Distributed Agent Search Efficiency

Abstract

Multi-agent systems where there exists a group of agents all working towards the same goal are gaining increased focus. One subset of these systems is groups of agents who have intermittent communication with one another. In an environment with limited communication the group of agents must act individually, while still completing their shared goal. Since each individual agent only knows about its experiences and those of other agents it meets, it is important that a system be developed to facilitate agent collaboration.

This thesis looks at one such application of a multi-agent system with limited agent communication where the agents are tasked with retrieving a set of objects – Distributed Agent Search. Within this application it is important that the agents act as a team even while not in constant contact with each other and that the search strategy used to retrieve the objects is the most efficient for the given environment. Through experimentation it was determined which of a set of developed search strategies performs most efficiently given different environment sizes, and varying communication.