Communication Convention Formation in Large Multiagent Systems

A fundamental challenge in large-scale multiagent systems (MAS) is to enhance network-related dynamical processes such as establishing social convention in a decentralized fashion by regulating the behavior of the artificial autonomous agents. In my dissertation, I plan to study the specific problem of communication convention formation as a dynamic semiotic process over a large-scale MAS. In a semiotic process, a group of artificial agents collectively invent and negotiate about a shared language system that is used for communication. The landscape of the communication system will be captured through a hierarchy of interdependent networks where a semiotic network is built on top of an agent network. I propose a network theoretic investigation to determine its implications on the dynamic convention formation process and plan to develop a generalized theory of communication convention suitable for large-scale networked MAS. This theory will support predictions of system properties such as convergence dynamics across different temporal regions. Finally, an effective convention formation mechanism based on the generalized theory will be presented and evaluated for a real-world application domain.