基于RBF神经网络自适应PID四旋翼飞行器控制  

Adaptive PID Control of Quadrotor Based on RBF Neural Network

作　　者：李砚浓 李汀兰[2] 姜艺[3] 范家璐 

机构地区：[1]东北大学信息科学与工程学院,沈阳110819 [2]东北育才学校科学高中部,沈阳110179 [3]东北大学流程工业综合自动化国家重点实验室,沈阳110819

出　　处：《控制工程》2016年第3期378-382,共5页Control Engineering of China

基　　金：国家自然科学基金项目(61304028);863项目(2015AA043802);辽宁省教育厅项目(L2014087)

摘　　要：四旋翼飞行器具有高非线性、强耦合、欠驱动等特点,飞行控制器设计困难。首先根据牛顿欧拉方程建立系统动力学模型;然后根据系统动力学模型推导出控制器设计模型,根据系统模型设计出串级控制策略以实现对系统姿态控制和位置控制的解耦;最后针对四旋翼飞行器的控制难点提出了基于RBF的多变量神经网络自适应PID控制方法。该方法具有神经网络自学习、自适应的能力,同时具有一定的非线性控制作用。仿真结果表明,该方法相对于常规PID控制方法具有更短的调节时间、更少的超调量、更好的抗扰动能力,同时,在模型参数变化的情况下该控制器比常规PID控制器的鲁棒性更强。Attitude and position control of the quadrotor is a challenging task because it is an under-actuated system with strong nonlinear, coupling characteristics. In this paper, the dynamics model and the state space function of the quadrotor are firstly established according to Newton Euler equation. Then, a cascade control scheme is proposed to decouple the control system and a multivariate neural network based on RBF adaptive PID control algorithm is proposed to realize robust control of the quad-rotor. This algorithm is not only capable of self-adaption and online learning, but also action of nonlinear control. The simulation results show that compared to the conventional PID controller, this controller has a shorter settling time, less overshoot and better resistance to disturbances. What's more, this controller has a stronger robustness than the conventional PID controller when the model parameters are changed.

关 键 词：四旋翼飞行器 神经网络 PID控制器 自适应 

分 类 号：V249.1[航空宇航科学与技术—飞行器设计] TP183[自动化与计算机技术—控制理论与控制工程]