|Ph.D Student||Barel Ariel|
|Subject||Gathering and Steering Swarms|
|Department||Department of Autonomous Systems and Robotics||Supervisor||Professor Alfred Bruckstein|
This thesis presents several results concerning multi agent systems. The agents in a swarm are assumed to be identical, anonymous (i.e. indistinguishable) and simple in the sense of having little or no memory (i.e. oblivious) and with limited sensing capabilities. First a critical survey of previous results concerning gathering algorithms is presented. Then a new paradigm of simple sensors is introduced. Here agents that randomly select their motion direction and can only detect the presence of other agents behind them are shown to be able to gather using a simple motion rule, allowing the agent to move forward only while no neighbouring agents are behind it. We here estimate the expected gathering time for the process and show in extensive simulations that the theoretical upper bound indeed holds for the random gathering process. After gathering is achieved the probabilistic motion rule causes the centroid of the cohesive swarm to perform a random-walk in the plane. The third part of the thesis introduces a novel idea of steering and guiding a swarm based on simple external control signals that is based on monitoring the fluctuations of swarm centroids and stopping the motion of all agents when the random movement is in wrong directions. This causes the swarm to perform a biased walk drifting in the desirable direction, without the need for the agents to know where their swarm is heading.