Multi-agent learning and control system using ants colony for packet scheduling in routers

Abstract

This paper describes a novel method of achieving packet scheduling in several routers of network, in order to optimize the end to end delay. We use a multi-agent system to model this problem, where each agent of this system tries to optimize the local scheduling and through a communication with each other, attempts to make global coordination in order to optimize the total scheduling. The communication between agents is done by mobile agents like ants colony. A pheromone-Q learning approach is presented in this paper, which consists to applying the standard Q-learning technique adapted to our architecture with a synthetic pheromone that acts as a communication medium speeding up the learning process of cooperating agents. © Springer-Verlag Berlin Heidelberg 2007.