A Scoping review of Deep Reinforcement Learning methods in Visual Navigation

Reinforcement Learning (RL) is a subset of Machine Learning that trains an agent to make a series of decisions and take action by interacting directly with the environment. In this approach, the agent learns to attain the goal by the response from its action as rewards or punishment. Recent advances in reinforcement learning combined with deep learning methods have led to breakthrough research in solving many complex problems in the field of Artificial Intelligence. This paper presents recent literature on autonomous visual navigation of robots using Deep Reinforcement Learning (DRL) algorithms and methods. It also describes the algorithms evaluated, the environment used for implementation, and the policy applied to maximize the rewards earned by the agent. The paper concludes with a discussion of the new models created by various authors, their merits over the existing methods, and a briefing on further research.