Piper-ViT-VS

Status: Planned. This example will explore adapting ViT-VS to the Piper robot. ViT-VS uses pretrained vision-transformer features, including a DINOv2 option, for visual servoing.

Reference implementation: ViT-VS on GitHub.

Planned scope:

• Review the reference implementation and identify the interfaces needed for Piper.
• Integrate the camera and robot interfaces, camera-to-tool calibration, and end-effector velocity commands.
• Evaluate visual convergence and robot tracking, documenting limitations and lessons learned.

Related foundations: Visual Servoing.