基于自适应神经网络的AUVs路径跟踪控制

Control of Path Following for AUVs Based on Adaptive Neural Network

为解决由于随时间变化水动力阻尼引起的参数变化和不确定性的问题,提出了基于径向基函数神经网络的未知评估算法,引入自适应算法以保证神经网络权值的最优评估。基于Lyapunov稳定性理论,设计一种自适应神经网络控制器以保证路径跟踪系统中所有误差状态都趋于稳定。为了验证该控制器的可行性,对系统施加如位置误差、方向误差等虚拟干扰,证明该控制器可将误差消减为零。另一方面,机器人在以恒定的速度行驶时,每个航点被指定一个适合半径的圆弧可以保证其有较高的精度。为了评估路径跟踪控制器的性能,提出直线型和直线加圆弧型路径方案。仿真结果表明,该控制器可以有效地消除机器人非线性和模型不确定性造成的干扰。

In order to solve the problems of parameters variation and uncertainty caused by time-varying hydrodynamic damping,the unknown assessment algorithm based on radial basis function（RBF） neural network is put forward,and an adaptive algorithm is introduced to ensure optimal evaluation of the neural network weights.Based on Lyapunov stability theory,an adaptive neural network controller is designed to ensure that all error states of the path following system tend to stablize.To verify the feasibility of the controller,virtual interferences such as position error,direction error,etc are applied to the system,which prove that the controller error can be reduced to zero.On the other hand,when the robot travels at a constant speed,each waypoint is assigned a suitable arc radius to guarantee a higher accuracy.In order to evaluate the performance of path following controller,linear and linear acceleration arc path programs are proposed.Simulation results show that the controller can effectively eliminate the robot nonlinearity and the interference caused by model uncertainty.