The dispatcher looks at the map — 40 trucks, and three haven't updated their position for 20 minutes. Is the transmission stuck? No network in the mountains? Or was the GPS antenna intentionally turned off? These are three different scenarios requiring different reactions, and the app must distinguish them — not just turn the dot gray. Our team has 5+ years of experience building such systems. We offer turnkey app development, including integration with popular telematics platforms and a backend for data processing. This is not just a map with markers — it's a real-time system where each point contains metadata: speed, direction, engine status, tracker battery level. If stationary for more than 5 minutes, the system automatically checks for network connectivity (tracker heartbeat) and ignition status. Only then can you distinguish a breakdown from a planned stop.
What the dispatcher actually needs
A monitoring app is not just "markers on a map." The dispatcher manages a fleet of 20–200 units and needs:
- Real-time data with metadata: coordinates + speed + direction + engine status + tracker battery level
- Route history: daily/weekly trail with geocoded stop addresses
- Geofencing: notification on entry/exit from warehouse, construction site, restricted area
- Alerts: speeding, prolonged idling with engine on, route deviation
All of this requires different client architectures. We have already implemented projects with fleets from 50 to 500 units, so we know how to scale the solution for any task.
How we receive telematics data in real time
Telematics units (Teltonika FMB, Wialon TK, Navixy OEM) send data via TCP or GPRS to a server. The mobile app does not connect to trackers directly — it receives the processed stream via WebSocket or MQTT client.
Data reception process:
- Tracker sends raw data to the server via the manufacturer's protocol (TCP, GPRS).
- The server parses the data, enriches it with meta information, and publishes to an MQTT topic.
- The mobile app subscribes to the topic via WebSocket and receives updates in real time.
On Android, we use OkHttp WebSocket with EventBus or SharedFlow to deliver updates to the ViewModel. On iOS, URLSessionWebSocketTask (iOS 13+) or Starscream for older targets. Updates arrive as JSON or protobuf — protobuf is preferable for large fleets: a packet of 40 vehicles × 10 fields in protobuf takes ~1.5 KB versus ~8 KB in JSON, saving 5x traffic. This reduces mobile internet costs for dispatchers. According to protobuf documentation, the binary format can reduce data size by 3-10 times compared to JSON.
How to ensure map performance with a large fleet
200 markers on a map with movement animation is already stressful. Key solutions:
Annotations instead of SVG overlays
On iOS, MKAnnotationView handles up to ~300 markers without noticeable lag. Beyond that, we use MKOverlay with a custom renderer that draws all points in a single CALayer. On Android, Google Maps SDK with MarkerOptions degrades after ~500 markers — we switch to Mapbox Maps SDK v10 with SymbolLayer based on a GeoJSON source: the entire fleet is updated with a single source.setGeoJson(featureCollection) call.
Server-side clustering
At zoom < 11, clusters are computed on the server (PostGIS ST_ClusterKMeans), and the client receives ready centroids with a counter. Local clustering (Supercluster) works for fleets up to 300–400 units.
Motion animation
Tracker positions update every 10–30 seconds — markers should not "jump." ValueAnimator with LatLngInterpolator (Android) or CABasicAnimation with CGPoint interpolation (iOS) — the marker smoothly "slides" to the new point.
Comparison of clustering methods:
| Method | Max markers | Latency |
|---|---|---|
| Supercluster (client) | ~400 | <100ms |
| PostGIS ST_ClusterKMeans (server) | Unlimited | <50ms (with index) |
Comparison of data formats:
| Format | Packet size (40 vehicles) | Traffic savings |
|---|---|---|
| JSON | ~8 KB | — |
| Protobuf | ~1.5 KB | 5x |
How route history and geocoding are built
A day's track is 2000–8000 points depending on the recording interval. We display it via Polyline / MKPolyline, but not the entire track at once: we load the visible map bbox and request points only for it. At "entire day" zoom, we discretize the track using the Douglas-Peucker algorithm on the server.
Stop addresses — reverse geocoding via Google Maps Geocoding API or OpenStreetMap Nominatim (self-hosted). We cache results in SQLite to avoid repeated requests when scrolling history.
What to do with alerts and notifications?
Speeding, geofence events, prolonged idling — triggers are calculated on the server, push notifications arrive via FCM/APNs. On iOS, we use UNNotificationCategory with UNNotificationAction — directly from the notification, you can open the map with the specific vehicle.
Geofences — GeoJSON polygons, checked with ST_Contains in PostgreSQL + PostGIS on every incoming tracker message. The mobile client only displays geofences and receives alerts — it does not compute intersections locally.
Common mistakes in notification implementation
- No background state handling: the app does not receive push if the user kills the process. Use
background fetchto reconnect to WebSocket. - Incorrect geofence setup: if a polygon is too complex (100+ vertices), PostGIS
ST_Containsslows down. Optimize polygons with Ramer-Douglas-Peucker simplification on the server. - Forgot about APNs/FCM tokens: when a new device is installed, the old token becomes invalid — always update it on the server.
What's included in development?
- Analysis of telematics device protocols and existing platform APIs
- Dispatcher interface design: map, vehicle list, history, notifications
- Development of all modules: data reception, map display, geofences, alerts, route history
- Load testing with simulation of 200+ trackers
- Publication on App Store and Google Play, push notification setup
- Documentation and dispatcher training
- Post-launch technical support
How long does development take?
MVP (map + real-time + history): 6–10 weeks. Full platform with geofences, alerts, mileage analytics, and reports: 3–5 months. Pricing is calculated individually after an audit of your infrastructure.
We guarantee stable operation under load of up to 500 vehicle units. Contact us for a project assessment — get a consultation on architecture and timelines. Request a cost estimate — it's free.