【摘要】 为解决无人机无模型路径规划的问题,提出一种环境信息未知情况下基于势函数(PF)奖赏的DQN路径规划方法。建立无人机在环境中的连续状态空间,将360°等分成若干个角度作为航向角建立无人机的动作空间,设计目标和障碍物对无人机的势函数奖赏,刻画不同动作对无人机的影响,并进行仿真实验。实验结果表明:PF-DQN算法能较好地实现无人机在环境信息未知下的无碰撞路径规划,且势函数奖赏能加快无人机路径规划网络的训练速度。更多还原

【Abstract】 In order to solve the problem of no-model path planning for UAV, a DQN path planning method based on potential function(PF) reward in the case of unknown environmental information is proposed. Establish a continuous state space of the drone in the environment, divide the 360° into several angles as the heading angle to establish the action space of the drone, design the target and obstacles to reward the potential function of the drone, and des cribe the difference more carefully. Carry out the simulation experiments. The results show that the PF-DQN algorithm can better realize the collision-free path planning of the UAV under the unknown environmental information, and the potential function rew ard can speed up the training speed of the UAV path planning network.更多还原