A multi-agent system which intelligently behaves under incomplete information environment

摘要

A multi-agent system is a system which solves a problem using two or more agents. However, existing systems mostly assume complete knowledge of each agent and the ideal cooperation between agents. In this research, we characterize the qualitative side of the imperfect knowledge of an agent using "default knowledge", while the quantitative side using "confidentiability". Under this circumstance, we provide a multi-agent system in which each agent performs commonsense reasoning under incomplete knowledge. Moreover, we introduce the "reliability" to indicate the confidential relation between agents, and realize communications which reflect mental states of agents. To verify the usefulness of this system, we use an example of a commercial transaction, and compare it with the conventional systems.