Push-test protocol - positive side first, fragile side spotted, axes aligned in the log, and cross-machine push counts stay qualitative

Symptom

Hand-push testing on hardware risked falls on a side sim had already flagged as fragile, and push counts invited apples-to-oranges comparison with sim numbers.

Context

Sim chirality was explicit: descendants were far more fragile in -y (fric-3000@kd1.2: +6 N*s survived 15/20 vs -6 N*s only 3-9/20) while the s1e control was perfectly symmetric (40/40). The protocol therefore: push the positive direction first, keep a spotter for the negative side; before any push, record which real-robot side corresponds to sim's +y in the log ("上机前对一次坐标"); and - citing the chaos lesson ("混沌课文") - real push results are used only as qualitative corroboration, never compared numerically with sim survival counts across machines.

Change

Push testing became a scripted, chirality-aware protocol with frame alignment as a logged precondition and an explicit epistemic limit on cross-domain count comparison.

Outcome

The fragile side was tested with protection informed by sim's quantified asymmetry; logs stayed interpretable because the frame correspondence was recorded before the first push.

Mechanism

Disturbance-response chirality is a real, quantifiable lineage property, so test order should follow measured fragility; and perturbation outcomes are chaotic in the details (divergent trajectories from tiny differences), so counts do not transfer across domains even when qualitative rankings do.

Applies when

• planning push/disturbance tests on hardware
• sim shows directional asymmetry in disturbance survival
• someone proposes comparing real push counts to sim counts