Session: 08-01-02 Maneuvering Simulation

Paper Number: 124022

124022 - Underactuated Unmanned Ship Trajectory Tracking Control Based on Lstm and Ddpg Under Environmental Disturbance 

Underactuated Unmanned Ship Trajectory Tracking Control Based on LSTM and DDPG under Environmental Disturbance

The trajectory tracking of unmanned ships is an important part of their autonomous navigation. Due to the underactuated nature of unmanned ships and the unknown environmental disturbances encountered during their navigation, it is difficult to achieve trajectory tracking of unmanned ships. To enhance the computational performance of unmanned ship trajectory tracking and save computational resources. Considering the unknown environmental disturbances on unmanned ships and rudder speed limitations. This article establishes a kinematic and dynamic model of a three degree of freedom underactuated unmanned ship. A unmanned ship trajectory tracking control model with event triggering mechanism was established based on LSTM neural network and reinforcement learning DDPG algorithm. When an event is triggered, the LSTM network is used as the actor network of the DDPG algorithm, outputting rudder angle and speed control commands to achieve trajectory tracking of unmanned ships. Finally, numerical simulation experiments were conducted, and the experimental results verified the effectiveness of the trajectory tracking control algorithm and demonstrated that the trajectory tracking model with event triggering mechanism can effectively enhance computational performance while achieving trajectory tracking targets.

Keywords: Environmental disturbances, Trajectory tracking, LSTM neural network, Reinforcement learning, Event triggering mechanism

Presenting Author: Ziming Wang Wuhan University of Technology

Presenting Author Biography: Family name: Wang
Given name: Ziming
Company: Wuhan University Of Technology
Department: School of Ship and Energy Power Engineering
Research direction: Unmanned ships, trajectory planning, trajectory tracking

Authors:

Ziming Wang Wuhan University of Technology
Shunhuai Chen Wuhan University Of Technology