The Efficiency of Stigmergy-Based Collaborative Transport

This project investigates whether stigmergy alone can coordinate decentralized agents to perform collaborative transport without explicit communication. In a 2D grid world with obstacles, simple transport agents deposit and follow a digital pheromone field that encodes local cues about object locations and recruitment needs. Heavy objects require k simultaneous carriers before they can move, creating a coordination challenge that cannot be solved by individual exploration alone. The model uses purely local rules: agents latch to nearby objects, reinforce pheromone signals when additional carriers are needed, and disperse after delivery to continue searching. Objects move toward a designated drop-off region using local navigation and decentralized target selection to reduce congestion and prevent stacking.

The system is evaluated through repeated simulation trials across varied swarm sizes, obstacle densities, and carrier requirements. Primary performance metrics include delivery success rate, time from discovery to delivery, and redundancy (wasted effort from excess carriers or repeated recruitment). Results show that stigmergy can reliably enable cooperative transport across a range of conditions, but also exhibits trade-offs: increasing agent count generally improves delivery speed and reliability while raising redundancy and congestion pressures. By relating performance to the ratio between available agents and required carrying capacity, the experiments provide practical guidance for sizing swarms and tuning pheromone parameters in cluttered environments. Overall, the findings support stigmergy as a viable coordination mechanism for multi-agent transport tasks where direct communication is limited or unavailable.