Optimal Sliding Mode Control of ROV Fixed Depth Attitude Based on Reinforcement Learning
Optimal Sliding Mode Control of ROV Fixed Depth Attitude Based on Reinforcement Learning
复制标题
基于强化学习的ROV定深姿态最优滑模控制
DOI：
10.1109/cyber53097.2021.9588177
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Li Zhigang
中科院分区：
文献类型：
--
作者：
Wang Fule;Qu Qiuxia;Yuan Baolong;Sun Liangliang;L. Yupeng;Guo Guanyan;Xiao Zupeng;Sun Liang;Li Zhigang
本文针对水下航行器测深控制系统提出了一种基于强化学习的积分滑模控制算法。由于非线性连续系统难以跟踪时变轨迹，通过引入新的状态变量，将最优跟踪问题转化为非线性时不变最优控制问题。采用自适应动态规划（ADP）算法求解非线性系统的HJB方程，找到近似最优策略。结合积分滑模控制，设计了近似最优滑模控制器。此外，利用Lyapunov方程验证了本文提出的控制策略能够保证系统的跟踪误差逐渐收敛于零，并且误差在小范围内也得到了验证。最后通过仿真实验验证了算法的有效性，增强了水下机器人在深度控制方向的抗干扰性和鲁棒性。
In this paper, an integral sliding mode control algorithm based on reinforcement learning is proposed for underwater vehicle depth determination control system. Since it is difficult for nonlinear continuous systems to track time-varying trajectories, the optimal tracking problem is transformed into a nonlinear time invariant optimal control problem by introducing a new state variable. The HJB equation of nonlinear systems is solved by adaptive dynamic programming (ADP) algorithm to find an approximate optimal strategy. Combined with integral sliding mode control, an approximate optimal sliding mode controller is designed. In addition, the Lyapunov equation is used to verify that the control strategy proposed in this paper can ensure that the tracking error of the system converges to zero gradually, and the error is also verified in a small range. Finally, the effectiveness of the algorithm is verified by simulation experiments, which enhances the anti-interference and robustness of the underwater robot in the depth control direction.