Building the Matrix: Comparing Methods of Communication and Swarm Structure Dynamics in Evolved Autonomous Swarms

Presenter: Joshua Rubini, CS, UI

Abstract:

Current research shows that for many problems, fully autonomous agents with the ability to pass messages to individuals in a local neighborhood are capable of being evolved to solve a diverse range of problems. However, this contrasts with many instances of natural evolution in which hierarchical control systems are found, particularly among more complex organisms such as many mammal and bird species. The "fully distributed swarm" suffers from three distinct problems: time-critical event handling, information staleness, and/or complex data mining, particularly in complex, dynamic problem spaces. This research attempts to show whether hierarchical communication and control structures can help overcome these problems.