Other
RH20T
Contact-rich robotic skill learning
Real-world multimodal manipulation collection with over 110,000 robot sequences, corresponding human demonstrations and language descriptions, spanning 147 tasks and 42 verb-level skills.
Access links
Primary places to inspect, download, read about, or reproduce the dataset.
Download
Not reportedPaper
Not reportedCode
Not reportedSupplementary
Not reportedQuick facts
- Vehicle type
- Other
- Environment
- Not reported
- Data origin
- Real-world
- Sensor count
- 2
- Ground truth
- Not reported
- Calibration
- Not reported
- Annotations
- Available
Sensor overview
High-level sensor availability before the detailed sensor records below.
Ground truth
Reference scope, method, rate, coverage, and provenance.
Not reported
Calibration and synchronization
Reported calibration level, reproducibility signals, and supporting notes.
- Level
- Not reported
- Processed parameters
- Not reported
- Raw calibration data
- Not reported
- Calibration targets
- Not reported
Calibration data
Calibration reproducibility details are not reported.
Notes
Camera extrinsics are calibrated before manipulation and checked per scene. Robot and force/torque frames are aligned; force sensors are tared; end-effector pose and wrench are transformed into each camera frame. Paper does not specify a reproducible raw calibration release.
Known or intentional limitations
Reported constraints and characteristics to check before using the dataset.
Fingertip tactile data are limited to configuration 7.
Foundation-model potential is not evaluated because of computing constraints.
Cameras
multiview-rgbd
- Model
- Not reported
- Setup
- Rgbd
- Count
- 1
- Effective cameras
- 2
- Modality
- Rgbd
- Resolution
- 1280 x 720
- Rate
- 10 Hz
- Shutter
- Not Reported
- Lens type
- Not Reported
- HFOV
- Not reported
- VFOV
- Not reported
- Notes
- RGB and depth images; binocular IR images also provided at 1280x720 and 10 Hz. Platform camera layout varies: 1–2 in-hand and 8–10 global cameras. Exact per-model counts are not specified.
IMUs
Not reported
GNSS
Not reported
LiDAR
Not reported
Radar
Not reported
Pitot tube
Not reported
Additional sensors
Other sensor
- Name
- Wrist force/torque sensing
- Count
- 1
- Rate
- 100 Hz
- Notes
- Six-axis force/torque. Configuration-dependent equipment includes OptoForce in Cfg 1, ATI Axia80-M20 in Cfg 2 and Franka sensing in Cfg 5; these are alternatives rather than simultaneous sensors.
Other sensor
- Model
- uSkin
- Name
- Fingertip tactile sensing
- Count
- 1
- Rate
- 200 Hz
- Notes
- Only configuration 7; reported tensor size 2x16x3 does not imply a physical count.
Other sensor
- Name
- Audio microphones
- Count
- 2
- Notes
- In-hand and global recordings. Table III prints 30 Hz for Audio without specifying whether this is a chunk cadence or waveform rate.
Other sensor
- Name
- Robot joint state
- Count
- 1
- Rate
- 10 Hz
- Notes
- Six/seven joint angles and torques; state channels are not a physical sensor count.
Other sensor
- Name
- Gripper Cartesian pose
- Count
- 1
- Rate
- 100 Hz
- Notes
- Reported size 6/7. Robot state is a manipulation input, not an independently validated trajectory reference.
Other sensor
- Name
- Gripper width
- Count
- 1
- Rate
- 10 Hz
Annotations
Reported annotation availability and task support.
- Availability
- Available
- Format
- Not reported
- Class count
- Not reported
Semantic segmentation
Not reported
Instance segmentation
Not reported
Object detection
Not reported
Optical flow
Not reported
Depth ground truth
Not reported
Citation
Citation key and BibTeX kept at the end of the page for reference.
Not reported
Dataset corrections
Notice a specific issue after reviewing the page?
Report an inaccuracy