Skip to content
Select language:English

Indoor positioning: how it works where GPS cannot reach

Updated 4 min read

Indoor positioning estimates where a person or object is inside a building, using locally available observations when satellite reception is insufficient. A position can help people find their way or locate equipment, but accuracy, update behaviour and coverage must be assessed in the real environment: no single sensor guarantees them.

Where GPS does not reach, a position inside a building is estimated from the observations available on site

GPS belongs to the family of global navigation satellite systems (opens in a new tab), or GNSS. Roofs and walls can obstruct signals; reflections can alter their paths. GPS.gov explains these limitations (opens in a new tab): a position shown on a phone does not prove that it is accurate enough to distinguish neighbouring rooms.

An indoor system therefore seeks different or complementary references. It must also relate the estimate to the correct building, floor and map. A Wi-Fi (opens in a new tab) network available for communication is not, by itself, a prepared positioning service.

Sources have different contributions and limitations; the choice depends on the target and available devices.

FamilyContribution and conditions
Inertial sensors (opens in a new tab)Accelerations and rotations help follow motion. Errors accumulate, so references are needed to correct the estimate.
Magnetic fieldMeasured variations can be compared with a map of the site’s characteristics. References must be collected and checked, accounting for changes to the building.
Radio signalsBLE (opens in a new tab), Wi-Fi (opens in a new tab), UWB (opens in a new tab) and RFID (opens in a new tab) provide different observations. Devices, infrastructure and measurement techniques determine what can be derived.
CameraScene features can help recognize a place and follow motion. Lighting, viewpoint and visibility must be suitable; uniform surfaces or obstructions reduce the available information.

This is not a product ranking. Architecture and integration details are in the technologies overview.

Sensor fusion (opens in a new tab) builds a common estimate from complementary observations. For example, inertial motion can connect two radio references; a later correction can constrain accumulated error. Combining data does not mean indiscriminately averaging it: quality, calibration and available references matter.

GiPStech describes this approach in multi-sensor indoor positioning (opens in a new tab). Combining sources can reduce reliance on one of them, but does not remove site preparation or guarantee identical performance in every building. Vision is also a project choice, rather than a requirement for every indoor solution.

Position, tracking and navigation are different results

Section titled “Position, tracking and navigation are different results”

NIST distinguishes localization from tracking (opens in a new tab): the former estimates location, while the latter follows it over time. A previous record is not equivalent to a current observation. A zone or floor also answers a different question from a point coordinate.

Navigation adds the search for a route to a destination. Useful guidance also needs a consistent map and a comprehensible representation of the space: an estimated point alone does not describe doors, stairs or accessibility.

In a multi-storey building, a technician needs to find a mobile item of equipment. Knowing that it is in the building may be insufficient; identifying its floor and zone could already narrow the search. If it moves quickly, the time of the latest location update also matters.

The initial requirement is therefore an operational decision: which information enables the technician to act? Granularity, update behaviour and devices are selected afterwards. This fictitious example does not assume automatic tracking of every object without a device or prepared references.

NIST stresses the need for representative testing (opens in a new tab): materials, floors, infrastructure and ways of moving affect results. A stationary point in an empty corridor does not represent every use.

QuestionWhat to check in the trial
Is the position sufficient for the task?Error and relevant worst cases, rather than just a mean value.
Are building, floor and zone recognized?Transitions between levels and areas with similar layouts.
Does information arrive in time?Update interval and delay relative to motion.
Where is a useful estimate missing?Coverage and behaviour during signal loss or service recovery.
Is the result repeatable?Representative routes and devices, handheld or pocket use, site conditions and reference maintenance.

ISO/IEC 18305:2016 (opens in a new tab) defines metrics and scenarios for testing and evaluating localization and tracking systems, primarily indoors. It provides a framework for measurements and their presentation: citing it does not by itself demonstrate that a solution meets the project’s requirement. Acceptance thresholds should be set before testing.

In operational applications, position has value when it helps find an asset and understand its context. Asset2Work connects assets with locations and integrates positioning in the app: the technology enables the service and should be assessed alongside the project’s data, map and process. An asset’s last associated location is not a universal guarantee of a live detected presence.

The reserved documentation on assets and locations explains the operational model. For public background, read What is indoor positioning? (opens in a new tab); the technology comparison and asset tracking cover related choices.