Dual-pheromone stigmergic framework for decentralized swarm search and rescue

Abstract Robot teams search large areas faster by spreading out and avoiding repeated visits. Centralized planners coordinate this by assigning regions using information about robot positions and progress. However, maintaining these assignments becomes difficult when communication is unreliable. Decentralized control reduces this dependence, but robots lack shared records of where others have searched. We present STIGSAR, which uses stigmergy: robots leave environmental traces that others can sense. Temporary repulsive traces discourage revisits. When one robot discovers a target, it deposits a persistent attractive trace to guide others toward it. In simulation, STIGSAR approached centralized coverage performance and made 7–12× fewer visits per covered cell than random walk. Compared with single-trace variants that completed every mission, STIGSAR reduced median normalized mission cost by 38–45%. Robotarium experiments demonstrated recruitment with virtual traces. RoboMaster robots deposited and sensed physical repulsive trails, continuing search after robot loss without direct communication or online task allocation.