An evaluation of quorum sensing mechanisms in collective value-sensitive site selection

摘要

Discrete collective decision making strategies enable robotic swarms to choose the best of a finite number of opinions. Future targeting and foraging problems will require swarms to make a site selection decision and unanimously respond to that decision. Collective value-sensitive decision models are a promising approach for swarm site selection, but current models do not ensure consensus. An existing quorum sensing strategy enables a swarm to alter collective behavior after a decision, but has not been evaluated under local communication restrictions. This paper combines these strategies to enable a swarm to detect when a majority of agents support a site and cease deliberation in order to rapidly build a consensus. Two quorum detecting rules are evaluated: a majority rule and a k-unanimity rule. Each rule relies on message queues that either persist throughout a decision problem, or are episodic and cleared each time an agent changes state. The proposed models are shown, in simulation, to sense collective decisions without decreasing the accuracy of the underlying decision strategy. Significantly, two models accurately detect decisions without increasing the original strategy's observed decision time.