Evaluation of dynamic task allocation strategies for distributed information systems

Presents the results of an evaluation of several agent-based dynamic task allocation strategies which we have developed for information systems spread over the Internet or other wide area networks (WANs). These strategies are based on the cooperation of agents organized in one community of reactive agents and another community of allocator agents, and on the learning skills of the allocator agents. Reactive agents represent the services available on WANs such as Internet. Allocator agents are able to cooperate and to learn about the states of the reactive agents. We compare our multi-agent-based strategies with negotiation-based dynamic task allocation strategies, by simulation. The results of this evaluation show that our strategies reduce the amount of message sending