PocketGo is a portable platform for collecting high-quality multimodal robot demonstration data to enable fast imitation learning.
Standard:
Handheld & arm-mounted gripper kits (≤2 kg payload)
GoPro camera: 5.3K@60fps, 155° FOV, HyperSmooth 6.0
Visual-inertial SLAM (ORB-SLAM3) for 6D pose tracking
Full software stack: data → calibration → training → deployment
Supports up to 4,000 demos/day
Optional:
Mini AI computer: i9 + RTX GPU (≥11 GB VRAM) for edge training/inference
