White Paper: Intelligent Route Optimisation for Fleet Operators

The Problem: Dead Kilometres and Idle Time

In any fleet operation, the most expensive kilometres are the ones without a passenger. Dead-running — the distance travelled between dropping off one passenger and picking up the next — represents a direct cost with zero revenue. For a typical urban taxi fleet, dead-running accounts for 35–45% of total kilometres driven.

Multiply that across a fleet of 200 vehicles operating 18 hours a day, and you're looking at thousands of dollars in wasted fuel, accelerated wear and tear, and missed earning opportunity — every single day.

FLEET's route optimisation module, currently in beta, is designed to attack this problem systematically.

How FLEET Route Optimisation Works

The engine operates on three layers:

1. Demand Prediction

Using historical booking data, time-of-day patterns, event calendars, and weather feeds, FLEET builds a rolling demand heatmap. This isn't a static model — it recalculates every 60 seconds, adapting to real-time conditions. If a concert lets out early or rain hits unexpectedly, the demand surface shifts and vehicles are repositioned accordingly.

2. Vehicle Positioning

Rather than waiting for a booking to come in and then finding the nearest car, the system proactively suggests repositioning moves. Drivers receive gentle nudges: "Move 800m north-west to Victoria Street — predicted demand spike in 12 minutes." These suggestions are optional, but drivers who follow them consistently see 15–20% higher utilisation rates in our beta cohort.

3. Multi-Stop Sequencing

For operators running pre-booked or shared-ride services, the sequencing engine solves the classic travelling-salesman variant in real time. Given a set of pickups and drop-offs with time windows, it computes the optimal route that minimises total distance while respecting every passenger's time constraint. The solver handles up to 40 waypoints per vehicle and returns a solution in under 200 milliseconds.

Beta Results

We've been running the route optimisation beta with three fleet partners across two Australian cities since January 2026. The early numbers are encouraging:

The 23% increase in jobs per shift is the headline, but the reduction in average pickup time matters just as much — it's the metric passengers actually feel.

Technical Architecture

The optimisation engine runs as a dedicated microservice within the FLEET cloud infrastructure. Key design decisions:

• Solver — custom constraint-satisfaction solver built on Google OR-Tools, with FLEET-specific heuristics for taxi/rideshare domain constraints
• Latency target — sub-200ms p99 for single-vehicle routing, sub-2s for full-fleet rebalancing
• Data pipeline — real-time GPS telemetry ingested via MQTT, aggregated into 60-second demand snapshots
• Privacy — all positioning data is ephemeral and not stored beyond the 24-hour operational window. No passenger location data is used in the demand model — only anonymised, aggregated booking counts by zone

What's Next

Route optimisation is currently available as a beta feature for FLEET customers with 50+ vehicles. We're actively expanding the beta cohort and refining the demand prediction model with each new market we enter.

Planned enhancements for the GA release include EV-aware routing (factoring in charge state and charger locations), multi-modal handoff for operators running both taxi and rideshare fleets, and an open API for operators who want to integrate optimisation signals into their own dispatch logic.

Interested in joining the beta? Contact our team to discuss eligibility and onboarding.