Dynamic Networks and Distributed Problem-Solving

Download

PDF (Conference Paper) - Published Version Download (656Kb)

Description/Abstract

The effect of dynamic networks on distributed problem-solving was examined using a multi-agent simulation environment. Synthetic agents were tasked with the problem of finding optimal solutions to a specific design problem, and they were allowed to communicate the results of their search efforts to other agents via a dynamically-evolving communication network structure. The growth of the network was determined by two parameters. One parameter determined the rate at which the network structure emerged, while the second determined the point at which the first network link was formed. Together, these parameters produced a reliable effect on collective problem-solving performance. Firstly, performance was negatively affected by the rate of network growth, with faster growth rates producing poorer performance. Secondly, performance was improved by introducing longer initial delay periods into the network formation process, a manipulation which also served to attenuate the decline in performance seen with increasing network growth rates. Of particular interest, the study found that networks with dynamic, constructive topologies delivered a better profile of performance relative to networks with fixed, static topologies. The results are discussed in relation to our understanding of how military coalition communication networks may affect performance outcomes in distributed problem-solving environments.