基于前视声纳信息的AUV局部路径规划研究

Study of Local Path Planning Based on Forward Looking Sonar for AUV

针对水下环境的不确定性,建立了前视声纳的视域模型。主要采用强化学习的方法对自治水下机器人(AUV)进行控制和决策,综合Q学习算法、BP神经网络法、人工势场法对AUV进行局部路径规划。在AUV与环境的试错交互中,借助于来自成功与失败经验的奖励和惩罚值,不断改进水下机器人的自治能力。并设计了AUV局部路径规划器,实现AUV在不确定环境下的避障任务。半实物仿真证明了算法的可行性与可靠性。

The visual model of Forward looking sonar is built according to uncertainty of underwater environment. The reinforcement learning is adopted to control and decision for Autonomous Underwater Vehicles (AUV), and Q- learning , BP neural net, artificial potential is integrated to local plan for AUV. The automation ability is promoted by success and fail experience through trial error with the environment. The local path planner is designed, and the obstacle avoidance task is realized. The feasibility and reliability are verified through semi- physical simulation.