A UAV Path Planning Algorithm Based on GRU and DDPG

The stability of UAV path planning algorithms is crucial for ensuring efficient task completion. In response to the issue of poor training stability in deep reinforcement learning algorithms for UAV path planning tasks in high-dimensional and complex spaces, this paper proposes a UAV path planning algorithm combining GRU and DDPG. First, the UAV action space is designed based on the concept of Artificial Potential Fields (APF). Second, GRU is incorporated to endow the DDPG network with memory capability, enabling it to make decisions based on previous states. Additionally, a new reward mechanism is proposed to address the problem of sparse rewards in traditional algorithms, which leads to poor planning performance. Finally, the effectiveness of the proposed method is validated through simulation experiments. The results demonstrate that the proposed algorithm significantly enhances the convergence and stability of the learning process and improves the performance of path planning.