Study of UAV Path Planning Problem Based on DQN and Artificial Potential Field Method
Jiawei Wang, Guangyu Lei, Jiandong Zhang · 2023
UAV path planning has become increasingly important, and reinforcement learning is one of the most rapidly developing fields of artificial intelligence, and studies have been conducted to apply reinforcement learning to UAV path planning. This paper firstly introduces the basic principle of DQN algorithm, and then improves the reward function of the DQN algorithm by combining with the artificial potential field method, and introduces direction reward to solve the problem of sparse reward in the DQN algorithm, and finally further improves the network performance. The article compares the performance of the improved DQN algorithm with that of the DQN algorithm through experimental simulations and finally finds that the DQN network combined with the artificial potential field method has a significant performance improvement compared with the conventional DQN network, which further verifies the effectiveness of the proposed model.