Definition
Point clouds represent 3D environments as unordered sets of points, each with coordinates (x, y, z) and optionally color (RGB) or normal vectors. They are produced by depth cameras (Intel RealSense, Azure Kinect), LiDAR sensors, or stereo vision systems. In manipulation, point clouds enable 6-DOF grasp prediction, object pose estimation, and scene understanding. Networks like PointNet, PointNet++, and 3D-Diffusion-Actor process point clouds directly. Compared to 2D images, point clouds provide metric 3D information essential for spatial reasoning in cluttered environments.
Why It Matters for Robot Teams
Understanding point cloud is essential for teams building real-world robot systems. Whether you are collecting demonstration data, training policies in simulation, or deploying in production, this concept directly affects your workflow and system design.