Task Allocation Strategies: A Study with a Multi-Agents System in Fully Distributed Information Systems

In this paper we present some task allocation technics useful on Internet or other Wide Area Network (WAN). These technics are based on the cooperation of multi-agents organized in one community of Reactive agents and another of Allocator agents. Reactive agents represent services availables on the WAN such as Internet search engines. Allocator agents' task is allocate the queries that they receive from users. To take such a decision, Allocator agents cooperate sharing theirs informations about the state of Reactive agents. An Allocator agent is able to characterize the behaviour of Reactive agents by means of a knowledge model we developed. In order to facilitate the composition of complex services such as retrieval of multiple language documents, we also define another agent type we call Sub-contractor agent.