R-KNav: 10,000 Hours of Sidewalk Rover Multimodal Data from 15 US Locations

R-KNav is a multimodal dataset derived from real-world operations of the R-Kiwi sidewalk rover fleet. Robot.com collected 10,000 hours of data across 15 locations in the United States to support the development of AI-driven autonomous robotics. The dataset is intended to catalyze the evolution of robotics foundation models.

Use Cases

• Training multimodal perception models based on sensor data from sidewalk rovers.
• Developing autonomous navigation algorithms based on real-world urban operation data.
• Benchmarking robot learning models on a large-scale, real-world dataset.
• Studying long-term robot deployment patterns across diverse geographic locations.

Strengths

• 10,000 hours of recorded multimodal data provides substantial scale for model training.
• Data originates from real-world operations of a commercial sidewalk rover fleet across 15 distinct US locations.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and specific file formats are unknown, which may limit suitability assessment.

Provenance

Source

Robot.com, derived from their R-Kiwi sidewalk rover fleet operations.

Collection Method

Collected from real-world operations of a commercial robot fleet.

Freshness

Last updated 2026-06-25 15:46:12; freshness should be verified.

Geography

15 locations in the United States.