Decentralized Time-Varying Formation with Dynamic Leader Selection

This paper presents a novel approach to time-varying formation for multi-agent systems for operation in an unknown environment. The time-varying formation uses a leader-follower system with a dynamic leader selection process. Leaders are determined by calculating the center of formation, and determining the position of a predefined goal point relative to the center. By incorporating a distributed protocol in which each agent calculates an estimate of the center based on incomplete information, the agents are able to determine their roles without requiring a central control system. A role negotiation process is developed for resolving edge cases. Two sets of simulations are conducted; the first set showing the capabilities and limitations of the center of formation estimation method, and the second set showcasing a team of robots navigating a series of environments using the proposed time-varying formation algorithm.