基　　金：国家863计划(2011AA7053016)

摘　　要：为解决逆控制方法较依赖系统精确模型及易受逆误差影响等问题,提出了一类非线性系统基于支持向量回归(Support Vector Regression,SVR)的鲁棒自适应控制方法,该方法通过支持向量回归辨识得到被控对象的逆模型,并构建自适应逆误差补偿环节,在线修正由于逆模型误差、外界干扰等不确定因素对控制系统性能的不良影响,使得系统能够快速、准确的跟踪参考模型输出。导出了支持向量回归的权值自适应调整率,并利用Lyapunov稳定性理论证明了系统闭环渐近稳定。通过对典型非线性模型的仿真研究,证明了此控制方案在解决一类非线性问题上的可行性,且鲁棒性较好。To solve the problem that nonlinear dynamic inverse method is relatively dependent on the accurate system model and easily affected by inverse error, a nonlinear robust adaptive controller design using Squares Support Regression （SVR） is proposed in this pa- per, in which SVR is used to identify inverse model of the controlled system and construct adaptive compensation term to eliminate the inverse error caused by uncertain factors, such as inverse model error or disturbances, and so on. This could make the system outputs track the outputs of the reference model rapidly and accurately. The updating rule of SVR parameters is derived, and the stability of the nonlinear system is proved by Lyapunov stability theory. Furthermore, the simulation results of a typical nonlinear model show the feasi- bility and robustness on solving a class of nonlinear problems of the proposed method.

关 键 词：支持向量回归 自适应控制 反馈线性化 非线性 鲁棒性 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]