Robotic Swarm Dispersion Using Gradient Descent Algorithm

Abstract

Robotic swarms can be used in applications ranging from mapping and exploring to sensing, emergency relief, and many others. The increasing sizes of these swarms makes it necessary to use simple inexpensive individual robots to keep a reasonable total cost. This usually leads to robots with very limited capabilities. With these limitations, swarm designers must rely on software to enable the robots to efficiently perform essential tasks such as deployment, congregation, and navigation. This paper addresses the particular problem of robotic swarm dispersion. We present and test a method that uses the strength of the robots' radio signals and the gradient descent algorithm to achieve an efficient swarm dispersion. The results show that, with swarms of 3, 5, 7, and 10 robots, the suggested method can consistently increase the average distance from a robot to the center of the swarm and consequently increase the swarm's coverage area.