The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec)
Online ISSN : 2424-3124
2023
Session ID : 2P2-C25
Conference information

Acquisition of Action Models for Autonomous Navigation of Mobile Robots Using Deep Reinforcement Learning Based on Sharing Positions in Robot Network
*Yuki MATSUBARAKazuyuki MORIOKA
Author information
CONFERENCE PROCEEDINGS RESTRICTED ACCESS

Details
Abstract

The purpose of this study is to develop an autonomous navigation system for mobile robots using deep reinforcement learning. Especially, this study assumes that multiple mobile robots are connected to the network and the positions of networked robots are shared each other. In this paper, deep reinforcement learning using the positions of the other robots is performed. Then, action models that can avoid the other robots and reach the destinations are obtained. Unity is used as a simulation environment for learning and an agent and the other mobile robots are placed in the environments. As the training result, the action model that the agent can arrive at goals while avoiding the other robots are obtained.

Content from these authors
© 2023 The Japan Society of Mechanical Engineers
Previous article Next article
feedback
Top