RoboSense Spatial Camera Tech Backs First High-Resolution Precise-Depth 3D Manipulation Dataset

On November 11, during CoRL 2026, RoboSense announced a strategic partnership with BitRobot centered on 3D perception and real-world data collection. The first public outcome is BitRobot's release of the RoboCap LiDAR-100 dataset, which the companies describe as the world's first high-resolution, precise-depth 3D manipulation dataset.

The dataset adopts RoboSense's 3D perception solution and spans 100 hours of real operations in kitchens, auto-repair shops, and car-detailing environments. Its multimodal recordings are intended to help robots understand the physical world and learn manipulation skills from practical workflows rather than narrow lab setups. The partners position RoboCap as a resource for embodied AI researchers who need task-level diversity.

RoboSense says the collaboration addresses a key embodied-AI bottleneck: collecting accurate depth data at high resolution during genuine tasks. By combining spatial-camera sensing with scalable real-world capture, the partners aim to improve training reliability for robots that encounter varied objects, tools, and lighting. They add that the release establishes a foundation for further work on data standards and larger real-world collection efforts. Such data is seen as increasingly important as robot learning shifts from simulation toward real-world deployment.