Location-Based Services

Satellite positioning made outdoor navigation effortless. Indoors it stops working — and that is where most of the interesting use cases are. I design and build indoor positioning systems, real-time locating systems (RTLS) and the location-based services on top of them: wayfinding, asset tracking, occupancy analytics and proximity-triggered experiences.

My approach is hybrid: no single technology is accurate, cheap and available everywhere, so I fuse several. And it is interoperable by default, so your positioning data is not trapped inside one vendor's platform.

Positioning technologies

Floor plan annotated with green and red markers indicating fingerprinting coverage quality per location — Wi-Fi fingerprint coverage mapped onto a building floor plan — a survey determines where a technique is reliable and where a second one is needed.

Choosing a positioning technology is an engineering trade-off between accuracy, infrastructure cost, battery life, privacy and how much of the building you can modify. In practice, the right answer is usually a combination — a coarse but always-available technique providing continuity, corrected by an accurate technique wherever it is available.

Bluetooth Low Energy

BLE beacons remain the most cost-effective way to cover a building. Suitable for room-level awareness, proximity triggers and, with enough density, metre-level trilateration.

iBeacon, Eddystone, AltBeacon and SemBeacon
Deployment planning and beacon density modelling
RSSI filtering, path-loss calibration and trilateration

Wi-Fi fingerprinting

Reuses the access points already installed in the building. A survey builds a radio map, after which devices are located by matching live measurements against it.

Fingerprint collection, cleaning and radio map generation
kNN, weighted kNN and probabilistic matching
Handling AP changes and radio map drift over time

Visual positioning & markers

Camera-based positioning, SLAM and fiducial markers give centimetre-level accuracy and orientation, which matters for augmented reality and robotics.

Fiducial marker detection and pose estimation
Semantic marker description via FidMark
Anchoring AR content to a real-world coordinate frame

Inertial & sensor fusion

Accelerometer, gyroscope and magnetometer data bridge the gaps between infrastructure-based fixes and smooth out a jittery track into a plausible path.

Pedestrian dead reckoning and step detection
Kalman and particle filters for multi-source fusion
Map matching against the building geometry

Built on OpenHPS

OpenHPS is the open-source hybrid positioning framework behind most of my work. Positioning techniques, filters and data sinks are expressed as nodes in a processing graph, so a system can be assembled from reusable parts and reconfigured without a rewrite. It is implemented in TypeScript and distributed as CJS, ESM and UMD, which means the same processing network runs on a server, in a browser, or on an embedded device.

Because OpenHPS is open source and vendor-neutral, you are not buying into a black box. It also serialises directly to RDF using the POSO vocabulary, so positioning output is interoperable data rather than an opaque coordinate stream.

Room used as an OpenHPS test environment with a tracked robot on a marked floor grid — **OpenHPS: An Open Source Hybrid Positioning System**
Technical report, 2020

Floor plan with fingerprinting coverage indicators — **Indoor positioning using the OpenHPS framework**
IPIN 2021

Mobile application displaying detected SemBeacons with RSSI and distance on an indoor map — **SemBeacon: discovering and detecting the position of physical things**
ACM IoT 2023

Smartphone showing an indoor beacon map of a building floor — **Interoperable and Discoverable Indoor Positioning Systems**
PhD dissertation, VUB, 2025

What you build on top

A position is not a product. The value is in the service it enables, and each of these has quite different accuracy and latency requirements — which is exactly why the positioning layer should be designed around the application rather than the other way around.

Indoor wayfinding

Turn-by-turn navigation in hospitals, airports, campuses and shopping centres — including accessible routing for visitors who cannot use stairs or need step-free paths.

Asset tracking & RTLS

Locating equipment, vehicles and stock in warehouses, hospitals and production facilities, with dwell time, utilisation and search-time analytics.

Smart building analytics

Occupancy, heat maps, flow analysis and space utilisation — aggregated and anonymised, so you get the insight without tracking individuals.

Proximity services

Context-aware content, geofencing and device discovery: services announce themselves to nearby users rather than requiring another app to be installed for every building.

Privacy and transparency by design

Illustration of a mapping application with a route, representing a survey on location data privacy — From my 2025 survey on the privacy and transparency of location data — third-party sale of location data was participants' single largest concern.

Location is among the most sensitive categories of personal data: it reveals where you live, work, worship and seek medical care. My 2025 survey on the privacy and transparency of location data found that users are broadly aware of this and consistently underestimate how many of their applications hold persistent location access.

I treat that as a design constraint rather than a compliance checkbox. Positioning can often be done entirely on the device, with only the derived result — or nothing at all — leaving it. Where data must be stored, a Solid pod keeps it under the user's control, with explicit and revocable access, instead of in an application silo. In practice this also makes GDPR obligations around data minimisation, purpose limitation and portability far easier to demonstrate.