PISC: Pheromone-inspired Self-organized Collective Cooperation

Pheromones enable decentralized coordination in collective systems, yet their role in shaping cooperation thresholds, organizational scales, and cognitive strategies remains poorly understood. We propose PISC, a pheromone-inspired model for multi-agent cooperative load transport that integrates search, recruitment, and carrying phases through dual virtual pheromone fields. Through two-dimensional simulations with 150 agents navigating obstacles, we systematically investigate how cooperation thresholds, perception radius, and pheromone parameters influence task completion time, collision frequency, and transport velocity. Results reveal that intermediate cooperation thresholds optimize the trade-off between traction benefits and collision costs, while excessively large organizational scales degrade efficiency across all phases despite varying optimal scales per phase. Dual-field pheromone coordination substantially enhances overall performance and stability, whereas single-field pheromone guidance yields localized benefits but introduces system-wide instability through agent dropouts and reconfigurations. Ablation experiments confirm that synergistic pheromone-mediated recruitment and transport mechanisms are essential for robust collective behavior. Comparative benchmarking against artificial potential field and occlusion-based visual servoing methods further demonstrates that PISC's dual pheromone fields provide spatial recruitment organization and continuous directional guidance that jointly yield superior task success rates and transport efficiency in obstacle-cluttered environments. These findings establish quantitative principles for designing scalable swarm intelligence systems under resource constraints and communication limitations.