Stanford Kuka Multimodal: 3,000 Episodes of Robotic Interaction Data

This robotics dataset contains 3,000 episodes and 149,985 frames of multimodal data collected from a Kuka robot arm. Released by the LeRobot team and associated with research paper 1810.10191, the collection provides 20 FPS video and time-series sensor data for a single robotic task.

Use Cases

• Training imitation learning models using the 149,985 video frames
• Analyzing robot trajectory consistency across 3,000 episodes
• Developing multimodal fusion algorithms combining video and time-series data

Strengths

• 3,000 total episodes providing high repetition for a single task
• High-frequency 20 FPS capture rate
• 149,985 total frames for visual model training
• MIT licensed for open research and commercial use

Limitations

• Limited diversity with only one task represented
• Metadata lists robot_type as unknown despite the title reference
• Geographic bias to US-based research settings

Provenance

Source

Stanford University / LeRobot (Arxiv 1810.10191)

Collection Method

Physical sensor and video capture from a Kuka robot arm

Freshness

Last updated September 2025.

Geography

United States