Real-time fleet oversight for autonomous shuttles

Challenge

Terhills is a resort in Maasmechelen, Belgium. It wanted guests to move between accommodations, dining and activity zones on autonomous shuttles that behave like a metro line rather than a set of individual vehicles. The resort partnered with EasyMile to run its EZ10 shuttles on site.

EasyMile's Site CC API is deliberately minimal. It lets you send one shuttle to a set of coordinates and streams back its telemetry. Scheduled lines, dispatching, collision avoidance, passenger information and operator tooling all had to be built on top.

Three groups needed different things: passengers inside a shuttle, guests waiting at a stop, and the staff running the fleet. All three views had to agree with vehicles moving in real time.

Solution

We built a real-time backend and one front-end for each of those groups.

In-Shuttle Display

A trip timeline inside every shuttle shows the current destination, the upcoming stops and progress along the route. It ships as a single-file build, so it installs cleanly on the in-vehicle hardware.

Station Kiosk

Every stop has a "next shuttles" board with live arrival times, destinations and intermediate stops. It is bilingual (Dutch and English) and shows announcements that operators publish for that station.

Operator Dashboard

A Mapbox view of the site shows shuttle positions, active trips and route conflicts. Operators create lines, assign vehicles, broadcast messages and tune operating parameters (shuttle speed, minimum spacing, door-opening time) without a developer.

Real-Time Backend

A GraphQL service turns the operator's lines into missions for EasyMile's API. The dispatcher assigns shuttles to scheduled lines and recomputes starting points mid-trip from telemetry. A collision-avoidance engine detects overlaps between routes. A state machine per shuttle reconciles the backend with the events each shuttle streams over an authenticated WebSocket.

Key Features

• One backend, three audiences: in-shuttle, station and operator apps share one data model
• Lines on top of vehicles: scheduled lines, mid-trip replanning and arrival times, built on a per-vehicle coordinate API
• Collision avoidance between shuttles on overlapping routes
• Operator self-service: stations, lines, fleet, announcements and operating parameters are all set from the dashboard
• Bilingual passenger screens (Dutch and English) with announcements per station
• Infrastructure as code for production and staging

Results

• Guests use the shuttles as a single transit service across the resort, not as separate vehicles
• Shorter waits and simpler navigation from real-time scheduling
• Resort staff run the service themselves: new lines, stations and announcements go live without a developer
• Room to grow: more shuttles, stations and routes fit the same model