Optimal physical human-robot collaborative controller with user-centric tuning

Collaborative robots are increasingly used in different fields of application to physically interact with the human (e.g., manufacturing, rehabilitation, etc.) In order to improve the physical collaboration performance, the interaction between the user and the robot must be safe, smooth, and intuitive. Indeed, this paper proposes a controller which is composed of three nested loops. A Passive Velocity Field Control (PVFC) defines the lowest control layer, ensuring the passivity of the system. The intermediate control layer is defined by a Cartesian impedance controller, managing the interaction between the user and the robot and sending to the PVFC a reference position. The outer layer is defined by a Linear Quadratic Regulator (LQR), detecting the intention of motion of the user and deforming accordingly the Cartesian impedance setpoint to follow such a motion. In addition, to enhance the collaboration performance for each user, a preference-based optimization approach is employed to tune the control parameters, implementing a human-centric tuning procedure. In such a way, the controller is customized for the specific user to establish the proper interaction with the robot. The proposed controller has been evaluated by making use of a Franka EMIKA panda robot as a test platform, comparing the achieved performance with a controller previously developed by some of the authors in a free-motion collaborative task along the z vertical direction. Achieved results show the improved performance obtained by the proposed controller. In addition, an assembly task has also been optimized to show the potential of the proposed control framework in complex and realistic situations.