Arian: A General Architecture for Advisable Agents

One of the most important issues in multi-agent systems is communication. The agents with more comprehensive information (which can be a human observer) advise other agents by sending messages. Similar agents may also communicate to transfer their experience. These messages are called advice. In this paper a general method, which is applicable for all agent architectures will be proposed for integrating advice. In this method advice is modeled by a case in a case-based reasoning system. It is the duty of the advice receiver to derive a case from each advice message. Each case is composed of a problem, which is circumstances that the advice will be used and a solution, which is the solution of the advice to that problem. In this paper a 2-layered architecture for refining the agent architecture and internal algorithms with the help of advice will be presented. Two main components of this system is a CBR which refines the parameters and the other is a translator, which translates messages to standard cases. An implemented example is presented to show how this proposed method works.