Location Streams

Concepts Covered

• Mobile location updates
• Freshness windows
• Stale coordinates
• Driver availability state
• Stream processing
• Write volume and sampling
• Operational lag

Definition

A location stream is a continuous flow of position updates from moving clients such as drivers, riders, delivery couriers, vehicles, or devices.

Each update says:

driver_id=d_42
lat=40.741
lng=-73.989
heading=180
speed=8.2m/s
timestamp=2026-05-15T10:14:22Z

The stream lets the system estimate where an entity is now. It is an estimate because mobile devices move, networks delay updates, GPS can be noisy, and clients can disappear without clean shutdown.

The Pain That Forces Location Streams

A naive system asks the driver app for location only when a rider requests a trip.

That fails because matching needs immediate candidates:

rider requests trip
  -> server asks nearby drivers for location
  -> drivers respond slowly or not at all
  -> matching waits
  -> rider sees spinner

Ride matching needs a recent view of supply before the request arrives. Location streams exist because the platform must continuously maintain a live, approximate map of available drivers.

Mental Model

Location is not a permanent fact. It is a decaying signal.

A coordinate from two seconds ago may be useful. A coordinate from five minutes ago may be dangerous.

Instead of thinking:

driver is at this point

think:

driver was observed near this point at this time

Matching systems should use freshness windows. A stale location should be ignored, downgraded, or verified before assignment.

How It Works

A common flow:

1. Driver app sends periodic location updates.
2. Realtime gateway or ingestion API accepts updates.
3. Stream processor validates, deduplicates, and normalizes updates.
4. Latest-location store updates current driver position.
5. Geospatial index moves driver between cells.
6. Matcher reads available drivers from nearby cells.

The system may persist raw location history for analytics, fraud detection, ETA models, or support workflows, but the matching path usually needs a fast latest-state view.

Tradeoffs

Higher update frequency improves freshness but increases:

• mobile battery usage
• network usage
• ingestion writes
• stream processing load
• geospatial index churn

Lower update frequency saves resources but creates stale matches and worse ETAs.

Production systems often adapt update frequency based on driver state. A driver actively navigating to pickup may update more often than a driver idle in a parking lot.

Operational Reality

Important signals:

• update rate per city
• late update rate
• stale location rate
• geospatial index update latency
• dropped update count
• mobile client heartbeat gaps
• matcher candidate freshness