Indoor positioning is not simply “GPS inside buildings”, nor is it synonymous with beacons, Wi-Fi, or UWB. It is a measurement, estimation, and integration chain that must turn imperfect signals into a useful position while making the service requirements and limitations explicit.
01 DEFINITION
What “indoor positioning” actually means
According to the NIST definition, indoor localization determines or estimates the position of a person, robot, or object equipped with an appropriate device inside buildings and underground structures. Tracking adds the time dimension: it repeats the estimate and makes the trajectory available to the party operating the service.
Positioning, localization, and tracking are often used interchangeably, but they are not the same as navigation and wayfinding. The former produce an estimate of where the target is; navigation computes a route, while wayfinding also includes maps, landmarks, signs, and interaction choices that help a person understand and move through a space.
A useful result is not merely a pair of coordinates. It must refer to the correct map, distinguish building and floor, express uncertainty, and identify what is being localized: a smartphone, a tag attached to an asset, a vehicle, a robot, or, in some systems, a person detected without a device. “Indoor GPS” conveys the idea of a dot on a map, but hides this variety of architectures.
02 THE INDOOR PROBLEM
Why GPS is not enough inside buildings
NIST notes that indoor systems cannot be compared without considering the building, infrastructure, calibration, maps, and operating conditions. Materials attenuate radio signals, while walls, floors, people, and metal surfaces create reflections and multiple propagation paths. A measurement that works in an empty corridor can behave differently in a crowded concourse or an equipment-dense ward.
GNSS is designed to receive satellite signals and estimate a position in a global reference frame. Inside a building those signals arrive weak, distorted, or not at all; the service often also needs to identify the floor, follow short movements, and respond in spaces where a few metres can mean a different room, ward, or side of a gate.
The environment is not static. Furniture, shelving, and machinery change; crowds absorb and block signals; doors and lifts alter possible paths; a smartphone may be held, pocketed, or differently oriented. An indoor solution must therefore manage uncertainty and maintain continuity, not merely produce an occasional point.
03 FROM MEASUREMENT TO SERVICE
From signal to position: how an indoor system works
Sensors, measurements, estimation, map, and application are distinct layers of the same chain.
The literature distinguishes physical technologies from the techniques that turn observations into position. The survey by Zafari, Gkelias, and Leung, for example, separates measurements such as received signal strength, time of flight, and angle of arrival from the technology families that make them available. The same technology can support several techniques, and the same technique can be applied to different signals.
The system acquires observations: radio signals, accelerations and rotations, magnetic-field variations, images, or proximity events. From them it derives distances, directions, characteristic fingerprints, or relative motion. Context is also needed before estimation, such as anchor locations, radio maps, sensor calibration, or building geometry.
An algorithm combines measurements and a motion model, estimates position and uncertainty, and then applies map constraints: non-crossable walls, corridors, stairs, lifts, and levels. Only after this chain does the result feed an application function, such as a blue dot, a route, an area alert, or an asset search.
04 TECHNOLOGIES AND TECHNIQUES
A family of tools, not one absolute winner
Physical technology, measurement technique, and algorithm are different choices that must be combined.
Radio technologies include BLE, Wi-Fi, UWB, and RFID, but they are not interchangeable. Android documentation on Wi-Fi RTT shows a ranging case: the device measures distance to compatible access points and can estimate its position through multilateration. BLE can provide proximity, strength, or direction; UWB supports fine time-based measurements; RFID covers identification and localization according to its tags, readers, and architecture.
Alongside radio there are geomagnetic fingerprints, inertial sensors, barometers, cameras, and map constraints. Fingerprinting compares an observed signature with a previously collected map; dead reckoning propagates motion from a known position; map matching removes or penalizes trajectories incompatible with the geometry. None of these terms alone identifies a complete system.
On the GiPStech website, the overview starts from Technology and continues through multi-sensor indoor positioning, visual positioning and VIO, adaptive hybrid localization, outdoor and vehicle continuity, augmented reality localization, and asset tracking with AsTra Wayra. These are complementary capabilities, not an absolute ranking.
A useful example of a specific limitation is Bluetooth RSSI alone: the GiPStech article shows why signal strength changes with the human body, orientation, reflections, and the environment. The point is not to exclude beacons, but to give them the right role within an architecture that manages instability.
05 MULTI-SENSOR FUSION
Why multi-sensor fusion matters
The IEEE survey on indoor fusion starts from a recurring limitation: in complex electromagnetic environments, a single source can lose accuracy or robustness. Fusion does not mean taking an indiscriminate average; it means estimating which observations are available, how reliable they are, and how they update the position state over time.
Inertial sensors describe rapid changes and motion continuity well, but their errors accumulate. A radio, magnetic, or visual reference can correct drift, while the map excludes impossible motion. The GiPStech Sensor Fusion overview illustrates this principle; adaptive hybrid localization extends it by selecting and weighting sources according to context.
Quality emerges from the whole chain: synchronization, calibration, motion model, outlier handling, uncertainty estimation, floor changes, and relocalization after an outage. The useful question is therefore not “which sensor is most accurate?” but “which combination keeps the service dependable under the site’s critical conditions?”.
06 FROM POSITION TO VALUE
From position to service
A coordinate becomes valuable when it changes an experience or a decision. For a person it can feed the dot on a map, the route to a department, an accessible exit, or contextual content. For an asset it can reduce search time, verify passage through an area, or connect location to an operational task.
Geofencing, presence, and analytics are not synonyms for positioning. They are functions that apply rules to the data: entering or leaving a zone, dwell time, flow aggregation, occupancy, or content activation. Likewise, augmented reality needs pose and spatial references, but also content, interaction, and application logic.
The Solutions pages show how position enters different contexts: healthcare, automatic presence detection, airports, Industry 4.0, retail, and public spaces. In each case, the target, update rate, infrastructure, and operational decision to support are different.
The articles on in-store AR navigation and the SILI 4.0 project show two very different examples: visual guidance on a smartphone on one side, and industrial localization applications on the other. They are specific cases that help explain how requirements and architecture change with the service.
07 DESIGN AND MEASUREMENT
How to choose and evaluate a solution
NIST guidance on testing reminds us that a fair comparison requires consistent scenarios, conditions, and metrics. Mean accuracy alone is not enough: error distribution, reliability, availability, latency, update rate, floor recognition, and behaviour during outages or context changes must also be observed.
- What must be localized, and which decision must it enable?
- Which accuracy, reliability, latency, and update rate are genuinely required?
- Which buildings, floors, critical areas, and operating conditions must be covered?
- Which infrastructure, calibration, and maintenance are acceptable?
- How are location data protected, retained, and integrated?
The project starts from the service and target: person or asset, available device, area and floors to cover, genuinely required accuracy, speed of movement, installation constraints, offline needs, and systems that must exchange data. Only then should the combination of technologies, maps, and algorithms be chosen.
Data governance and responsibilities must finally be made explicit: who can see location, at what granularity, for how long, on what legal basis, and with which security measures. The best solution is not the one promising the smallest number under ideal conditions, but the one that demonstrates on site that it meets the requirement with sustainable cost, maintenance, and organizational impact.
FROM REQUIREMENT TO FIELD TEST
Indoor positioning is a system decision
Describe the environment, who or what must be localized, and the service you want to enable. We can translate the use case into measurable requirements and a verifiable field test.
