Genrobot Introduces DAS 2.0, Advancing Human Data For Robot Learning
(MENAFN- EIN Presswire)
SILICON VALLEY, CA, UNITED STATES, November 13, 2026 /EINPresswire / -- GenRobot has introduced its DAS 2.0 series, featuring DAS Dex 2.0, a next-generation glove for capturing dexterous hand-motion and tactile data, and DAS Ego 2.0, a head-mounted multimodal data-capture device offering a first-person perspective. Together, they synchronize visual, hand-motion, and tactile data to support scalable collection of real-world human manipulation data for robot learning.
More videos do not necessarily improve embodied AI training. Occlusion can distort action labels; similar trajectories can produce different contact outcomes; and interference caused by wearable devices, signal interference, or measurement drift can undermine consistency. Short grasping clips may miss approach, repositioning, and transport. DAS 2.0 combines complementary sensing and whole-body recording.
Three Configurations for Different Data Needs
Ego 2.0 captures first-person video and head–hand trajectories across operators, objects, and environments, including failed attempts and recovery actions. Ego + Wrist Camera adds close-up views of approach direction, finger adjustments, hand–object motion, and local deformation, although blur, occlusion, and alignment limitations remain. Ego 2.0 + Dex 2.0 adds direct joint and tactile measurements, helping distinguish visually similar actions with different physical outcomes.
Broad-coverage data expands task diversity, wrist-camera footage adds detail, and direct measurements constrain critical states. Dual-view tactile learning requires additional synchronized Ego + Wrist Camera + Dex recordings. Whole-body data aligns environmental observations, hand interactions, and body motion within shared temporal and spatial reference frames.
Ego 2.0: Wider Coverage and Finer Detail
Wider Vision. Richer Context.
Ego 2.0 features two 5-megapixel primary cameras and four 2-megapixel auxiliary cameras. Under equivalent conditions of aspect ratio, individual camera field of view, and working distance, upgrading the primary cameras from 2 MP to 5 MP increases linear sampling resolution by approximately 58% along each image dimension, reducing the projected linear dimension of the object surface represented by each pixel by approximately 36.8%. The overall horizontal field of view expands from 270° to 340°. The specified fingertip-tracking accuracy is 4 mm; reconstruction accuracy also depends on optics, exposure, calibration, occlusion, and algorithms.
Dex 2.0: Joint Motion and Tactile Sensing
Capture Motion. Measure Contact.
Dex 2.0 combines proprietary semi-flexible piezoresistive electronic skin with a side-mounted biomimetic exoskeleton. Positioning the exoskeleton along the sides of the fingers reduces interference when picking up objects from surfaces, adjusting objects within the hand, or manipulating objects in confined spaces. Its sensing architecture aims to distinguish actual contact from finger flexion, material deformation, preload caused by wearing the glove, and channel crosstalk. GenRobot reports consistent tactile responses in a month-long test involving repeated movements under comparable conditions.
Miniature magnetic encoders and kinematic calibration provide a 21-degree-of-freedom (21-DoF) representation of hand motion. Each glove incorporates 16 tactile sensors and 980 sensing elements, with a minimum detectable force of 0.1 N.
From Paired Tactile Learning to Whole-Body Tasks
Synchronized Ego + Wrist Camera + Dex recordings pair visual observations with direct joint and tactile measurements. Through paired supervision and knowledge distillation, future Ego + Wrist Camera configurations may estimate whether contact has occurred, contact regions, and trends in contact intensity, with confidence scores. Predicted values must remain distinct from measured labels.
Estimating current contact states relies on present and historical observations; predicting future contact outcomes must also account for intended actions. Vision cannot resolve every hidden contact state or reliably estimate absolute force without calibration; critical robot control still requires feedback from the robot itself.
Whole-body recordings connect planning, coordination, and execution, showing how approaching a target, adjusting body position and torso posture, and extending the arms contribute to tasks such as lifting, turning, and carrying objects. Reliable execution still requires robot-specific feedback and control.
Synchronized Hardware for Continuous Collection
A shared clock, hardware triggering, and exposure control enable head–hand temporal alignment at approximately 1 ms. At a hand speed of 0.5 m/s, a 1 ms offset corresponds to about 0.5 mm of movement, versus 5 mm at 10 ms. These figures describe timing-related displacement, not total tracking error.
Ego 2.0 supports hot-swappable magnetic battery packs that can be replaced without powering down, each providing approximately 2.5–3 hours of operation. Capture-status checks and batch post-processing help reduce interruptions and unusable data.
Together, these capabilities aim to help models progress from isolated interactions toward continuous-task learning. GenRobot will scale human data collection to advance embodied AI.
More videos do not necessarily improve embodied AI training. Occlusion can distort action labels; similar trajectories can produce different contact outcomes; and interference caused by wearable devices, signal interference, or measurement drift can undermine consistency. Short grasping clips may miss approach, repositioning, and transport. DAS 2.0 combines complementary sensing and whole-body recording.
Three Configurations for Different Data Needs
Ego 2.0 captures first-person video and head–hand trajectories across operators, objects, and environments, including failed attempts and recovery actions. Ego + Wrist Camera adds close-up views of approach direction, finger adjustments, hand–object motion, and local deformation, although blur, occlusion, and alignment limitations remain. Ego 2.0 + Dex 2.0 adds direct joint and tactile measurements, helping distinguish visually similar actions with different physical outcomes.
Broad-coverage data expands task diversity, wrist-camera footage adds detail, and direct measurements constrain critical states. Dual-view tactile learning requires additional synchronized Ego + Wrist Camera + Dex recordings. Whole-body data aligns environmental observations, hand interactions, and body motion within shared temporal and spatial reference frames.
Ego 2.0: Wider Coverage and Finer Detail
Wider Vision. Richer Context.
Ego 2.0 features two 5-megapixel primary cameras and four 2-megapixel auxiliary cameras. Under equivalent conditions of aspect ratio, individual camera field of view, and working distance, upgrading the primary cameras from 2 MP to 5 MP increases linear sampling resolution by approximately 58% along each image dimension, reducing the projected linear dimension of the object surface represented by each pixel by approximately 36.8%. The overall horizontal field of view expands from 270° to 340°. The specified fingertip-tracking accuracy is 4 mm; reconstruction accuracy also depends on optics, exposure, calibration, occlusion, and algorithms.
Dex 2.0: Joint Motion and Tactile Sensing
Capture Motion. Measure Contact.
Dex 2.0 combines proprietary semi-flexible piezoresistive electronic skin with a side-mounted biomimetic exoskeleton. Positioning the exoskeleton along the sides of the fingers reduces interference when picking up objects from surfaces, adjusting objects within the hand, or manipulating objects in confined spaces. Its sensing architecture aims to distinguish actual contact from finger flexion, material deformation, preload caused by wearing the glove, and channel crosstalk. GenRobot reports consistent tactile responses in a month-long test involving repeated movements under comparable conditions.
Miniature magnetic encoders and kinematic calibration provide a 21-degree-of-freedom (21-DoF) representation of hand motion. Each glove incorporates 16 tactile sensors and 980 sensing elements, with a minimum detectable force of 0.1 N.
From Paired Tactile Learning to Whole-Body Tasks
Synchronized Ego + Wrist Camera + Dex recordings pair visual observations with direct joint and tactile measurements. Through paired supervision and knowledge distillation, future Ego + Wrist Camera configurations may estimate whether contact has occurred, contact regions, and trends in contact intensity, with confidence scores. Predicted values must remain distinct from measured labels.
Estimating current contact states relies on present and historical observations; predicting future contact outcomes must also account for intended actions. Vision cannot resolve every hidden contact state or reliably estimate absolute force without calibration; critical robot control still requires feedback from the robot itself.
Whole-body recordings connect planning, coordination, and execution, showing how approaching a target, adjusting body position and torso posture, and extending the arms contribute to tasks such as lifting, turning, and carrying objects. Reliable execution still requires robot-specific feedback and control.
Synchronized Hardware for Continuous Collection
A shared clock, hardware triggering, and exposure control enable head–hand temporal alignment at approximately 1 ms. At a hand speed of 0.5 m/s, a 1 ms offset corresponds to about 0.5 mm of movement, versus 5 mm at 10 ms. These figures describe timing-related displacement, not total tracking error.
Ego 2.0 supports hot-swappable magnetic battery packs that can be replaced without powering down, each providing approximately 2.5–3 hours of operation. Capture-status checks and batch post-processing help reduce interruptions and unusable data.
Together, these capabilities aim to help models progress from isolated interactions toward continuous-task learning. GenRobot will scale human data collection to advance embodied AI.
Mona Wang
GenRobot
Visit us on social media:
LinkedIn
X
Legal Disclaimer:
EIN Presswire provides this news content "as is" without warranty of any kind. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the author above.
You just read:
GenRobot Introduces DAS 2.0, Advancing Human Data for Robot Learning News Provided By Arc Growth LLC November 13, 2026, 04:00 GMT Share This Article Distribution channels: Technology
EIN Presswire's priority is author transparency. We do our best to weed out false and misleading content. The content above is
the sole responsibility of the author who makes it available. If you have any complaints, kindly contact the author above.
Legal Disclaimer:
MENAFN provides the
information “as is” without warranty of any kind. We do not accept any
responsibility or liability for the accuracy, content, images, videos,
licenses, completeness, legality, or reliability of the information
contained in this article. If you have any complaints or copyright issues
related to this article, kindly contact the provider above.
Comments
No comment