Abstract:
Central place foraging is a problem domain which consists of finding and delivering resources situated throughout an unknown environment to a singular collection depot. Foraging behaviors are the primary benchmark application of swarm robotics, which is the study of the complex group behavior that emerges from the local interactions of many simple individuals. A common issue within central place foraging approaches is inter-robot interference, a significant detractor from scalable group performance. To address this problem we propose a novel technique for central place foraging, the Multimodal approach. This technique separates a preliminary search phase from collection behavior, locating all of the resources within the environment before any are picked up, storing and sharing these locations amongst all of the agents. This information is then used by the collecting agents in order to select resources in areas in which there are no other agents, mitigating the effect of interference. The application of this approach to various simulated problem formulations resulted in a significant performance increase as compared to a baseline approach. This lends to our conclusion that a separation of search and collection can lead to the incorporation of more advanced routing techniques that further improve the performance of the foraging task.