Joringellcc-by-4.0
CRIA: Contact-Rich Interaction Anomaly Anticipation Dataset
A real-world multimodal dataset for studying anomaly anticipation in contact-rich robot manipulation, containing 2,852 trajectories from 18 manipulation tasks with synchronized multi-view RGB videos, robot state/action signals, force/torque measurements, and frame-level anomaly labels.
Downloads120
Episodes2852
Why This Matters for Physical AI
CRIA enables research in anomaly detection and anticipation for contact-rich manipulation by providing real-world demonstrations with force/torque sensing and frame-level anomaly annotations, critical for developing robust and safe robotic systems that can detect and prevent task failures.
Technical Profile
- Modalities
- rgbproprioceptionforce_torque
- Robot Embodiments
- Franka FR3
- Action Space
- joint_positions
- Environment
- lab
- Task Types
- manipulationinsertionwipingpressingpath_constrained_manipulationarticulated_object_manipulation
- Episodes
- 2852
- Data Format
- LeRobot
- Annotation Types
- reward_labelsaction_labelsframe_level_anomaly_labelsinteraction_onset_annotations
- License
- cc-by-4.0
Access
Need custom rgb data?
Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.
Request a Sample Pack