E2 / Confirm the grasp before transfer

01 / QUESTION

Use sustained height and bilateral contact to decide whether transfer should proceed.

METHOD

Same scene and motion; 0/25/50 mm × baseline/guarded × three repeats = 18 rollouts. At 4 s, check height and bilateral contact over the last 0.2 s.

RESULT

Six biased failures cancelled transfer; three successful zero-offset trajectories remained identical. Task completion did not improve.

OBSERVATIONS

At 25 mm bias, post-gate horizontal path fell from 246.221 mm to a numerical residual; the cube was still not lifted.

CONCLUSION

For these fixed conditions, a task precondition prevented empty transfer without fixing pickup or establishing generalization.

LIMITATIONS

Privileged simulator state, deterministic repeats and a one-time check; not visual control, a hardware emergency stop or slip recovery.

REPRODUCE

Eight boundary tests; 18 independent CSV audits; nine matching prefixes, nine reproduced baselines and three identical successful pairs. Six recordings and state replay included.

1. Engineering cases

   Embodied practice II: if the grasp failed, skip the transfer

   An 18-rollout MuJoCo comparison of a lift/contact gate, with recorded trajectories, videos, runnable code and explicit limits.

   2026-09-2912 MINPublished
2. Tutorials

   Embodied AI Practice I: Pick up a cube in MuJoCo, then deliberately miss

   Run a contact-based pick-and-place task, compare0,25 and50mm target biases, and preserve videos, state trajectories and explicit success criteria.

   2026-09-2818 MINPublished
3. PROJECT / 003

   Hohoo Embodied Experiments

   Seven MuJoCo teaching experiments, from pickup bias to stage contracts, with raw trajectories, decisions and independent audits.

   2026-09-28Experiment
4. LAB / 004

   VL01 / Pickup target offsets

   What changes when pickup x targets shift by 25 / 50 mm with the controller fixed?

   2026-09-28Completed