Publication
Weight perception in exoskeletonsupported teleoperation
Mareike Förster; Su-Kyoung Kim; Michael Maurus; Shivesh Kumar; Bigna Lenggenhager; Elsa Andrea Kirchner
In: The Third Neuroadaptive Technology Conference. Neuroadaptive Technology Conference (NAT-2022), October 9-12, Lübbenau, Germany, Pages 59-61, n.n. 10/2022.
Abstract
Despite advances and research in the area of teleoperation (e.g. (1-3)) over the last decades and the
development of active exoskeletons that provide force feedback to the human for a transparent and
intuitive interaction (4), it is still challenging to measure the subjective transparency of a
teleoperation system. But this information is of high interest to be able to define the sensibility of
the system and to optimize the interfaces used (5). To address the question of transparency, we
measure the ability of a teleoperator to distinguish weights in two different conditions: weights
added to the exoskeleton (“teleoperation OFF”) or to a robot under teleoperation (“teleoperation
ON”, see Fig. 1a and 1b). In the latter case, weight information was transferred using force feedback
between two different robot platforms, i.e., RH5 Manus humanoid [6] and Recupera-Reha
exoskeleton [7]. The implemented teleoperation framework utilizes the force control available in the
exoskeleton and kinematic control on the humanoid. The movement intention of the human inside
the exoskeleton is transferred to the humanoid robot using workspace scaling. On the other hand,
forces and torques felt by the end-effector of the humanoid are scaled and transferred as additional
end-effector forces in the exoskeleton’s inverse dynamic model using HyRoDyn [8] to enable force
feedback. Experiments were performed during a single-arm teleoperation setup without transferring
movements from the human to the robot but only force feedback from the robot to the human via
the exoskeleton. Both conditions, i.e., “teleoperation ON” condition (Fig. 1b) and “teleoperation
OFF“ condition (Fig. 1a) were compared when adding weights either to a basket attached to the
robots end effector or to the end of the exoskeletons’ hand interface structure. An adaptive
procedure was used to determine the perception threshold to optimize sampling level.
14 participants, between 21 and 30 years old and right-handed, took part in this study. Prior to the
experiments, the minimally perceptible weight during teleoperation was defined at 200g. During
weight changes the position of the arm of the exoskeleton and of the humanoid were predefined.
The exoskeleton was under force control and the controlled robot (used in the teleoperation
condition) was under position control to only transmit forces caused by the weight added to a basked
that was fixed to the end effector of the robot. To avoid visual and auditory cues the participants
were blindfolded and wearing noise cancelling headphones. Each experiment starts with an empty
box. After 30 seconds the start of the experiment is declared. 7 subjects started in teleoperation
mode and 7 subjects with the non-teleoperation condition. Weights were changed every 10s.
Weights were removed or added, depending on the participants’ answers according to the weighted
up-down staircase method (WUDM) by Kearnbach (9). If the participant perceived no gravitational
forces, weight was added until it was perceivable. If they perceived weight, it was reduced. In this
procedure for each trial the size of the increasing steps is weighted according to the last three
answers of the participant. The weight is decreased one step after each correct response and
increased three steps after each incorrect one leading to a convergence level of 75%. A total of 20
trials were carried out per condition.