Active Search Promotes the Collective Behavior of Underwater Robots With Limited Field of View

Abstract

Vision is a critical sensing modality for underwater robots. However, a limited field of view (FoV) significantly hampers vision-based distributed swarm control, making it difficult for the swarm to maintain cohesion and preventing the emergence of typical collective behaviors. In this study, we present a novel model to address this challenge. By incorporating an active search strategy for situations where no neighbors are detected within the FoV into a classical potential field-based collective model, the proposed approach improves swarm cohesion without compromising the system's capacity for emergent behaviors such as swarming, flocking, and circling. To gain a deeper understanding of the new model, we also investigate the collective behavior under different FoV and different group sizes. Theoretical analyses further elucidate the conditions for circling behavior, providing insights into the model's dynamics. Finally, underwater experiments with three physical robots validate the model's effectiveness in real-world scenarios with limited FoV, offering a new perspective on vision-based collective behavior in underwater robotic swarms.