Proceedings of the 2005 IEEE Congress on Evolutionary Computation. Multi-Agent Cooperation Using the Ant Algorithm with Variable Pheromone Placement

The Ant Algorithm was created by examining real life ant colonies and developing an algorithm to use the concept of “stigmergy ” to approach multi-agent problems with distributed control. As agents work on tasks, more agents attempt difficult tasks. Task deadlock occurs when agents attempt impossible tasks indefinitely. Previous research avoids task deadlock through adaptive attenuation factors. This research investigates increasing algorithm effectiveness through variable pheromone placement. Results of computational experiments are presented demonstrating the increased effectiveness of the new algorithm. Key Words: multi-agent system, cooperation, ant algorithm, variable pheromone placemen