Pollen Robotics has released Grabette, an open-source handheld device designed to record real-world robot manipulation data without requiring a robot or teleoperation rig. The system uses a fisheye camera and an RGBD depth camera to capture 6-DoF trajectories, allowing users to generate clean datasets that can be processed via a browser-based dashboard and shared on the Hugging Face Hub.
The device is built from standard off-the-shelf components, including a Raspberry Pi, Pi camera, and OAK-D depth camera, ensuring it is low-cost and reproducible. Data is stored as camera-local 6-DoF cartesian poses and gripper states, outputting to the LeRobot dataset format for compatibility with various learning methods. The release includes an example pipeline demonstrating how to train policies on recorded demonstrations.
The goal is to lower the barrier to entry for robot learning by enabling anyone to contribute to a large, collaborative manipulation dataset. By making data collection as easy as recording a video, Pollen Robotics aims to solve the supply problem of diverse real-world data that currently hinders the field.