Adaptable Decentralized Task Allocation of Swarm Agents

Scalable task allocation in dynamic real-world domains often requires efficient, robust, decentralized, and adaptable approaches. Response threshold reinforcement is a biologically-inspired model of probabilistic action that has been shown to lead to efficient task allocation among swarm agents that do not reason or communicate, making it a highly scalable and low cost solution. The model leads agents to specialize, resulting in reduced costs of interference and task switching, as well as to improved efficiency and adaptability to initially unknown environments. While initial specialization of this and other models is investigated in much of existing literature, subsequent re-adaptation to domain changes is seldom verified. Our goal is to investigate the robustness of response threshold reinforcement to various environmental changes, as well as to compare this model to other decentralized approaches.