SYSTEM Cited by 1 source
Google Maps¶
Google Maps is Google's consumer mapping, routing, navigation, and ETA product. From the sysdesign-wiki's perspective it is interesting as a planet-scale routing + ETA-prediction system that has to condition predictions on per-trip features (lane type, modality, historical traffic, live probe data) and serve them under tight latency budgets to a consumer app.
Stub page. The wiki currently has one ingested Google Maps source and it captures only the intro of the HOV-ETA post. The entries below reflect what that source substantiates; this page will expand as more Maps posts are ingested.
Subsystems surfaced so far¶
- HOV-specific ETA surface (2025-06-30 launch) — a dedicated ETA prediction for routes that include HOV lanes, conditioned on a trip classifier that separates HOV trips from non-HOV trips. Motivation: HOV lanes are measurably faster than general lanes during peak (Utah Salt Lake Valley, ~16%), so a single lane-agnostic ETA mispredicts for HOV-eligible drivers (Source: sources/2025-06-30-google-hov-specific-etas-google-maps).
Design shape (inferred from one ingested source)¶
- Split ETA surfaces by trip category rather than trying to fold category information into a single ETA regression. The HOV launch explicitly says "HOV-specific ETAs", implying a separate surface selected at serving time by a classifier.
- Driver-facing route option rather than silent re-ranking — users opt in to HOV-inclusive routes, same UI shape as other modality splits Maps exposes.
Seen in¶
- sources/2025-06-30-google-hov-specific-etas-google-maps — Google Research announcement of HOV-specific ETAs + the classification system behind them.