Automatically tracking and analyzing the behavior of live insect colonies

Abstract

We introduce the study of {\it live} social insect colonies as a relevant and exciting domain for the development and application of multi-agent systems modeling tools. Social insects provide a rich source of {\it traceable} social behavior for testing multi-agent tracking, prediction and modeling algorithms. An additional benefit of this research is the potential for contributions to experimental biology --- the principled techniques developed for analyzing artificial multi-agent systems can be applied to advance the state of knowledge of insect behavior. We contribute a novel machine vision system that addresses the challenge of tracking hundreds of small animals simultaneously. Fast color-based tracking is combined with movement-based tracking to locate ants in a real-time video stream. We also introduce new methods for analyzing the spatial activity of ant colonies. The system was validated in experiments with laboratory colonies of {\it Camponotus festinatus} and several example analyses of the colonies' spatial behavior are provided.