A Review of Path Planning Methods for Mobile Robots

Path planning is an indispensable core problem in the autonomous navigation of mobile robots. Whether a safe and efficient trajectory can be planned directly determines the usability of robots in scenarios such as warehousing, delivery, medical assistance and autonomous driving. Scholars at home and abroad have accumulated abundant research results on this problem. Nevertheless, existing methods have respective trade-offs in application scope and performance limits, and no universally recognized optimal solution has been reached so far. This paper sorts out relevant methods. These methods are divided into two categories: traditional methods and intelligent methods. The traditional category includes graph search algorithms (Dijkstra, A*, D*), artificial potential field methods and sampling-based planning algorithms (PRM, RRT, RRT*). The intelligent category covers intelligent optimization methods represented by genetic algorithms, particle swarm optimization and ant colony optimization, as well as two data-driven approaches: reinforcement learning and deep learning. The core idea, typical application scenarios and limitations of each type of method are elaborated one by one. Finally, a horizontal comparison table is provided, and the author puts forward observations on several noteworthy research directions including hybrid planning frameworks, multi-robot coordination and uncertainty modelling.