robonar2024CC-BY-4.0
RONAR (RoboNar Dataset)
A real-world multimodal dataset paired with natural language narrations collected on a Stretch SE3 mobile manipulator performing household tasks in home environments. The dataset supports behavior transparency, risk estimation, and failure recovery with annotated failure cases and expert-labeled recovery instructions.
Downloads236
Episodes70
Likes1
Why This Matters for Physical AI
RONAR enables natural language grounding of robot experiences for transparency and failure analysis, advancing interpretable embodied AI systems that can articulate their actions and reasoning in real-world settings.
Technical Profile
- Modalities
- rgbdepthproprioceptionlanguageodometry
- Robot Embodiments
- Stretch SE3
- Environment
- homekitchenliving roomhallway
- Task Types
- manipulationnavigationgraspingobject_detectionpick_and_place
- Episodes
- 70
- Data Format
- ROSBAG
- Annotation Types
- language_instructionsfailure_labelsrecovery_descriptionsobject_detectionnarration
- License
- CC-BY-4.0
Community Signals
Top 50% by downloads
Academic Citations2
- Agentic LLM-based robotic systems for real-world applications: a review on their agenticness and ethics2025 · Frontiers in Robotics and AI
- A survey on state-of-the-art deep learning applications and challenges2025 · Engineering Applications of Artificial Intelligence
HuggingFace Discussions1
Access
Need custom rgb data?
Claru builds purpose-built datasets for home applications with dense human annotations and quality assurance.
Request a Sample Pack