TIMING CONTROLLER FOR HUMAN-ROBOT OBJECT HANDOVER: IMPLEMENTATION AND EVALUATION

Abstract

This work compares two controllers for human-robot object handovers. The proposed Timing controller allows users to specify timing parameters for the handover and computes velocity control using Model Predictive Control (MPC). This controller provides feedback if it cannot satisfy those constraints, which can be used to better tune the controller. We implemented the controller on a collaborative robot with two objective functions: minimum cumulative jerk (MCJ) and minimum cumulative error (MCE). For each, we conducted a user study to compare it with a baseline Proportional Velocity (PV) controller. Users specified the controller parameters and performed handovers with the robot using both controllers. We found that the timing controller with the MCE implementation can provide better user experience and fewer failures compared to the PV controller. Our findings could help towards the design of better controllers for Human-Robot handovers.