Real-World Accuracy Improvement for Cooperative Manipulation using Model Predictive Control

Robot systems have the potential to replace Computerized Numerical Control machines in the whole production line, even in challenging milling processes. When the capabilities of a single robot are insufficient, cooperative manipulation with multiple robots can be a solution. But until now, it has not seen utilization in reality due to the worse accuracy and difficulties in load distribution. Based on the model predictive control method in our theory-oriented work (Ye et al., 2024), the current paper presents the implementation and the real-world experimental results of two physically coupled robots moving a milling spindle. The novel implementation comprises the online identification of uncertain kinematic parameters and the real-time capable solution of optimal control problems. In the experiments, we show that the performance of our approach surpasses state-of-the-art control methods for cooperative manipulation in terms of optimized load distribution and path tracking deviation (reduced by at least 70.7%). Further, the deviation is reduced by 61.8% compared to that of a single robot.