MuJoCo velocity estimation: smoother signals can mean worse control

01 / QUESTION

54 actual simulation rollouts compare finite differences, EMA and an oracle reference across position noise and observation delay.

METHOD

Freeze source and protocol, execute local comparisons, archive raw records and independently recompute with hash checks.

RESULT

54 actual simulated rollouts and 108000 rows; 13 pass the final-300ms contract. Smoothing reduces some noise but worsens delayed tail error.

OBSERVATIONS

54 actual simulated rollouts and 108000 rows; 13 pass the final-300ms contract. Smoothing reduces some noise but worsens delayed tail error.

CONCLUSION

Derive finite-difference noise amplification and audit smoothing lag, force clipping and stable completion.

LIMITATIONS

Fixed planar actuator; no arm, vision, ROS2 or hardware. Full VL01 and author-review status remain unchanged.

REPRODUCE

The article supplies a pinned commit, locked environment, separate output directory and audit command.

1. Engineering cases

   MuJoCo velocity estimation: smoother signals can mean worse control

   54 actual simulation rollouts compare finite differences, EMA and an oracle reference across position noise and observation delay.

   2026-10-0911 MINPublished
2. 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.

   2026-09-28Experiment
3. Engineering cases

   Delayed MuJoCo feedback: can an old observation predict now?

   Twenty-four paired traces compare delay, position noise and constant-velocity prediction, retaining incomplete recoveries.

   2026-10-079 MINPublished