RoboSense, BitRobot Launch 100-Hour LiDAR 3D Manipulation Dataset

The significance here is less the dataset itself than the pairing of a LiDAR hardware vendor with a data-collection partner: physical-world manipulation data is becoming a product line, not just a research byproduct. The kitchen, auto repair and car detailing settings matter because they are messy, contact-heavy environments where vision alone tends to struggle, so multimodal 3D capture is the differentiator. Who this affects: embodied-AI model developers and robotics teams that currently rely on camera-only datasets. The open question is whether 100 hours is enough to generalize beyond these three domains, and whether hardware-linked data becomes a durable moat or a commodity.