Using Meta-agents for Multi-agents in Networks

摘要

In this paper, we propose an approach using meta-agents for monitoring multi-agents and controlling their behaviour while moving between states in a network. The meta-agents are built on the agents in the system and then used to inspect the agents' behaviour when reaching a result as well as perceiving the reason for that result. Each meta-agent can comprise everything from one agent to dozen or more agents depending on the task assigned to the system. The benefit of using meta-agents for a multi-agent system is the ability to provide the fastest way between nodes under given circumstances and to handle a vast number of nodes in graphs and networks such as travelling salesman problem (TSP).