A Survey of Path Planning Algorithms Based on Deep Reinforcement Learning

Traditional path planning methods have obvious limitations in complex and changing environments. Firstly, the shortcomings of these traditional methods are discussed, and then deep reinforcement learning as a new solution is introduced. The principles, advantages and disadvantages of 3 deep reinforcement learning methods including value function based, strategy based and hybrid value based strategies, as well as their representative research results in various application fields in recent years are summarized. The representative algorithms are tested on a unified platform, and actual comparative analysis is performed. Finally, the challenges and research prospects of path planning techniques based on deep reinforcement learning are summarized.