Human–Robot Cooperative Heavy Payload Transportation Based on Whole-Body Model Predictive Control

Abstract

Human-robot collaborative transportation is crucial for expanding robot abilities in smart manufacturing, requiring robots to recognize and follow human intentions while alleviating the human load and ensuring safety. To address this issue, a collaborative framework is proposed that integrates a collaborative controller with a whole-body controller. The collaborative controller uses the object’s motion states and robot force sensor data, along with the object’s center-of-mass dynamics model, to predict the desired position and orientation of the end effector. At the same time, a kinematic MPC is used to generate optimal joint and base velocity commands. The whole-body controller integrates a velocity command controller with impedance control, introducing end-effector load compensation to ensure safe and stable collaborative operation. A closed-loop stability analysis is conducted directly on the complete system dynamics, proving that the velocity tracking error is uniformly ultimately bounded under bounded disturbances and an explicit effective-gain condition. Simulation and physical experiments validate the effectiveness of this framework. The submitted experiment videos were recorded in real-world scenarios, demonstrating the system’s ability to achieve secure, reliable, and efficient human-robot collaboration. Note to Practitioners—This paper presents a human-robot collaborative transportation framework aimed at improving the efficiency and safety of collaboration between robots and humans. Traditional methods typically focus on capturing human joint behaviours to predict human intentions, enabling human-robot cooperation. However, these methods increase system uncertainty, computational overhead and hardware costs, and their effectiveness may not be ideal. Therefore, we propose a framework where the robot focuses on the object’s state and the contact forces between the robot and the object to achieve collaborative tasks. With just a wrist force sensor and a vision sensor capable of reasoning the object’s orientation, adequate collaborative transportation can be achieved. Additionally, our framework introduces impedance control, reducing the stiffness of the manipulator joints and ensuring safety during the collaboration process. In the future, we will conduct experiments with different types of robots to adapt the system to more industrial scenarios.