Handoff requires time to reconstruct the situation
Aliases: takeover time · situation reconstruction · takeover readiness
What it is
Takeover situation-reconstruction time runs from a control request until the person understands goal, robot state, environmental change, prior automation action, and remaining options. Hands on a controller do not imply effective control; a mental model must be restored.
Why it happens
During autonomy, attention may move elsewhere and continuous proprioceptive and causal cues disappear. A request requires resampling, anomaly detection, short-horizon prediction, and strategy. Removing protection immediately transfers authority before awareness, prompting hesitation or overcorrection.
Studying it
Out-of-loop duration, secondary task, request content, and time budget can vary while measuring gaze return, state comprehension, first correct input, stable control, and conflict. Control contact and safe control recovery are separate. Scenarios should include subtle prior automation actions.
Where it stops holding
Continuously engaged operators need less reconstruction but still make mode errors. Expertise cannot remove new environmental change. Fast hazards require continued automatic protection, and extra time cannot compensate for missing information.
Applying it
- Include current goal, location/trajectory, trigger, recent automation action, and risk in the request.
- Retain safety constraints until understanding and stable input are confirmed.
- Measure prompt-to-safe-control distributions across disengagement durations and size the window from a high quantile.