Dynamically Refining the Task Models of Agents in a Multi-Agent System: A Less Communication Intensive Approach

Action selection of agents in a given environment is governed by the utility they gained from the resulting environment state. In a multi-agent system (MAS), the autonomy of agents can lead to a situation for multiple agents to perform similar or identical tasks if they independently try to maximize self utilities. Therefore, it is important to dynamically identify the autonomous capabilities of the agents in the MAS and modify them to avoid any task overlapping. This paper presents a less communication intensive corporation and coordination strategy for refining the task models of agents on the fly using a collective reasoning process.