|
Abstract : |
Abstract. This paper addresses issues in developing open multiagent systems, in which it is easy to expand the functionality by adding new agents with new capabilities, and which facilitate interoperability with other agent systems. We argue that an open multiagent system should define the following support elements for agent communication: 1. A common set of speech acts to define the types of messages that an agent might send to another agent. 2. A common service ontology by which the agents can describe their capabilities to each other, and reason about which agents have the capabilities needed to execute specific tasks. 3. A common set of prescriptive conversation policies to define the acceptable exchanges of messages between agents. In addition to the above, we also discuss the utility of having a matchmaking agent that can reason over agent capabilities to recommend agents for specific tasks, where the capabilities and requirements are defined using a common service ontology. This ensures that the semantics of matching agent capabilities to task requirements remains the same across the multiagent system. 1, |
