Answers are grounded in this paper and cite their sources.
Swarms of drones and other mobile agents often rely on a few informed members to keep the group aligned, yet how many are needed, and how the communication rule affects the result, is rarely quantified under realistic conditions. We addressed this with a Vicsek model of 3000 agents in which informed agents (heading noise 0.10) are mixed with uninformed agents (noise 0.50), with density, perception range and speed calibrated to field data from starling flocks. Simulations ran on graphics processing units using a memory-optimized periodic cell list and an exact k-nearest-neighbour search, both checked against brute force at start-up. Critical points were located by an adaptive Binder-cumulant bisection that needed 7–8 runs instead of the 25–41 of a uniform grid. About 10% informed agents ordered the swarm, and finite-size scaling over 1000–8000 agents placed the threshold for large swarms near 11%. With range-based communication the starling parameters lay at the order–disorder transition. Communication with the seven nearest neighbours made order 2.6 times less sensitive to density, and density independent when speed was scaled with neighbour spacing, but required about nine neighbours. These results give design rules for leader–follower swarms: minimizing informed agents and communication links, and scaling speed for robustness.
No takes yet. Share an insight, caveat, or question.
Deniz et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: