Each page follows one control cycle of a policy: what the cameras and joint encoders deliver, how the network turns them into tokens, and how many actions it commits to before looking again. The 3D view is the NexArm from our MuJoCo model, placed in the bowl-stacking workcell with the same front, top and wrist cameras the policies are trained on. The project background is in Building a Robot Learning Pipeline for the 6-DOF NexArm.
ACT
A CVAE-trained transformer that turns the NexArm's camera frames and joint state into 100 future joint targets, then runs all of them before it looks again.
- 100 · actions per chunk
- 100 · actions run before re-planning
- 30 Hz · control rate
- 902 · encoder tokens (3 cameras)
TurboVLA
DINOv3 image tokens and a BERT-encoded instruction are fused by six bi-attention layers, and an ACT-style decoder predicts 16 actions, of which the NexArm runs 8 before looking again.
- 16 · actions per chunk
- 8 · actions run before re-planning
- 30 Hz · control rate
- 588 · vision tokens (3 cameras)