Purely vision-based collective movement of robots

Abstract Collective movement inspired by animal groups promises inherited benefits for robot swarms. However, while animals only rely on local senses, robots often use global information or explicit communication, introducing weaknesses to the swarm. To address these vulnerabilities, bio-inspired decentralized swarms have been a focus for decades. Yet, creating robots that move efficiently together using local sensory information remains an extraordinary challenge. Here, we present a decentralized, purely vision-based terrestrial swarm, where robots achieve polarized motion with highly effective collision avoidance exclusively through simple visual interactions. They compute everything on board based on their individual camera streams, without central processing or communication. Using robot experiments and agent-based simulations, we show that with this model, even with a strictly limited field of view and within confined spaces, ordered group motion can emerge. Our results offer a multitude of practical applications from hybrid societies to advanced vision-based robot swarms operating in ever-changing environments.