Skip to content

feat: add vision-relative front-flap pregrasp and teleop data path - #50

Closed
Alpenj wants to merge 6 commits into
pro/mujoco-shoe-followup-01from
pro/vision-relative-manipulation-01
Closed

Alpenj wants to merge 6 commits into
pro/mujoco-shoe-followup-01from
pro/vision-relative-manipulation-01

Conversation

@Alpenj

@Alpenj Alpenj commented Sep 4, 2026

Copy link
Copy Markdown
Owner

Scope

  • derive the measured shoe-box pose from simulated H201 metric depth without MuJoCo body pose or segmentation IDs
  • target the robot-facing front tuck flap right-hand grasp region
  • solve four guarded right-arm IK segments and add a bounded right-wrist RGB edge correction
  • reuse the existing MuJoCo keyboard teleop, camera payload, episode quality gate, and DAPIER-native ACT dataset
  • block the stale physical box-top target after the SIM target correction

Verification

  • 48 focused MuJoCo/data unit tests passed
  • 8 hardware-runner mock tests passed; no device opened
  • accepted compact pregrasp episode loaded as 65-frame 12D state/action RGB-D ACT data
  • CPU one-step ACT train/checkpoint round-trip passed

Not yet complete

  • operator visual confirmation of the corrected front-flap path
  • system udev alias reinstall
  • physical H201/wrist calibration and fresh camera frames
  • physical segmented visual-servo motion
  • right-arm lid opening and left-arm shoe extraction

This remains Draft. SIM and mock evidence does not authorize physical execution.

@Alpenj

Alpenj commented Sep 4, 2026

Copy link
Copy Markdown
Owner Author

Superseded by cleanly rebased Draft PR #51 after the parent PR #49 branch advanced. No hardware execution was performed.

@Alpenj Alpenj closed this Sep 4, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant