Research on performance comparison of patrol path planning techniques for mobile robots in nuclear power plants

Abstract

Path planning, as a basic problem of mobile robots, is important in the application of industrial patrol robots. This paper takes the nuclear power plant as an example and solves the problem of multi-target patrol path of patrol robot. Firstly, this paper processes multi-target points applying Euler's formula to obtain a reasonable order of patrol target points. Then three path planning algorithms, A* algorithm based on graph search, RRT* algorithm based on sampling and Q-learning algorithm based on reinforcement learning, are applied on path planning, and combined with Minimal-Jerk algorithm for optimizing trajectories. The performance of the results is finally compared using two evaluation metrics. The acceleration variance and route analysis in planar images are used as evaluation metrics in this paper. It is considered that the smaller the acceleration variance, the more smoothly the robot can move. The gentler the route is, the more effective the robot movement is.