Decentralized Nonlinear MPC for Robust Cooperative Manipulation by Heterogeneous Aerial-Ground Robots

Cooperative robotics is a trending topic nowadaysas it makes possible a number of tasks that cannot be performedby individual robots, such as heavy payload transportationand agile manipulation. In this work, we address the problemof cooperative transportation by heterogeneous, manipulator-endowed robots. Specifically, we consider a generic number ofrobotic agents simultaneously grasping an object, which is to betransported to a prescribed set point while avoiding obstacles.The procedure is based on a decentralized leader-followerModel Predictive Control scheme, where a designated leaderagent is responsible for generating a trajectory compatible withits dynamics, and the followers must compute a trajectory fortheir own manipulators that aims at minimizing the internalforces and torques that might be applied to the object bythe different grippers. The Model Predictive Control approachappears to be well suited to solve such a problem, becauseit provides both a control law and a technique to generatetrajectories, which can be shared among the agents. Theproposed algorithm is implemented using a system comprisedof a ground and an aerial robot, both in the robotic Gazebosimulator as well as in experiments with real robots, where themethodological approach is assessed and the controller designis shown to be effective for the cooperative transportation task.QC 20201109