Multi-Agent Machine Learning :A Reinforcement Approach

Abstract

Learning is a process of improving the performance of a system based on its past experiences. This method intervenes when the problem seems too complicated to solve in real time, or when it seems impossible to solve the problem in a classic way. As an example of learning methods we cite reinforcement learning. This method of learning is often used in the field of robotics and agents. It aims to determine a control law for a mobile robot or agent in an unknown environment. This kind of technique applies when it is assumed that the only information on the quality of the actions performed by the robot is a scalar signal that presents a reward or a punishment, the learning procedure aims to improve the choice of actions in order to maximize the rewards. One of the most used algorithms for solving this learning problem is the Q-Learning algorithm that is based on the Q-Function, and to ensure the generation of this last function and the proper functioning of the learning system. , the action performed by the mobile robot in its environment is ensured by the use of a selection function, this action is evaluated by rewards and punishments