基于PID神经网络的三自由度直升机控制

摘要

使用一种PID神经网络控制器对具有仿射非线性强耦合的三自由度直升机系统进行仿真和实时控制,效果达到预期目的.PID神经网络多变量控制具有很强的适应性,根据神经网络的BP学习算法,通过神经网络的离线仿真可以确定神经网络权值的初值.再通过在线学习可以实时地更新神经网络权值对非线性系统进行自适应控制.PID神经网络输出层之间彼此交叉,实现了三自由度直升机俯仰轴和横滚轴的解耦控制.横滚轴在水平方向上产生的力矩驱动旋转轴在水平面的旋转.把横滚轴作为输入,使用传统PID控制器使得旋转轴达到了满意的控制效果.%Using a PID neural-network multivariable controller to simulate and real-time control the affine nonlinear and strongly coupled 3-DOF Helicopter system,achieve expected effect. PID neural-network multivariable controller has strong adapt?ability. According to BP neural-network learning algorithm,initial network weights can be determined by offline simulation based on mathematical model of 3-DOF Helicopter. And through online learning neural-network weights can be updated in real-time adaptive control of nonlinear systems. Since the PID neural-network output layers cross each other,the pitch axis and horizontal roll?er of the 3-DOF Helicopter have been decoupled. The torque generated by the horizontal axis in the horizontal direction drives the rotation of the rotary shaft in the horizontal plane. Using the horizontal axis as the input,using the traditional PID controller makes the rotation axis to achieve a satisfactory control effect.