Human-Robot Collaborative Lifting Motion Prediction and Experimental Validation

This paper presents a human-robot collaborative symmetric lifting motion prediction using inverse dynamics optimization. The human and robot arm are modeled in Denavit-Hartenberg (DH) representation. A floating-base rigid body with 6 global degrees of freedom (DOFs) is similarly modeled as a three-dimensional (3D) table. A set of grasping forces characterizes the human-table and robot-table interactions. The joint torque squares of human arm and robot arm are minimized and subjected to physical and task related constraints. During lifting, the design variables include the cubic B-spline control points of joint angle profiles of the human arm, robot arm, and table. In addition, the discretized grasping forces are also treated as design variables. Both numeric and experimental human-robot lifting was performed with a 2 kg table. The simulation reports the human and robot arm's joint angle profiles, joint torque profiles, and grasping force profiles. These profiles were validated with experimental data, which was collected using a motion capture system, force sensors, and the robot operating system (ROS). The human and robot arms' joint angle and torque profiles demonstrate a similar trend in the experimental environment. The grasping force comparison implies that the human and robot share the load while lifting together.