Path planning for indoor Mobile robot based on deep learning

Abstract

This paper aims to give an optimal path planning of a mobile robot in a known indoor environment. An algorithm based on deep learning, ray tracing algorithm, waiting rule, and Rapidly-exploring Random Tree is proposed to solve the problem of obstacle avoidance and path planning. Firstly, GoogLeNet is used to classify obstacles. Here, it helps to distinguish between static and dynamic obstacles. Secondly, for the static obstacle avoidance, the ray tracing algorithm is proposed to avoid the obstacles which are identified by GoogLeNet. And for the dynamic obstacle avoidance, this paper proposed a waiting rule for dynamic obstacle avoidance. Thirdly, the RRT method plans a path from the start point to the goal point. The novelty of this paper is that the type of obstacles is distinguished by deep learning, and the two different kinds of obstacles used ray tracing algorithm and waiting rule to avoid obstacles collision, respectively. In experimental results, rapidly-exploring random tree is compared with genetic algorithm, and particle swarm optimization method in static environment, and it is compared with artificial potential field approach in dynamic environment. Experiments are carried out in the two-dimensional environment and successfully applied to the path planning of mobile robots in multi-obstacles environment, and they prove the feasibility of the algorithm.