Five Tips to Prevent AR Drift: Practical Accuracy Improvements for the Field

Table of Contents

• What is AR drift?

• Tip 1: Minimize error with high-precision positioning

• Tip 2: Correct device heading and orientation

• Tip 3: Create a stable AR tracking environment

• Tip 4: Perform careful initial alignment

• Tip 5: Use advanced technologies to address drift at the root

• Simple surveying with LRTK

• FAQ

What is AR drift?

Use of AR (augmented reality) technology is progressing in construction, civil engineering, and other field sites. Simply pointing a smartphone or tablet to overlay digital information—such as a 3D model from the drawings or routing of pipes—onto the real scene makes it easy to share the finished image on site or give work instructions intuitively. However, when AR is actually used on site, a common problem is that “the display is misaligned with reality.” This phenomenon, called AR drift, refers to virtual models overlaid in AR gradually shifting away from their real-world positions over time or as the user moves.

Why does AR drift occur? Major causes include the following. One is low device positioning accuracy. Standard built-in smartphone GPS can have errors of several meters, and that positional error directly causes the virtual object to be displayed in the wrong place. A second cause is device heading and orientation errors. If a smartphone’s electronic compass points incorrectly or the gyroscope exhibits drift (tiny errors accumulating over time), the displayed direction will be off and model alignment will break down. A third cause is unstable tracking due to the surrounding environment. AR apps estimate device pose by detecting feature points from the camera feed, but in monotone scenes with few features, in dark areas, or when viewing glass surfaces, tracking becomes unstable and models can slowly drift or jump. A fourth cause is initial alignment error. When placing a model in AR, the initial alignment is sometimes done manually, and even a small misalignment at that stage causes overall inconsistency. In larger sites, slight angular or positional errors become large offsets at a distance. A fifth cause is time-dependent drift: sensor cumulative errors and environmental changes can cause AR displays to gradually diverge over time.

These factors can cause the AR 3D model to disagree with the real-world position (the so-called “misalignment”), and on-site personnel may end up distrusting the technology. For example, if an AR marker indicates “drill a hole here” but the actual location is off by several dozen centimeters, there is a risk of drilling in the wrong place. Workers may end up remeasuring with tape or string line, negating the benefit of using AR. Preventing AR drift is therefore essential to fully leveraging AR’s advantages. Below are five concrete tips to improve AR display accuracy on site.

Tip 1: Minimize error with high-precision positioning

The first tip to prevent AR drift is to use the most accurate positioning information available. A smartphone’s standard GPS can have errors of about 5-10 m (16.4-32.8 ft), and that error directly translates into virtual model misplacement. An effective countermeasure is to utilize centimeter-class positioning technology such as RTK-GNSS (cm level accuracy, half-inch accuracy). RTK (real-time kinematic) uses correction information from a base station to correct satellite positioning errors in real time, dramatically reducing errors from several meters to a few centimeters.

For example, with an RTK-capable positioning device, you can determine a device’s position outdoors with centimeter-level accuracy and place AR objects based on absolute coordinates. With high-precision positioning, virtual models are unlikely to shift even as users walk around the site, so the AR display remains consistently correct and its reliability increases dramatically. The more accurate the positioning information, the less effort is required for fine-tuning initial placement or later corrections. For wide-area sites that use AR, consider introducing such high-precision positioning. Recently, compact RTK-GNSS receivers that can be attached to smartphones have appeared, making centimeter-level positioning—previously possible only with specialized surveying equipment—easier to achieve.

Tip 2: Correct device heading and orientation

The second tip is to accurately correct the device’s (smartphone or tablet) heading and orientation. AR apps determine where to display virtual objects based on the device’s orientation and tilt; if the device’s heading sensor (electronic compass) is off or the gyroscope exhibits drift, the displayed model will not align with reality. This is especially important outdoors, where magnetic interference from steel frames or heavy machinery can cause the compass to indicate the wrong north.

Compass calibration is a basic task before using AR. Hold the smartphone and move it in a figure-eight motion to correct electronic compass deviation. When you move to a different spot on site where you plan to use AR, check again that the compass points correctly. Because the compass can be disturbed by metal objects or power lines, it’s a good idea to step away and recalibrate if possible. Also, if you use AR for a long time and sensors accumulate drift, restarting the app to reset sensor values can be effective. Some devices include high-performance IMUs (inertial measurement units) with superior attitude estimation; for AR use, choosing newer or higher-end models can help.

Tip 3: Create a stable AR tracking environment

The third tip is to prepare an environment in which AR tracking is stable. AR tracking estimates the device’s position and orientation in real time by detecting objects and patterns (feature points) in the camera image. If tracking becomes unstable, the device can lose its pose, causing AR models to drift away from reality. To maintain tracking accuracy, pay attention to the following environmental factors.

• Is the scene rich enough in feature points?: Monotone walls or floors lacking patterns provide few feature points, making it easy for AR to lose tracking. When necessary, place temporary markers or distinctive objects, or change your position slightly to include a more feature-rich background.

• Brightness and lighting stability: Extremely dark environments or strong backlighting make it hard for the camera to capture clear images and detect features. On site, add lighting or change the time of use to maintain appropriate brightness for AR.

• Measures for glass, mirrors, and other reflective surfaces: Pointing the camera through glass or having mirrors in view can cause AR to confuse real objects with their reflections. In reflective environments, adjust the camera angle or temporarily cover reflective surfaces if needed.

• Device performance: Tracking depends on the device’s computing power. Older devices may lag and degrade accuracy, so use the latest-generation devices or those with strong AR support where possible. Many recent devices include a LiDAR scanner (a laser-based depth sensor); such depth sensors help detect planes and understand space, stabilizing tracking.

By addressing the above points and preparing the environment, the AR’s self-localization can remain accurate and virtual models will stick closely to their intended real-world positions. Conversely, ignoring environmental factors increases the chance that even a high-performance AR app will gradually drift.

Tip 4: Perform careful initial alignment

The fourth tip is to carefully perform the initial alignment of the AR display. When placing 3D models or drawings in real space with an AR app, there is an initial step to align position and orientation. If misalignment occurs here, no amount of tracking or positioning accuracy afterward will restore consistency with reality. This is especially true when displaying large models over wide areas: small angular or scale errors in the initial alignment can translate into large positional deviations at distant points.

A way to improve initial alignment accuracy is to align multiple on-site reference points. For example, when displaying a building layout in AR, identify corresponding corners or landmark points of existing structures on the model and fine-tune so that position and orientation match at two or more points. This produces better overall consistency than aligning to a single point. If possible, using pre-surveyed control point coordinates is also effective. By obtaining coordinates of known on-site points (such as building foundation corners) with surveying equipment and aligning the model to those coordinates, you can achieve a more reliable initial alignment. Note that this method requires specialized knowledge and effort, so introduce it as appropriate for the site.

In short, the initial step is crucial. If you properly match the model to reality at the start, AR displays will remain stable during subsequent work. Conversely, sloppy initial setup will eventually lead to drift and loss of trust. Make it a habit to take the time for careful initial alignment when using AR on site.

Tip 5: Use advanced technologies to address drift at the root

The final tip is to adopt the latest technologies to fundamentally address AR drift. In recent years, various technological innovations have appeared to solve AR misalignment. In addition to the previously mentioned RTK-GNSS, techniques such as VPS (visual positioning service) and cloud anchors are becoming available. VPS estimates highly accurate self-location by comparing the camera view to a database of spatial features stored in the cloud. Cloud anchors are shareable reference points in AR space; once placed, they allow virtual objects to be displayed in the same location when revisited. Using these methods makes it easier to maintain alignment over wide areas and suppress drift during long-term operation.

Comprehensive solutions combining hardware and software have also emerged. For example, systems that combine a small RTK-GNSS receiver attachable to a smartphone with a dedicated AR app provide end-to-end high precision from positioning to AR display. These systems greatly reduce the effort of initial alignment and deliver AR that does not shift at all even when users move. In other words, a time has come when technology can fundamentally eliminate AR misalignment without relying on tedious marker placement or repeated resets.

Of course, usability and cost must be balanced for field use, but to maximize the value of AR it is important to proactively consider adopting such advanced technologies. Rather than merely “making do” with on-site workarounds, adopting a technology-first mindset to solve the problem will significantly expand AR’s practical potential.

Simple surveying with LRTK

Based on the points above, LRTK is attracting attention as a decisive solution to prevent AR drift. LRTK is a new solution developed to fundamentally solve on-site AR misalignment. Concretely, LRTK combines an ultra-compact RTK-GNSS receiver that can be attached to a smartphone with a dedicated AR app, allowing centimeter-level high-precision positioning data to be fed into AR so 3D models align precisely with real space. This ensures that virtual models remain correctly positioned as users move around the site. The bothersome marker placement and repeated position resets are unnecessary, and once set up even wide-area sites can achieve “drift-free AR.”

Moreover, LRTK can be used as a simple surveying tool. With the dedicated app, you can record coordinates at a tapped point with centimeter-level precision, measure distances and height differences, and perform tasks that previously required specialized surveying equipment or multiple people using only a smartphone. For example, you can mark stake positions on site based on design coordinates in the drawings, or measure and verify the as-built shape immediately after construction—anyone can perform these tasks with intuitive operations. The system is designed so that no advanced surveying knowledge is required; after registering on-site control points during initial setup and defining the coordinate system, you can start using the app and AR immediately.

LRTK dramatically improves on-site positioning accuracy and work efficiency. By seamlessly overlaying digital construction data onto reality without suffering from AR drift, it is expected to be a trump card for “on-site DX” (digital transformation). If AR drift is a problem for you, consider evaluating LRTK. For more details, check official LRTK information.

FAQ

Q: What is AR drift? A: AR drift refers to the phenomenon where virtual objects displayed in AR appear shifted from their intended real-world positions. Causes include device GPS and sensor errors and tracking failures due to the surrounding environment. When the displayed position gradually changes over time or as the user moves, this is described as AR drift.

Q: Can using markers or QR codes prevent AR misalignment? A: Placing markers (image markers) or QR codes on site and using them as references for AR alignment can temporarily improve accuracy. However, maintaining stable accuracy across wide outdoor sites is difficult. Markers are effective only within the area where they are placed, and they require scanning by the camera each time. They may peel off or become dirty in wind and rain, creating detection risks. As a fundamental measure, technologies that maintain high-precision alignment over wide areas without relying on markers—such as RTK-GNSS—are more effective.

Q: Does using LRTK require specialized knowledge or special equipment? A: LRTK is designed so site personnel can operate it. Basically, you need only a smartphone (or tablet) and a small LRTK-compatible GNSS receiver. The dedicated app is intuitive and does not require surveying expertise. Registering on-site control point coordinates during the first setup takes some familiarization, but it can be learned after a few uses. No complex configurations or bulky equipment are necessary, and one smartphone per person can deliver high-precision AR and surveying.

Q: How accurate is positioning with LRTK? A: Depending on environmental conditions, LRTK typically achieves positioning accuracy on the order of a few centimeters (cm level accuracy, half-inch accuracy). Compared with standard smartphone GPS errors of several meters, LRTK delivers accuracy that is a small fraction of that. In open-sky conditions with good satellite reception, repeated measurements at the same point yield highly reproducible coordinates. Accuracy may decrease somewhat under overpasses or under tree cover where satellite signals are weak, but it still significantly outperforms conventional GPS.

Q: Can LRTK be used indoors? A: Because LRTK uses GPS satellites for high-precision positioning, it is primarily intended for outdoor use. In buildings or underground areas where GPS signals do not reach, it is difficult to ensure high accuracy at present. However, there are cases where you can use outdoor-surveyed control points near a building to derive relative indoor positions, so with some ingenuity LRTK can assist in partial indoor positioning. As indoor positioning technologies evolve and integrate with LRTK, indoor use will expand.