Social Insects: A Domain for AI Research

Tucker Balch - Georgia Tech

Abstract

Social insect societies are wonderfully complex, robust and adaptive, even though they are composed of apparently simple individuals. Bees, for instance, are able to allocate resources (workers) to various tasks both inside the colony and outside, across areas of tens of square miles without a command structure. Task allocation is apparently accomplished as the workers follow simple stochastic rules mediated by millions of individual to individual interactions within the hive each day. At the BORG Lab we are applying machine vision, machine learning and other CS tools to the task of understanding social insect behavior. Our lab houses 3 ant colonies and 4 bee hives as subjects for our research. I will describe our work with these interesting creatures. This is work with Frank Dellaert, Zia Khan, Rohit Sharma, Andrew Stein, Adam Feldman, and Rande Shern. Thanks also to Manuela Veloso and the MultiRobot Lab.

Back to the Main Page