Scalability is often lauded as one of the advantages of decentralized multi-robot systems and robot swarms. Theory and many experimental works predict that with increasing swarm density, we will observe a gradual decay of performance. In our work, we have taken a closer look at the scalability of robot swarms in different settings and we have noticed that the predicted decay does not always appear to be gradual. Instead, the performance splits into two phases, potentially causing catastrophic failures at near-optimal swarm densities. In this talk, I will provide an overview of our empirical and theoretical analyses of scalability behavior in robot swarms and the resulting considerations for the design of robot swarms.
Image created in DALL-E by Maria Ott.