A Reinforcement Learning Approach for Shortest Path Navigation in Automated Guided Vehicles for Medical Assistance

Abstract

With the advent of tremendous innovation in the field of autonomous navigation, the day-to-day chores of humans has been simplified. Autonomous navigation is used in all types of vehicles including aerial, ground and underwater vehicles. They are being used in vast domains like surveillance, delivery, self driving cars, etc. In addition to this, they also find application in the defence sector where they play a crucial role through surveillance in the border. The real challenge for the autonomous vehicles lies in obstacle detection and avoiding it intelligently. Another challenge for experts in the autonomous navigation field is identifying the shortest path. Hence, we propose a novel framework to tackle the obstacle detection problem and to identify the shortest path using a reinforcement learning approach that could be helpful in delivering food and medicines to the old and the special need people. The autonomous vehicle is trained and tested in the grid environment on both Q learning and Double Q learning approach. The performance of the algorithms is evaluated, and the metrics are discussed in the chapter.