Ultra-wideband (UWB) positioning is typically accurate to 10–30 cm where tags and base stations (UWB anchors) can see each other, and to several tens of centimetres in cluttered, metal-heavy halls, with tag-to-base-station ranges of a few tens of metres indoors.
Those figures hold only when the system is planned for the building. This guide explains what an accuracy figure says, what degrades it, how far UWB reaches and how to write a testable requirement.
What is the difference between accuracy and precision?
Accuracy is how close reported positions are to the true position. Precision is how tightly repeated positions cluster, whether or not they are in the right place.
A still tag whose positions scatter over 5 cm, all 40 cm east of the truth, is precise but not accurate. A useful system needs both, and a single datasheet number rarely says which one it describes. Ask four questions:
- Which statistic? A mean hides the outliers that cause wrong alerts. Ask for the 95th percentile (P95), the error that 95 % of positions stay within.
- 2D or 3D? Height is usually the weakest axis, because base stations sit at similar heights.
- Static or moving? A tag on a tripod scores better than a badge on a walking person.
- Where? An empty test hall is not a racked warehouse.
What affects UWB positioning accuracy?
UWB measures distance by timing radio pulses. Light travels about 30 cm in a nanosecond, so each nanosecond of timing error is about 30 cm of distance error. A bandwidth of 500 MHz or more gives pulses about two nanoseconds long, short enough to pick out the direct path before reflections arrive. Anything that hides that path or weakens the geometry costs accuracy.
| Factor | Effect | What helps |
|---|---|---|
| Geometry | Ranging errors are multiplied by the dilution of precision (DOP) | Surround the area; avoid straight lines |
| Non-line-of-sight (NLOS) | Blocked paths make distances read long, by tens of centimetres or more | Enough base stations that every point sees several directly |
| Multipath | Reflections matter little while the direct path is strong | Keep base stations clear of large metal surfaces |
| Human body | Absorbs signals at 6–8 GHz and shadows chest-worn badges | Wear badges high; add base stations |
| Metal | Racks and machines shadow aisles | Plan base stations per aisle |
| Update rate | Moving tags lag behind their reported position | Higher rate for fast vehicles |
| Filtering | Less jitter, more lag on turns | Tune per tag type |
| Calibration | Wrong base-station positions bias whole zones | Careful survey or automatic calibration |
Why does base-station geometry matter so much?
Each distance places the tag on a circle around a base station, and the position is where the circles meet. Circles crossing at wide angles barely move when a distance is slightly wrong. Circles crossing at shallow angles, as when all base stations stand along one wall, slide a long way for the same error.
Dilution of precision (DOP) is this multiplier: a DOP of 3 turns a 10 cm ranging error into roughly 30 cm of position error. In a straight line of base stations a tag can even appear mirrored on the wrong side.
What do bodies, metal and walls do?
A blocked direct path is the main source of large errors indoors. The signal arrives either through material, which slows it, or by a longer reflected route. Both make the distance read long, never short.
The human body absorbs UWB strongly, so a chest badge can lose every base station behind its wearer. People-tracking therefore needs more base stations than asset tracking for the same accuracy. Metal racks and machines leave shadows, especially in long aisles.
How do update rate and filtering change the picture?
The update rate does not improve a single position, but it decides how old the latest one is. A forklift at 10 km/h covers about 2.8 m per second, so at 1 Hz its reported position can trail by up to 2.8 m; at 10 Hz, by about 28 cm. Filters smooth jitter but add delay, so use light filtering for alerts and heavier filtering for analytics.
How accurate is UWB in different environments?
The table gives indicative 2D accuracy at the 95th percentile for systems planned for their environment. Poor geometry or too few base stations can make results several times worse.
| Environment | Typical conditions | Indicative accuracy (P95, 2D) |
|---|---|---|
| Sports field or open yard, base stations around the edge | Line of sight almost everywhere | 10–30 cm |
| Open production or assembly hall | Mostly line of sight, some machines | 20–50 cm |
| Warehouse with high racking | Aisles shadowed by racks and goods | 30 cm–1 m, better with base stations planned per aisle |
| Hospital, office or care home | Walls and doors between rooms | Reliable room and zone level; tens of centimetres inside well-covered rooms |
| Badges worn by people, any site | The wearer’s body shadows the badge | Noticeably worse than tags with a clear view |
How far does a UWB signal reach?
EU and US rules limit UWB to a very low power density, a mean of −41.3 dBm/MHz, which protects other radios and keeps the range short. Indoors with clear line of sight, a tag and a base station can usually range over a few tens of metres; larger datasheet figures come from open space. Plasterboard, wood and glass let the signal through with some loss, brick and concrete weaken it strongly, and metal blocks it.
Range is not coverage. A 2D position needs at least three base stations in view and a 3D position at least four, ideally from different directions, so base stations sit much closer together than their maximum range. On a square grid, one base station per 200 m² means about 14 m between neighbours; one per 500 m² means about 22 m.
How should you specify an accuracy requirement?
Start from the decision the data supports, not from the best number on offer. Over-specifying accuracy adds base stations without adding value.
| What you need to know | Accuracy usually enough |
|---|---|
| Which room or zone a person or asset is in | 1–3 m |
| Which aisle or lane a vehicle is in | 0.5–1 m |
| Which bay, workstation or pallet slot | 0.3–0.5 m |
| Which tool at a workstation, or an athlete’s running line | 10–30 cm |
Then write the requirement so that it can be measured:
- Name the zones. Open floor, racked aisles and yards each get their own figure.
- State the statistic. For example “95 % of 2D errors below 50 cm”.
- State the tag and its motion. A forklift roof tag at 10 km/h is not a badge on a walking person.
- State update rate and availability. For example “at 2 Hz, with a position 99 % of the time”.
- Agree the test. Surveyed reference points, normal working conditions, results per zone. ISO/IEC 18305 gives a common vocabulary.
Where to go next
To see how the measurement method affects accuracy, battery life and capacity, read TWR vs TDoA vs AoA. To turn an accuracy requirement into a base-station layout, continue with planning a UWB RTLS deployment and try the base station estimator. Terms such as NLOS and DOP are explained in the glossary, and if you want to test accuracy on your own site, book a demo.
Frequently asked questions
How accurate is UWB indoors?
What is the range of UWB?
Does UWB need line of sight?
Can UWB go through walls?
What is dilution of precision (DOP)?
Does a higher update rate make UWB more accurate?
How can I test accuracy before buying?
Sources and further reading
- ISO/IEC 18305:2016 Real time locating systems: test and evaluation of localization and tracking systems
- R. B. Langley, Dilution of Precision, GPS World, May 1999
- D. Dardari et al., Ranging with ultrawide bandwidth signals in multipath environments, Proceedings of the IEEE, 2009
- 47 CFR § 15.517: technical requirements for indoor UWB systems (US)
- Commission Implementing Decision (EU) 2019/785 on radio spectrum for ultra-wideband equipment
Published . Last reviewed by the Noccela engineering team.