欠驱动USV神经网络自适应轨迹跟踪控制

Adaptive neural network trajectory tracking control for underactuated unmanned surface vehicle

摘要:

为解决欠驱动USV系统在模型不确定性和未知海流干扰下的水平面轨迹跟踪问题,基于反步法与自适应技术,提出一种非线性鲁棒轨迹跟踪控制策略. 针对欠驱动USV参数不确定问题,运用神经网络自适应方法对欠驱动USV系统未知函数进行估计和逼近；然后,运用动态面方法获取虚拟变量的导数,不仅控制律结构简单,易于工程实现,而且有效地减小了传统反步法中虚拟变量直接求导的复杂性；针对未知时变海流的干扰,设计了一种指数收敛海流观测器,有效估计未知缓慢时变海流速度；其次,基于李雅普诺夫理论证明了所设计的控制器能够保证运动轨迹收敛于期望值,并且保证了该轨迹跟踪闭环控制系统所有信号最终一致有界；最后,在控制输入受限条件下,为检验该控制器的跟踪性能,选取圆轨迹作为参考轨迹,仿真实验表明,该控制器能够有效地实现预期轨迹,神经网络自适应方法对系统未知函数可有效估计和逼近,对未知时变海流干扰具有较强的鲁棒性,轨迹跟踪误差和速度跟踪误差均收敛到零附近的一个邻域内,从而验证了所提出控制器的有效性.

Abstract:

To deal with the problems of horizontal trajectory tracking control of underactuated unmanned surface vehicle (USV) in the presence of model uncertainties and unknown ocean currents, a robust nonlinear trajectory tracking controller for underactuated USV was proposed based on the backstepping method and adaptive technique. For the model uncertainties of underactuated USV, the adaptive neural network technique was employed to estimate and compensate the unknown model uncertainties. Then, the derivatives of virtual control variables were obtained by dynamic surface control method. The control law was simple in structure and easy to be realized in engineering, and it greatly reduced the complexities of the traditional backstepping method. For the disturbances of unknown time-varying currents, an observer was designed to estimate the velocity of unknown time-varying currents. Next, based on Lyapunov’s direct method, it was proved that the designed controller could ensure that the motion trajectory converged to the expected value, and that all signals of the trajectory tracking closed-loop control system were finally uniformly bounded. Lastly, under the condition of limited control input, in order to verify the tracking performance of the controller, the circular trajectory was selected as the reference trajectory. Simulations were carried out and results show that the controller could accurately track the desired trajectory and had strong robustness for model uncertainties and unknown time-varying currents. In addition, the unknown functions of the system were effectively estimated and compensated by adaptive neural network technique, which verified the effectiveness of the proposed tracking control scheme.