Virtual Testing and Policy Deployment Framework for Autonomous Navigation of an Unmanned Ground Vehicle Using Reinforcement Learning

摘要

The use of deep reinforcement learning (DRL) as a framework for training a mobile robot to perform optimal navigation in an unfamiliar environment is a suitable choice for implementing AI with real-time robotic systems. In this study, the environment and surrounding obstacles of an Ackermann-steered UGV are reconstructed into a virtual setting for training the UGV to centrally learn the optimal route (guidance actions to be taken at any given state) towards a desired goal position using Multi-Agent Virtual Exploration in Deep Q-Learning (MVEDQL) for various model configurations. The trained model policies are to be transferred to a physical vehicle and compared based on their individual effectiveness for performing autonomous waypoint navigation. Prior to incorporating the learned model with the physical UGV for testing, this paper outlines the development of a GUI application to provide an interface for remotely deploying the vehicle and a virtual reality framework reconstruction of the training environment to assist safely testing the system using the reinforcement learning model.