A study on the optimal route design considering time of mobile robot using recurrent neural network and reinforcement learning

Abstract

Recently, the robots market is growing rapidly, and robots are being applied in various industrial fields. In the future, robots will work in more complex and diverse environments. For example, a robot can perform one or more tasks and collaborate with people or other robots. In this situation, the path planning for the robots to perform their tasks efficiently is an important issue. In this study, we assume that the mobile robot performs one or more tasks, moves various places freely, and works with other robots. In this situation, if the path of the mobile robot is planned with the shortest path algorithm, waiting time may occur because the planned path is blocked by other robots. Sometimes it is possible to complete a task in a shorter time than returning or performing another task first. That is, the shortest path and the shortest path do not coincide with each other. The purpose of this study is to construct a network in which the mobile robot designs the shortest path planning considering shortest time by judging itself based on environment information and path planning information of other robots. For this purpose, a network is constructed using a recurrent neural network and reinforcement learning is used. We established the environment for network learning using the robot simulation program, V-Rep. We compare the effects of various network structures and select network models that meet the purpose. In the future work, we will try to prove the effect of network by comparing existing algorithm and network.