Tactile Sensors Are an Effective Substitute for Wrist Cameras in Contact-Rich or Occluded Manipulation
Tactile Sensors Are an Effective Substitute for Wrist Cameras in Contact-Rich or Occluded Manipulation
This research paper evaluates the effectiveness of finger-mounted tactile sensors as a substitute for wrist cameras in manipulation tasks involving contact-rich or occluded environments. The authors show that incorporating low-cost tactile arrays into a visuotactile ACT policy enables policies to recover performance advantages typically associated with wrist cameras, using scene and egocentric vision combined with touch. Across six real-world task families, adding tactile sensing improved success rates by 31.7-60.0 percentage points on three contact-rich tasks, outperforming vision-only policies with wrist cameras on all evaluated tasks. The results support the use of tactile sensing as an effective alternative to wrist cameras in scenarios where maintaining physical contact is critical.
The study demonstrates that finger-mounted tactile sensors can outperform wrist cameras on contact-rich manipulation tasks by directly measuring physical interaction without requiring visual access, achieving higher success rates across placement, insertion, and occluded retrieval scenarios.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
task families count
6
View original evidence
visual access. In this work, we show that this feedback enables policies to recover the performance advantage of wrist cameras using scene and egocentric vision together with touch. Our approach incorporates two low-cost tactile arrays into a visuotactile ACT policy, without wrist images during training or deployment. We evaluate six real-world task families spanning placement, insertion, occluded retrieval, and sust
Open source S3contact-rich tasks count
3
View original evidence
ained-contact tool use. On three contact-rich tasks, we independently vary wrist-camera and tactile availability while keeping the demonstrations, policy architecture, and optimization settings fixed. Adding wrist cameras improves success by 18.3-28.3 percentage points over vision without wrist images. Adding tactile sensing instead improves success by 31.7-60.0 percentage points and yields higher observed success th
Open source S4tactile arrays count
2
View original evidence
visual access. In this work, we show that this feedback enables policies to recover the performance advantage of wrist cameras using scene and egocentric vision together with touch. Our approach incorporates two low-cost tactile arrays into a visuotactile ACT policy, without wrist images during training or deployment. We evaluate six real-world task families spanning placement, insertion, occluded retrieval, and sust
Open source S3
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
visual access. In this work, we show that this feedback enables policies to recover the performance advantage of wrist cameras using scene and egocentric vision together with touch. Our approach incorporates two low-cost tactile arrays into a visuotactile ACT policy, without wrist images during training or deployment. We evaluate six real-world task families spanning placement, insertion, occluded retrieval, and sust
Open source S3
ained-contact tool use. On three contact-rich tasks, we independently vary wrist-camera and tactile availability while keeping the demonstrations, policy architecture, and optimization settings fixed. Adding wrist cameras improves success by 18.3-28.3 percentage points over vision without wrist images. Adding tactile sensing instead improves success by 31.7-60.0 percentage points and yields higher observed success th
Open source S4
Tactile Sensors Are an Effective Substitute for Wrist Cameras in Contact-Rich or Occluded Manipulation
Open source S1
Source：arXiv Robotics — research abstracts · arxiv.org