Force-based Perception and Control Strategies for Human-Robot Shared Object Manipulation

Licentiate thesis, 2019

Physical Human-Robot Interaction (PHRI) is essential for the future integration of robots in human-centered environments. In these settings, robots are expected to share the same workspace, interact physically, and collaborate with humans to achieve a common task. One of the primary tasks that require human-robot collaboration is object manipulation. The main challenges that need to be addressed to achieve a seamless cooperative object manipulation are related to uncertainties in human trajectory, grasp position, and intention. The object’s motion trajectory intended by the human is not always defined for the robot and the human may grasp any part of the object depending on the desired trajectory. In addition, the state-of-the-art object-manipulation control schemes suffer from the translation/rotation problem, where the human cannot move the object in all degrees of freedom, independently, and thus, needs to exert extra effort to accomplish the task.
To address the challenges, first, we propose an estimation method for identifying the human grasp position. We extend the conventional contact point estimation method by formulating a new identification model with the human applied torque as an unknown parameter and employing empirical conditions to estimate the human grasp position. The proposed method is compared with a conventional contact point estimation using the experimental data collected for various collaboration scenarios. Second, given the human grasp position, a control strategy is suggested to transport the object in all degrees of freedom, independently. We employ the concept of “the instantaneous center of zero velocity” to reduce the human effort by minimizing the exerted human force. The stability of the interaction is evaluated using a passivity-based analysis of the closed-loop system, including the object and the robotic manipulator. The performance of the proposed control scheme is validated through simulation of scenarios containing rotations and translations of the object.
Our study indicates that the exerted torque of the human has a significant effect on the human grasp position estimation. Besides, the knowledge of the human grasp position can be used in the control scheme design to avoid the translation/rotation problem and reduce the human effort.