Adaptive foraging in robotic swarms

Foraging is a canonical problem in robotics as it can be used to represent many other tasks. Bucket-brigading is a known control strategy for reducing interference in large swarms of robots. In this thesis, we examine the foraging problem and bucket-brigading solution from the perspective of the spatial distributions of robots and pucks in the foraging task. We allow these distributions to inform the robots' foraging behavior. We give a variation on bucket-brigading---adaptive ranging---in which the controller adapts to varying and nonuniform robot distributions. We provide a second variation---relocation---in which the swarm adapts to varying and nonuniform puck distributions. We describe a custom swarm foraging simulation tool, with which we show that these enhancements outperform the conventional bucket-brigading approach. We discuss robotic foraging in the context of the ideal free distribution from behavioral ecology, and propose a statistical test for measuring how well the distributions are matched.