Integrating heterogeneous reasoning in a multi-agents system

The paper addresses the problem of integrating several heterogeneous reasoning models to solve a common problem. These reasoning models are implemented in a multi-agents system. Each agent represents one domain reasoning model. The agents cooperate because none of them is able to solve a given overall problem. We deal with the problem of heterogeneity by modeling cooperative reasoning which, an agent performs to contribute efficiency to the common problem solving. The most important point to develop is the architecture constructed to allow an agent to support the two levels of reasoning: cooperation and domain reasoning. This architecture is sufficiently flexible to: allow the increasing of the variety and the number of agents, deal with approaches developed by distributed artificial intelligence researchers, integrate new ones and be domain independent. Another issue concerns the mechanisms necessary for the agents to communicate despite of heterogeneous knowledge representations. A consensus representation is developed to allow each agent instantiating its local representation without any translation mechanisms.<>