Non-Prehensile Cooperative Object Transportation with Omnidirectional Mobile Robots: Organization, Control, Simulation, and Experimentation

Abstract

Distributed robotics promises to increase flexibility and robustness in robotic automation. The robotic network needs to adapt self-reliantly to different scenarios. Similarly, robustness advantages can be put into practice only if robots can join and leave the robotic network at any time. In that regard, cooperative object transportation constitutes an appropriate model problem to study whether a cooperation scheme is fit to realize the potential of distributed robotics. This paper proposes an optimization-driven scheme for the transportation of, potentially non-convex, polygonal objects which shall be pushed by a group of omnidirectional mobile robots. For dynamic control, distributed model predictive control is used, whereas a custom distributed optimization algorithm is proposed to deal with organization. The latter algorithm is, essentially, a distributed version of augmented Lagrangian particle swarm optimization. Experiments with purpose-built robots show that the scheme can deal with different scenarios without scenario-specific parameter tuning as well as with robots joining and leaving, realizing the promises of cooperative robotics.