Cooperation in Swarms of Robots without Communication

Swarm robotics aims to use a large group of relatively simple robots to solve tasks that can hardly be achieved by a single robot in the group. Compared to single robot systems with increased capability, a swarm robotic system may have advantages in robustness, flexibility and scalability. However, designing cooperative behaviors for a swarm robotic system is a challenging problem, especially when the robots may not have communication capabilities and thus only know local information. For a swarm of miniature mobile robots that cannot communicate explicitly, this thesis studies fully decentralized solutions of two problems. For the problem of cooperative transport, the thesis presents a strategy to push an object that is large enough to occlude the robots' perception of the goal of the transportation. For the problem of pattern formation, the thesis investigates algorithms based on the Brazil nut effect that can organize the swarm of robots into an annular formation. These problems are studied using physics-based computer simulations as well as experimental implementations based on the e-puck robotic platform. The simplicity of the solutions make them suitable for applications that require the individual robots to be as simple as possible. Example application scenarios could be micro robot swarms working in the human body.