Split over $n$ resource sharing problem: Are fewer capable agents better than many simpler ones?

A new paper presented at the 15th International Conference on Swarm Intelligence (ANTS 2026) tackles a fundamental question in multi-agent systems: when resources are limited, should you build a few highly capable agents or many simpler ones? The researchers—Karthik Soma, Mohamed S. Talamali, Genki Miyauchi, Giovanni Beltrame, Heiko Hamann, and Roderich Gross—formulate this as the 'split over n' resource sharing problem, where a group of n agents equally shares a common resource like monetary budget, computational power, or physical size.

Using a multi-agent coverage case study where each agent's disk-shaped footprint area scales as 1/n, the team's formal analysis reveals nuanced trade-offs. The initial coverage rate actually increases with more agents, but performance heavily depends on how agent speed scales with size. If speed decreases proportionally with agent radius, groups of all sizes perform equally well. However, if speed decreases proportionally with footprint area, a single larger agent outperforms any group of smaller ones. Computer simulations add another layer: resource splitting increases individual agent failure rates, suggesting that distributing resources too thinly can compromise system reliability. These findings provide a framework for engineers to determine the optimal level of distributiveness when designing multi-agent systems under tight resource constraints.