An Adaptive Multi-Agent System: Genetic Approach

Actual software systems are developed for increasing complex applications. This complexity implies development of systems able to react to events and to, carry out organizational changes. The MAS technology gives interest solutions for that. Self-organization is an interesting capacity for this type of system, it makes it possible to face the dynamic of its environment. We are interested to study how genetic algorithms can help emergence of organizational structures in adaptive multiagent systems. We propose a multiagent system evolving in a dynamic environment. An agent supervisor reorganizes the task agents when it observes disturbances of the system. The supervisor agent applies a genetic algorithm to modify the cooperation beliefs of the task agents in order to obtain a cooperative organisation. The system evaluation is done according to a cooperation criterion and an external observer appreciation of the system