Model-based estimation of interaction force between human and collaborative robot

Abstract

Nowadays, the increasing demand for flexibility in the production process has spawned a new kind of equipment called collaborative robots that allows the human presence in their workspace and collaborate with each other on tasks. It has gained increasing attention because it combines the efficiency of robots and the dexterity of human beings through physical human-robot interaction. This dissertation aims at addressing two practical problems for handling human-robot interaction: the robot dynamics parameters identification, and extending the contact perception application from the end-effector to any position of the manipulator using the measurement of motor current. The robot dynamics parameter identification is formulated as a semi-definite programming problem derived from the linearized robot dynamics model and constrained by the physical consistency of the parameters. An experiment design procedure is made to generate trajectories that are optimal in the sense of the persistence of excitation. This dynamics identification framework turns out to work well under simulation with a small validation error. Based on the estimation of dynamic parameters, model-based contact detection and disturbance torque estimation are studied. Based on the decoupling property of disturbance torque estimation through momentum observer, a closed-form solution of contact point location and contact wrench estimation algorithm is obtained. The algorithm is tested under simulation and demonstrated high efficiency and reasonable accuracy.