Agent network topology and complexity

In this paper, we examine two multi-agent based representations of SATs and further experimentally study the topologies of resulting agent networks. We show that different representations will make agent networks manifest different topologies. In one presentation, the resulting agent network show obviously small-world topologies. Generally speaking, a small-world topology will computationally harden a search process. Therefore, we propose a guiding design principle that to solve a search problem by a multi-agent system, it should avoid having small-worlds among agents and it should maintain balanced intra- and inter- agent computational complexity. Categories and Subject Descriptors I.2.11 [Artificial Intelligence]: Distributed Artificial Intelligence—Multiagent systems General Terms Experimentation, Measurement Keywords Agent network, Small-world, Computational complexity

• Type: Conference
• Year: 2003
• Where: ATAL
• Authors: Xiaolong Jin, Jiming Liu
• Keywords: Agent Network, ATAL 2003, Multi-agent Based Representations, Small-world Topologies