基于支持向量机的自修复飞行控制及仿真研究

针对多输入多输出非线性耦合系统的控制问题,提出一种多输入多输出函数的支持向量机回归方法和一种基于支持向量机的逆控制方法。采用该回归算法对多输入多输出非线性耦合系统的逆系统进行辨识,再与原系统复合而成为一个伪线性系统,就可完成对非线性耦合系统的线性化解耦系统,最后设计相应的PID控制器以获得优良的动、静态特性与抗干扰能力。对一个两输入两输出非线性耦合系统的仿真结果表明该逆控制方法是一种较神经网络逆控制方法更为有效的控制律重构方法。

Abstract：

An algorithm of SVMs regressing to a multivariable inputs and multivariable outputs （MIMO） function and a inverse control based on SVMs were proposed for controlling MIMO nonlinear correlative systems. The inverse system of MIMO nonlinear correlative system was identified by MIMO SVMs regressing algorithm. The original system connected with the inverse system will become linear and irrelated. PID controller will be designed to improve the dynamic and static capability of the system. Simulation results show that SVMs inverse control is a more valid way of control reconfiguration than BP neural networks.