LAB / 020
E11 / Completion and current validity
Seven MuJoCo rollouts and a paired replay separate historical completion from current validity, including false alarms and expiry boundaries.
EXPERIMENT RESULT
01 / QUESTION
Seven MuJoCo rollouts and a paired replay separate historical completion from current validity, including false alarms and expiry boundaries.
METHOD
Seven real MuJoCo rollouts compare latched events and continuous state on identical packets: 2ms physics, 20ms captures, 250ms qualification and 60ms expiry.
RESULT
All first completions occur at 7.102s; force revokes validity at 8.022s, silence/replay becomes UNKNOWN at 8.042s.
OBSERVATIONS
One false bad capture causes false revocation; validity returns at 8.282s. Five sensor-fault pairs retain identical physical trajectories.
CONCLUSION
Retain historical completion while updating present validity; unknown is not known physical failure.
LIMITATIONS
Observation only, no regrasp. Single-capture revocation still causes false alarms. A fixed scene and synthetic sensor faults do not prove hardware safety or complete VL01.
REPRODUCE
The pinned repository includes protocol, raw results and independent auditing; the article provides reproduction commands.
KEEP EXPLORING
Related content
- Engineering cases
Embodied AI practice (11): placement completed, but is it still valid?
Seven MuJoCo rollouts and a paired replay separate historical completion from current validity, including false alarms and expiry boundaries.
- PROJECT / 003
Hohoo Embodied Experiments
Two MuJoCo teaching scenes—pick-and-place and planar force control—record contracts, tuning, clearance and delayed-feedback outcomes, including failures.
- Engineering cases
Embodied AI IX: a release command is not verified placement
Apply single-capture and sustained-window checks to five actual trajectories, including forged observations and post-completion disturbance.
- Engineering cases
Embodied AI X: why can a recovery budget stop a recoverable task?
Twelve rollouts compare unlimited recovery, a two-resume cap and a 400 ms cooldown, separating bounded behavior from task completion.