基于神经网络的自主水下机器人动态反馈控制

Dynamic feedback control based on ANN compensation controller for AUV motions

针对自主水下机器人数学模型的强非线性及所受海流干扰无法确定等特点,设计了基于神经网络补偿器的动态反馈控制算法。通过对自主水下机器人系统数学模型研究,将系统分解为近似线性部分与非线性不确定部分。通过动态反馈控制实现对分解出的近似线性部分进行初步控制,利用神经网络所具有的自适应控制实现对不确定模型与干扰项进行补偿控制,提高自主水下机器人运动控制的鲁棒性。通过Lyapunov稳定判据证明了此控制算法的收敛性,通过Matlab数字仿真平台验证了此算法的抗干扰能力,在自主水下机器人半物理仿真平台上分别采用所提出的控制算法与PID控制算法对自主水下机器人系统定深运动进行了控制。研究结果表明所提出的算法具有更好的动态性能与强鲁棒性。

One algorithm of artificial neural network （ANN） compensator coupled with dynamical state feedback control was designed for motions of autonomous underwater vehicle （AUV）. The math model of AUV was decomposed to one extended linear subsystem and an unknown nonlinear subsystem. The fore part was controlled by dynamic state feedback control algorithm which was improved on feedback parame- ters described by some functions of time, poses, and surge speed as well as the last one being controlled by ANN whose parameters are online adjustment. Theories are proved and developed by Lyapunov theory which denoted that the tracking error can be converged to zero. Some simulation results with Matlab plat- form show that the advised control algorithm holds the capability of anti-disturbance. Meanwhile, these semi-physical environment experiences make it clear that the advised control algorithm can have better dy- namic oerformanee than PID control algorithm.