Dynamic expansion of formation using local information

This work presents a behavior-based, decentralized approach for agent formations. The formation dynamically grows from a single agent to a large number of agents. The shape is such that it always makes a regular polygon and hence appears to be a circular in the end. This formation could also be expanded to other shapes. We achieve this global behavior of making dynamic formation in a group of agents by using only local sensing and interaction. The algorithm is very simple and it solves the problem with agents having camera, sonar, laser and a radio link. The basic idea is to keep immediate leader or friend in its view by panning the camera at the angle required for making a shape, and at a fixed distance so as to maintain the formation. While in the group how to keep the formation and change it with the addition of a new agent is the main issue we are dealing here which has not been done according to our information. We test our algorithm on simulation as well as on some physical agents and discuss the issues related to physical systems.