Single view depth estimation based formation control of robotic swarms: Fundamental design and analysis

Abstract

This paper presents a practical formation motion control scheme for robotic swarms based on single view depth estimation. The single view depth estimation for each robot in the swarm is performed using a single non-sophisticated camera on the agent, and the prior information about the heights of the robots and other objects in the estimation. Here, a non-sophisticated camera means one that has limited field of view and limited resolution. First, the vision mechanism is analyzed and a single view depth estimation scheme is designed. Then a set of decentralized control laws, to be incorporated with depth (distance) estimation scheme, are introduced to move the robotic swarm in formation from an arbitrary initial position to an arbitrary final position without deforming the formation shape. The robots do not have any global positioning sensors, and they do not communicate with each other. The stability and performance of the overall system are analyzed mathematically.