Deep Reinforcement Learning for Autonomous Vehicles - State of the Art

Abstract

Reinforcement learning is considered to be one of the strongest paradigms in the AI domain, which can be applied to teach machines how to behave through environmental interaction. Recently the concept of deep reinforcement learning (DRL) was introduced and was tested with success in games like Atari 2600 or Go, proving the capability to learn a good representation of the environment. At this moment there are few implementations of DRL in the autonomous driving field. In this paper, we present the state of the art in deep reinforcement learning paradigm highlighting the current achievements for autonomous driving vehicles.