Live GPS Tracking + Geofence Automation for E-Commerce Logistics

The blue-dot problem

Most fleets already have GPS. What they do not have is tracking. There is a difference. A raw telematics feed puts a blue dot on a map. Somebody in the control room still has to watch it, interpret it, and act on it. On a 25-truck night convoy running the Bangalore–Hyderabad and Mumbai–Pune corridors, nobody watches all 25 dots. So the wrong turn, the three-hour dhaba halt, and the breakdown on a dark stretch all get discovered late — usually when the customer calls to ask where their shipment is.

Tracking that earns its name does the watching for you. It knows which trip each truck is on, where the origin and destination fences are, what the allowed halt budget is, and what "off route" looks like on this specific lane. That is the gap between a GPS device and a tracking system, and it is the entire subject of this post.

The live layer: 30-second GPS

Everything sits on one heartbeat: a sync cycle that runs every 30 seconds for the whole fleet. On each cycle the system pulls live position from the telematics provider and updates a per-vehicle live state — latitude and longitude, speed, ignition status, and two timers that matter more than people expect: stopped-since and driving-since. Those two timers are the raw material for halt alerts and driver-safety monitoring later.

The feed is multi-provider by design, with WheelsEye and Veyron adapters live, so a tenant is not welded to a single telematics vendor. If you switch hardware, the tracking logic above it does not change. During each cycle the system also drops a GPS breadcrumb on every active trip, building a trail that both powers the customer route map and, over time, teaches route templates from completed trips. The live fleet map itself is scoped to in-transit vehicles — a live tracking view never drops a vehicle that is mid-trip.

Geofence auto-promotion at origin and destination

This is the feature that turns GPS into an operations tool. Every trip has an origin and a destination with fixed coordinates inherited from the lane. Around each, the system arms a geofence — a 200-metre radius measured by Haversine distance. When the truck crosses into the fence, the trip status advances automatically along the lifecycle:

Two design decisions keep this honest on Indian highways. First, a manual-hold override: if a human reverts a status because the automation jumped ahead of reality, the revert sticks and the fence will not immediately re-promote it. Automation should never fight the ground team. Second — and this is a hard law across the platform — no trip ever auto-closes. Geofence automation will carry a trip all the way to REACHED, but a human always confirms completion with a stage mark that carries the team's date and time. The books never close a trip the ground crew has not verified.

The one deliberate exception is repositioning: EmptyTrip legs, which carry no cargo and no customer SLA, are allowed to auto-close on geofence. That is a founder-confirmed exception, not a loophole in the no-auto-close rule.

Touching-stop detection

Real e-commerce lanes rarely go clean point-to-point. A trip touches an intermediate FC, a hub, or a milk-run stop on the way. Traxium models these as touching stops — intermediate points on a trip, each with its own geofence and an allocated minutes budget. As the truck moves through the lane, the sync detects arrival, departure, and overrun at each touching stop. If the truck sits at an intermediate stop past its allocated minutes, that overrun is visible immediately, not reconstructed after the fact from a driver's phone call.

Alongside touching stops, the system tracks generic stop events: any time a vehicle stays stationary longer than 10 minutes, it is logged with start and end epochs and a duration. That historical stop record is what later lets the customer-facing page infer why a truck paused without anyone typing it in.

Route-deviation and halt alerts

Two failure modes cause most of the late-night scrambles: the truck goes the wrong way, or the truck stops and nobody knows why.

Route deviation and backward movement

On each 30-second cycle, movement is compared against the expected route. Both route deviation and backward movement raise alerts. A driver who misses the highway exit and takes the long road, or a truck that is somehow heading back toward origin, surfaces as an alert instead of as a surprise at the destination FC.

Halt escalation

Halts follow a schedule tuned so it does not cry wolf. Alerts escalate at 10, 30, and 60 minutes, with further escalation in a breakdown scenario, and exactly one MOVING_AGAIN notification is fired per recovery so the control room is not spammed as the truck crawls through a toll queue. Underneath, a 3 km/h GPS noise floor ignores the position jitter that WheelsEye-class devices produce when a truck is actually parked, so a stationary vehicle does not fake "movement." Vehicle-diagnostics and GPS sync-health endpoints round this out, so when a device goes quiet you can tell a dead tracker from a truck in a tunnel.

Driver-safety night watch

Indian long-haul runs at night. A fatigued driver on the Mumbai–Pune corridor at 3 AM is a safety and an SLA risk at once. The driving-since timer feeds a safety monitor with two windows calibrated for that reality:

• Night watch: 3+ hours of non-stop driving between 9 PM and 8 AM.
• Day monitor: 4+ hours of non-stop driving between 8 AM and 9 PM.

The continuous-driving timer resets after a 5-minute stop — a threshold deliberately chosen so a genuine break at a dhaba or a fuel stop counts as rest, while a 90-second crawl through a checkpoint does not. Combined with the noise floor, this keeps the safety alerts credible enough that dispatchers actually act on them.

The customer share page that never leaks a speed

Here is the feature that protects your relationship with Amazon and Flipkart. Every trip can generate a public share-tracking page at a token URL (/public/track/{token}) that you send to the customer. It shows a clean green route polyline drawn from the GPS breadcrumbs, the current location with an "Open in Maps" button, and — this is the important part — smart stop labels inferred from duration and time of day: a long midday pause reads as a meal break, a brief one near a plaza reads as a toll, a night halt reads as rest.

What it deliberately does not show is any raw operational data. No raw speed. No delay figure. No driver-fault attribution. The messaging is diplomatic on purpose. Your customer sees a shipment moving competently down the highway with sensible, human-readable stops — not a dashboard that hands them ammunition for a penalty conversation. The route extends live on each 30-second refresh, so the customer watches genuine progress. To keep the public map fast, breadcrumbs are thinned to points at least 500 metres apart with a 500-point cap.

Internally, your dispatchers get the honest version: an ETA engine that shows a confidence window only when a trip is uncertain or running late, so a calm on-time trip stays calm on the board. Two audiences, two truths, one trip.

Tracking readiness checklist

Use this to judge any tracking setup — Traxium or otherwise — against what an e-commerce lane actually needs.

Tracking is not the end of the story — it is the front door. Because the geofence promotes the trip stage automatically, that same trip flows straight into finance clearance, GST invoicing, and driver settlement without anyone re-keying arrival times. The dot on the map and the number on the invoice come from the same event.

Frequently asked questions