Author(s)

M.G.M. Madden & P.J. Nolan

Abstract

Reinforcement learning is a form of artificial intelligence in which an agent acquires and improves skills based on receiving positive and negative rewards when it performs actions within an environment. This paper describes a system which uses an extended reinforcement learning algorithm to generate reactive control strategies. It is applied to the control of a vehicle in a simulated traffic environment. 1 Reinforcement Learning 1.1 Introduction Reinforcement learning is an artificial intelligence methodology whereby an autonomous agent, the learner, can acquire and improve skills within an envi- ronment without having an explicit teacher. It has been defined by Sutton [12] as the learning of a mapping from situations to actions so as to maximize a reward or reinforcement signal. The learner i

Keywords