Omnibot: A Scalable Vision-Based Robot Swarm Platform

Abstract

Introducing vision sensing into swarms presents three challenges for developing robot platforms. First, the vision system requires a wide field of view to perceive surrounding robots. Second, vision algorithms demand high computational power, which poses a challenge for real-time vision-in-the-loop simulation. Third, as the swarm scale increases, managing the system becomes increasingly demanding. The main contribution of this paper is the development of a novel mobile robot swarm platform to overcome these challenges. 1) Each robot features a 360-degree omnidirectional vision system comprising four cameras, allowing each robot to detect and interact with the surrounding robots. 2) It has a novel ROS-based distributed swarm simulation system, which can effectively utilize the onboard computational resources of multiple robots to achieve parallel vision-in-the-loop simulation. 3) It features a novel swarm management system that allows real-time monitoring and debugging of multiple robots. These innovative designs provide a novel swarm platform that can facilitate the study of versatile vision-based swarm tasks.