A traffic intersection has a pedestrian at the corner, not yet committing to cross. An autonomous vehicle is approaching at 20 km/h. The vehicle could make a conservative decision (brake and prepare to stop) or an optimistic decision (assume the pedestrian won't cross and proceed at current speed). Why is the conservative decision preferred despite being less efficient?
AConservative decisions are legally required by traffic law
BConservative decisions are always optimal; efficiency is never relevant
CConservative decisions fail more gracefully: if the pedestrian does cross, braking prevents a fatal collision; if the pedestrian doesn't cross, modest delay is acceptable. Optimistic decisions have catastrophic failure (pedestrian collision) if prediction is wrong
DPedestrians always cross at intersections, so conservative assumptions are always correct
This hybrid is the practical solution to the planning tradeoff. Full open-loop is too brittle; full closed-loop is too slow. The receding horizon approach gives the best of both: real-time response to surprises while maintaining long-horizon goals.