Distributed model predictive formation control with discretization-free path planning for transporting a load

This paper proposes a comprehensive control scheme tailored to the task of letting a formation of omnidirectional mobile robots transport a plate through an unknown environment purely by normal and friction forces between the robots and the plate. The scheme includes a graph-based path planning strategy that can deal with obstacles of arbitrary polygonal shapes without any discretization of the environment. The strategy permits a natural implementation of a memory functionality that allows the robots to navigate through maze-like environments by successively constructing an approximative map of the environment during the motion. Most characteristic for the proposed approach, the robots are controlled using distributed model predictive control which makes it possible to enforce constraints on the movements of the robots to allow for a successful transportation of the plate with only acceptable slipping between the robots and the plate. The proposed scheme is carefully and successfully tested in various simulations posing rather sophisticated challenges both to its navigation and formation control subaspects.