Autonomous vehicle controls using Reinforcement Learning

Abstract

In this paper, we explore a reinforcement learning algorithm to train an agent to drive a vehicle in the OpenAI Gym environment called CarRacing-v0. Gym is an open-source repository created by OpenAI which provides a toolkit for developing and comparing various Reinforcement learning algorithms. The gym library is a collection of environments that can be used to work out reinforcement learning algorithms. Learning to drive in the CarRacing-v0 environment is challenging since it requires the agent to finish the continuous control task by learning from pixels. To tackle this challenging problem, we explored an approach called Deep Q Learning. In this paper we further demonstrate a method to train the agent which learns from raw pixels without providing any hand-crafted features. Some minor environment specific changes were made but the base agent was not provided any knowledge regarding car racing.