A Distributed Optimal Control Framework for Multi-Robot Cooperative Manipulation in Dynamic Environments

This article proposes a control framework for cooperative robot manipulation based on distributed optimisation. Besides transporting a jointly grasped rigid object, whole-body avoidance of dynamic obstacles for the entire multi-robot system in uncertain environments is also part of the task. Since the physical connection through grasping can significantly reduce the operation capability of the system, internal performance of individual robots should be optimised during task execution. For this purpose, a framework for the characterisation of robot joint-space constraints and their inclusion into a highly multi-goal optimisation problem for the aforementioned reactive manipulation is designed. A computationally and communicationally efficient distributed optimal control algorithm based on primal decomposition is then developed for the multi-robot system to reach the optimum without generating internal conflicting forces in spite of different fields of view. Experiments on a multi-arm platform are conducted for validation. Results not only show effective coordinated avoidance of moving obstacles, but also demonstrate the benefit of optimising internal performance during cooperation.