Ant-Inspired Dynamic Task Allocation via Gossiping

Abstract

© Springer International Publishing AG 2017. We study the distributed task allocation problem in multi-agent systems, where each agent selects a task in such a way that, collectively, they achieve a proper global task allocation. In this paper, inspired by specialization on division of labor in ant colonies, we propose several scalable and efficient algorithms to dynamically allocate the agents as the task demands change. The algorithms have their own pros and cons, with respect to (1) how fast they react to dynamic demands change, (2) how many agents need to switch tasks, (3) whether extra agents are needed, and (4) whether they are resilient to faults.