Formation Control Algorithms With Limited or No Communication

Abstract

Formation control refers to a collective behaviour of multi-agent systems where individual agents come together to form a pattern, often geometric. These formations can enable multi-agent systems to function more effectively in a broad range of applications. Many formation control algorithms require centralized decision making, communication between agents or a centralized decision maker and other factors that increase per-agent cost and reduce the robustness and scalability of multi-agent systems. To this end, we introduce two algorithms that operate using local decision making and limited or no communication. The first algorithm is a communication-free and index-free algorithm based on polar indicator distributions. The second is a progressive assignment algorithm using limited, situated communication that deterministically assigns agents a position in the objective formation along a convex spiral directed path graph. We also present an extension of the second algorithm for 3-dimensional formation definitions. The first algorithm is demonstrated in a physical experiment using ground-based agents while the second one is simulated using micro air vehicles (MAVs) in a physics-based simulator.