evaluateZone every simulated second.Drone-side vision vs a fixed home camera: complementary, not rivals
| Drone-side vision (OpenLander, MDPI Drones 8(1):21) | Fixed home camera (DropZone) | |
|---|---|---|
| When it sees the zone | Only on final approach, a few seconds from ~13 ft | Continuously, minutes before arrival |
| Dog walks in 30 s before (scenario 1) | Sees it at the last moment → abort, wasted trip | Sees it early → Alexa asks, pet door, HOLD, then drop |
| Pool just outside the drop point (scenario 2) | Outside its FOV at 13 ft; can't know | Owner marked it once; audit + receipt "near the pool" |
| Stale survey: umbrella added after mapping (scenario 3) | Sees it on arrival → abort | Sees it now; household can move it before the drone launches |
| Last-second events (final ~2 s) | Wins: it is looking right then | Advisory may already be sent |
| Areas the home camera can't see / homes without cameras | Wins: works anywhere | Needs a camera covering the zone |
| Authority | Authoritative (aircraft's own sensors) | Veto-only: HOLD or NO_ADVISORY, never "safe" |
Full write-up with sources: docs/COMPARISON.md. Everything on this page is simulated: scripted actors, a ground-truth detector (projected 3D boxes, no occlusion, no misses) and a simulated carrier. The zone logic, advisory and proof-of-drop are the real modules from public/core/.