Adaptive Collective Responses to Local Stimuli in Anonymous Dynamic Networks

The work proposes a framework where many simple agents collectively change “phase” in programmable matter when local stimuli appear or disappear. Each agent only detects stimuli nearby, and agents communicate by passing tokens along edges in a graph to coordinate transitions between an “aware” state (stimuli present) and an “unaware” state (stimuli absent). The authors introduce an Adaptive Stimuli Algorithm designed to remain robust even when multiple stimulus-related message waves compete, including cases described as possibly adversarial. The algorithm also supports graphs whose connections (edges) can be reconfigured over time in a controlled way. As an example application, the framework is used for a foraging task where food sources can be discovered, removed, or shifted at arbitrary times. The agents are intended to self-organize so that if food stays long enough they gather into a single large component around it, but if no food has existed recently they switch to a search phase and spread out to look for new food; the process is described as indefinitely repeatable.