Delft Haptics Lab

Robustness of Haptic Shared Control Against Model Inaccuracies During Telemanipulation

Abstract

This MSc project was performed by Jeroen van Oosterhout, under supervision of David Abbink.

Abstract To improve teleoperated task performance and control effort, haptic shared control can assist the human operator along a safe and optimal path with continuous guiding forces from an automated controller. But previous research tested such shared control on tasks that were accurately defined in the controller (giving flawless guiding forces), while in a practical implementation guidance will be flawed due to inaccurate model or sensor information. This research investigated the effect of zero, small (7.5 [mm]) and large (17.5 [mm]) guiding errors on task performance and control effort. Included in the experiment were high and low transparency (quality of reflected forces from the environment) to test if that affects the ability to deal with wrong guiding.
In a human factors experiment subjects were provided with high and low transparency and inaccurate haptic shared control. The subjects performed a three dimensional virtual reality peg-in-hole type task (30 [mm] diameter; 0.1 [mm] clearance). The results showed that small guiding errors improved task performance and control effort with respect to unguided operations for both transparency levels. Thus despite small guiding inaccuracies haptic shared control still aides operators during telemanipulation while transparency does not affect that aid.