As the interest in, and demand for, personal and service robots increase, research studies in human-robot interaction, especially involving physical contact between robots and humans, have become increasingly important. As an example of physical human-robot interaction (pHRI) research, cooperative object transport has been substantially investigated by many researchers. However, there is a gap in studies on bimanual cooperative object transport, a carrying mode essential for transport of large objects.

This research investigated human-human haptic interaction during cooperative bimanual transport of a large object. Eight pairs of human subjects, leader and follower dyads, were instructed to carry a large object with both hands and move cooperatively in the anteroposterior direction. The study focused on two haptic cues, the average rate of change of force (ARCF), and the interaction duration (tinteraction) employed by leaders and followers during the initiation phase of object transports. A custom-designed frame mounted with a load cell and an accelerometer was built to measure haptic interaction and transport movement in the anteroposterior direction. The experimental data showed that the leaders employed a repetitive ARCF during the initiation phase across trials. The other haptic cue tinteraction for followers to respond to the leader’s initiation was also found to be repetitive. Modeling of the abovementioned findings was conducted. The impedance model of a human arm during the initiation phase of bimanual cooperative transport was computed from the experimental data. Next, the expected interaction duration (t̂_interaction) computed by inputting the average rate of change of force to the impedance model was compared to the average value of the actual interaction duration (t̅_interaction) collected through the human-human study. The comparison showed that ̂t_interaction was larger than t̅_interaction but in a comparable range. The findings of the humanhuman bimanual cooperative object transport study and subsequent modeling provide a basis for future development of a controller for human-robot cooperative transport of a large object.