UniTac: Whole-Robot Touch Sensing Without Tactile Sensors

1[Brown University]
[Teaser Figure]

Abstract

Robots can better interact with humans and unstructured environments through touch sensing. However, most commercial robots are not equipped with tactile skins, making it challenging to achieve even basic touch-sensing functions, such as contact localization. We present UniTac, a data-driven whole-body touch-sensing approach that uses only proprioceptive joint sensors and does not require the installation of additional sensors. Our approach enables a robot equipped solely with joint sensors to localize contacts. Our goal is to democratize touch sensing and provide an off-the- shelf tool for HRI researchers to provide their robots with touch-sensing capabilities. We validate our approach on two platforms: Franka robot arm and Spot quadruped. On Franka, we can localize contact to within 8.0 centimeters, and on Spot, we can localize to within 7.2 centimeters at around 2,000 Hz on an RTX 3090 GPU without adding any additional sensors to the robot.


Methodology

[Network Figure]

Design of UniTac-Net. UniTac-Net is a four-layer MLP with either a regression or a classification output head. It takes proprioceptive feedback (joint torques and positions) as input and predicts the contact (if any).


Results

Live contact localization on Spots. UniTac is able to detect continuous changes in contact location.

[Franka Qualitative]

Live contact localization on FR3 robot arm. Top row: A human applies touch to the robot. Middle row: The system localizes the contact point on the robot’s mesh. Bottom row: Normalized joint torque changes are displayed (different colors indicate distinct joint sensors). The torque plots show the unique signature look for each touch location, which we utilized to localize the contact.


Physical Human Robot Interactions

pHRI deployment on Spot.

Posture change includes sitting and lying down.

Body expression includes wiggling and playing bow.

Motion action includes leg lifts, forehand/haunches shift/turn, and sidepass.

pHRI deployment on Franka.

Franka picks up a cube of a different color, based on where the touch is on its joint links.

Acknowledgements

This work is supported by the Office of Naval Research (ONR) grant #N00014-22-1-259. Wanjia was supported by a UTRA award and a Randy Pausch Fellowship.

Contact

Wanjia Fu (contact email)

Hongyu Li (contact email)