基于DDPG算法的海上无人救援技术研究

针对海上无人救援过程中遇险目标的漂移及如何快速靠近的问题,提出一种基于深度强化学习理论的目标追踪算法,使无人搜救船在与环境交互的过程中学习到自主驾驶追踪漂移遇险目标的最优驾驶决策。在SART的辅助下,通过自主学习能够使搜救船以最短的时间追踪到漂移遇险目标。在Gazebo物理仿真器中建立三维仿真环境,基于ROS系统分别设计直线漂移轨迹和不规则漂移轨迹仿真实验,通过多次自主学习训练,验证所提方法的有效性。

MARINE UNMANNED RESCUE TECHNOLOGY BASED ON DDPG ALGORITHM

Aiming at the problem of drifting distress target and the way of approaching quickly in the process of unmanned rescue at sea,a target tracking algorithm based on theory of deep reinforcement learning is proposed,which makes unmanned rescue vessel learn to autonomous driving to track drift target optimal decision during the interaction with environment.With the assistance of SART,the vessel got close to the drift distress target in shortest time through self-learning.A three-dimensional simulation environment was established in the Gazebo physics simulator.The simulation experiments of linear drift trajectory and irregular drift trajectory were designed respectively based on ROS.The effectiveness of the proposed method is verified through multiple independent learning and training.