Constructing Large Scale Cooperative Multi-Agent Systems from Semantic P2P Networks

Abstract

When we construct “Smarter Planet” by gathering all information and cooperating all objects of physical and digital world in the era of Internet of Things, the agents as hubs and gateways play very important roles. As the Network becomes huge scale, large scale multi-agent systems become more and more important. Multi-agent systems play important roles in various applications. However, large scale multi-agent systems encounter many big issues such as scalability and inter-operability. This chapter presents a different approach of large scale multiagent system architecture based on semantic P2P Network- Virtual Hierarchical Tree Grid Organizations (VIRGO) in mathematic terms. Other than unstructured and DHT-based structured P2P networks, VIRGO keeps the semantic meanings of the nodes according to their roles in the communities. In VIRGO approach, the agents are identified as domain names classified by the semantic meaning of roles in the organizations. The agents construct the VIRGO network by joining the virtual groups with the same names as their domains.We proof by mathematics that in this approach each agent has at least one path to communicate with any other agent and the performance is effective as maximin hops is less than log(N), where N is total number of agents. The performance bottle-neck of tree-like topology can be avoided by cache policies. Here, we also introduce a SQL-like language for inter-operation among agents. The application cases of large scale multi agent systems based on semantic P2P networks are also described.