In previous work, we developed a decentralized framework for formation
control by mobile robots. Each robot simultaneously estimates the current swarm shape,
by local information diffusion in a communication network, while controlling its own motion
to drive the swarm to the desired shape. These continuous-time estimation and control
laws result in provably correct behavior, but they make unrealistic assumptions for implementation
on actual robots.
This paper describes the first experimental implementation of decentralized estimation
and control based on information diffusion. We develop the hardware, software, and
communication protocols that adapt the theoretical approach to an actual implementation.
Experiments demonstrate that the robots successfully estimate the first and second inertial
moments of the swarm, using scalable local communication, while simultaneously moving
to control these moments.
