AR drift problem in pile-driving guidance: How to prevent pile position misalignment?

Premise

• You are an AI that translates Japanese into natural, readable English.

• Please translate only the provided text into English. Maintain Markdown headers, lists, line breaks, and inline code exactly as they are, and do not add summaries or supplementary explanations.

• As the text will be used for an article, translate as faithfully as possible for SEO purposes.

• Output only the English-translated text (plain text), and do not return JSON or any extra text.

[Unit notation rules for translation]

• Do not convert speeds (m/s), temperatures (℃), masses (kg), etc.

• In addition to the SI notation in the original text, add the imperial equivalent in parentheses: 〇〇 m (△△ ft), 〇〇 cm (△△ in), 〇〇 mm (△△ in)

• There must be a half-width space between the number and the unit, and use units ft / in (do not add a plural s)

• Rounding must not exceed the number of decimal places in the original text - For integer m, round ft to 0–1 decimal place (examples: 1 m → 3.3 ft / 3 ft, 10 m → 32.8 ft / 33 ft) - For m with 1 decimal place, ft to 1 decimal place (example: 1.5 m → 4.9 ft) - For m with 2 decimal places, ft to 2 decimal places (example: 1.23 m → 4.04 ft) - 30 cm → in use one decimal place (11.8 in) - 12 mm → in use two decimal places (0.47 in)

• ± and ranges should be accompanied by the corresponding values in the same format and with the same number of digits (examples: ±2 cm (±0.8 in), 1-2 m (3.3-6.6 ft))

• "cm精度" should be accompanied by cm level accuracy (half-inch accuracy)

Please translate the following input into English.

Table of Contents

• What is pile-driving guidance using AR

• What is the AR drift problem

• Risks and impacts of deviations at pile-driving sites

• Main causes of AR drift

• Conventional countermeasures and their limitations

• Fundamental solution through high-precision positioning

• Precise pile-driving guidance achieved by simple surveying with LRTK

• FAQ

What is AR-based pile-driving guidance?

In recent years, initiatives to utilize AR (augmented reality) technology at construction sites to guide pile driving (installation of foundation piles) have attracted attention. Conventionally, surveyors determined pile positions based on design drawings and marked the ground with layout markings or chalk lines to indicate pile centers. With AR pile-driving guidance, the design pile positions are overlaid on live site images captured by smartphone or tablet cameras. Site staff only need to place piles following the virtual markers visible on the screen, greatly streamlining the cumbersome surveying work that used to require tape measures or transits.

By using AR-guided pile driving, even inexperienced workers can intuitively understand "where to drive the piles," so improvements in construction accuracy and increased work efficiency can be expected. It can prevent positional marking mistakes on drawings and reduce rework caused by placement misalignments. However, AR also has its own challenges. A prime example of this is the phenomenon called 'AR drift'. In the next chapter, we will explain this AR drift issue in detail.

What is the AR drift problem

AR drift refers to the phenomenon where AR-rendered displays drift away from their real-world positions over time or with movement. For example, a virtual marker for a stake that should be on the ground may appear slightly offset horizontally or appear to float on the screen. A virtual object that was initially correctly aligned can gradually lose its position and orientation as the user walks around—this is AR drift. Put simply, it can be described as the 'misalignment' or 'drift' of AR displays.

When AR drift occurs, the reliability of the position information shown in AR is compromised. If the virtual overlay shifts even by a few centimeters (a few in) from the intended installation position, it can be critical for precision construction. Especially for tasks where positional accuracy is important, such as pile driving, misalignment in AR can lead to a serious error known as pile position misalignment. In severe cases, the display can be off by tens of centimeters or more (tens of in or more), and in such cases you can no longer trust AR guidance and end up having to re-check positions with a tape measure or surveying instruments. In other words, unless the AR drift issue is resolved, the AR pile-driving guidance that was introduced risks becoming unusable in practice.

Risks and Impacts of Deviations Occurring at Pile-Driving Sites

A deviation in pile positions at a construction site is a serious risk that directly affects the quality and safety of structures. Foundation piles are critical elements that support buildings and structures, and precise positions and spacing are established during the design phase. For example, even a deviation of just a few centimeters (a few inches) from the specified position can cause interference with adjacent structures or affect the overall load-bearing balance of the foundation. Deviations in pile positions that exceed allowable tolerances require corrective measures or reinforcement work as construction defects, leading to extended construction schedules and increased costs.

The occurrence of deviations in AR-guided pile-driving not only invites these risks but also affects on-site credibility. If workers trust the AR display and set piles accordingly, only to find upon re-survey that the positions do not match, distrust of the AR technology itself will increase. A system that has lost trust will no longer be used on site, wasting the investment in digital technology. Conversely, if AR display deviations can be prevented and consistently accurate position guidance provided, site personnel can confidently use AR, which directly leads to improved construction quality and efficiency. Since AR is used for pile-driving guidance, suppressing pile position deviations caused by AR drift is extremely important.

Main causes of AR drift

So, why does AR drift occur? The causes can mainly be classified into the following factors.

• GNSS (GPS) accuracy limits: Position measurements from GPS built into smartphones and tablets commonly have errors on the order of a radius of several meters (several ft). Because staking positions on design drawings require accuracy to within several centimeters (several in), if the device’s position information is off by several meters (several ft), large discrepancies will occur already at the initial display stage.

• Sensor (bearing/attitude) errors: A smartphone’s electronic compass (magnetometer) and gyroscope also include small errors and drift. For example, if the bearing is off by 2 degrees, an object 10 m (32.8 ft) away will be displayed tens of centimeters (several in) off in position. Slight errors in the device’s tilt angle also appear as height or positional offsets of distant virtual objects.

• AR engine tracking errors: AR apps estimate device motion by capturing feature points in the camera image. However, tracking becomes unstable in feature-poor, monotonous scenes, and slight computational errors accumulate when moving over wide areas. As a result, even if virtual objects are aligned correctly at first, they may become slightly misaligned after movement.

• Environmental factors and coordinate inconsistencies: Surrounding steel structures or strong magnetic fields can disturb electronic compass readings. Also, if the design data and the site’s survey coordinate system don’t match, the model can appear to be translated even when placed in the correct location. Specifying the wrong reference point on the drawings or failing to perform coordinate conversion on sites using a local coordinate system can cause large positional offsets.

As described above, device-specific positioning and sensor errors, tracking errors inherent to AR technology, and issues with the environment and data settings interact in complex ways to cause AR drift. Conversely, if these factors are eliminated one by one, the potential to suppress AR drift becomes apparent.

Conventional countermeasures and their limitations

Various measures have been implemented in the field to prevent AR drift. Let’s look at the main countermeasures and their respective limitations.

• Device sensor calibration: Before using the AR app, you calibrate the smartphone's electronic compass and reset the gyroscope reference. Specifically, this involves swinging the device in a figure-eight to adjust the magnetic sensor and placing it on a level surface to correct gyroscope drift. These steps can reduce sensor errors to some extent but cannot eliminate them completely. As you move or as time passes, drift accumulates again.

• Alignment using markers or known points: This method places QR codes or marker boards at key locations on site and reads them with the camera to correct the virtual model's position. Alternatively, the design model can be overlaid on real photos in the AR app and manually fine-tuned to match known points such as building corners or boundary stakes. These techniques can correct initial offsets to some degree, but when moving across a wide area they can cause viewpoint-dependent drift to reappear, and they require the hassle of placing and reading markers repeatedly at multiple locations.

• Combination with basic surveying: Instead of relying solely on AR, you measure critical points with a transit or GPS as usual, then adjust the AR display by applying offsets to match those measurements. For example, you might survey only one reference point used in the design on site and apply that difference to the entire AR model. This improves accuracy, but once in the end, surveying is necessary, it can no longer be called a "convenient AR display." On site, preparation of surveying equipment and specialized skills are still required.

The measures above have some effect, but they are not a fundamental solution. Manual calibration and marker placement are cumbersome and increase on-site workload. Misalignments caused by device limitations inevitably remain, so they do not achieve "completely eliminating drift." Traditionally, each site relied on ingenuity to "somehow suppress the drift," and there was no decisive solution to the AR drift problem.

Fundamental solution through high-precision positioning

The key to fundamentally eliminating AR drift is the adoption of high-precision positioning technology. Among these, high-precision GNSS (satellite positioning) known as the "RTK method" is the most effective means for countering AR drift. RTK (Real Time Kinematic) is a technique that dramatically increases the positioning accuracy of GPS and similar systems by using correction data from a base station, achieving horizontal positioning of ± a few centimeters (± a few in) and vertical positioning also on the order of a few centimeters (a few in), compared to ordinary GPS with errors of several meters (several ft).

If a device can determine its position accurately to the centimeter (cm level accuracy, half-inch accuracy), aligning a virtual model with the real world becomes significantly easier. If a smartphone knows its absolute coordinates almost perfectly, then just "displaying a virtual stake marker at the design coordinates" will make the real and virtual positions nearly coincide. The idea is that the alignment of drawing data and on-site positions, which used to be a struggle, will be handled automatically by machines and come together smoothly.

Furthermore, high-precision positioning information also suppresses drift in the AR engine. By continuously feeding the absolute coordinates obtained by RTK back to the AR system, misalignments caused by camera tracking errors and sensor errors can be automatically corrected. As a result, even when users walk around, occurrences like "virtual models drifting in midair" are less likely to happen. It also prevents situations where a virtual object that was once aligned drifts out of place on its own and requires readjustment.

High-precision satellite positioning used to require specialized equipment and expertise, but in recent years simple RTK solutions that can be used with smartphones have emerged. By attaching a dedicated compact receiver to a smartphone and receiving correction information, anyone can now perform centimeter-level positioning in real time on site (half-inch accuracy). By leveraging such technology, the ideal situation where “a model in AR perfectly aligns with reality and does not drift even when moved” can be achieved.

Precision stake-driving guidance achieved through simplified surveying with LRTK

One example of a solution that makes high-precision positioning easy to use on site is LRTK. LRTK consists of an RTK-capable GNSS receiver that can be attached to a smartphone and a cloud service, and is designed so that even non-specialist surveyors can easily perform centimeter-level positioning (cm level accuracy, half-inch accuracy) and AR display. By incorporating "simplified surveying" with LRTK, you can resolve the AR drift issues introduced earlier all at once.

For example, if you pre-register the coordinate values of each stake from the design drawings into an LRTK app, a virtual stake position marker will be displayed on your smartphone screen at the exact location on the ground when you are on site. With ordinary GPS the initial display can be off by several meters, but thanks to high-precision positioning the model appears in an almost exactly correct position from the start. After that, simply place the stakes at the positions shown on the screen, and you can lay out the design positions without using surveying instruments.

Also, because LRTK provides continuous high-precision coordinates even while moving, there is almost no concern that a virtual stake marker once aligned will progressively drift as you walk around. The AR stake display remains continuously fixed at its designated position on the ground, preventing the situation where "you notice the virtual marker has moved away from the actual position." This is precisely what can be called "drift-free AR staking guidance".

Furthermore, by utilizing LRTK, stake-positioning tasks that previously relied on skilled surveyors can be carried out by the on-site staff themselves. This is because, by following the guidance on the smartphone screen and simply moving according to guide displays such as "Move ○ cm (○ in) to the north," anyone can reach the accurate position. Expensive surveying equipment and complicated procedures are not necessary. Because a single person can mark stake positions in a short time, effects such as reductions in personnel and shortened work times can also be expected.

By incorporating simple surveying with LRTK in this way, the accuracy and reliability of AR pile-driving guidance improve dramatically. By solving the AR drift problem, AR displays that always show the correct position are no longer a "fun demo" but a construction support tool fit for practical use. To fully leverage digital technology on site, the use of high-precision positioning solutions will be the key going forward.

FAQ

Q. What preparations and equipment are needed to perform pile-driving guidance using AR? A. Basically, you need a smartphone or tablet that can be used on site and a compatible AR app. Recent smartphones and tablets have AR functions built in, so AR alignment displays are possible without dedicated surveying instruments. However, to improve accuracy, it is ideal to have a high-precision GNSS receiver that can connect to the smartphone (for example, an RTK-capable device such as LRTK). It is also important to preload design data, such as the pile installation coordinates, into the app and prepare by aligning it with the site’s coordinate system.

Q. Is AR-guided pile driving possible with only 2D drawing data? A. Yes, it is possible. Even without a 3D model, you can indicate pile locations by projecting points and lines from the plan onto the ground using AR. For example, by extracting pile center positions from the design drawing’s CAD data or from a coordinate list and marking those locations in the AR app, you can confirm on site where piles should be driven. If a 3D model is available, you can also check vertical positions and perform clash checks, but for the purpose of guiding pile positions, 2D information is sufficient.

Q. How accurate is AR-based pile location display? A. If you use only a standard smartphone’s GPS and AR features, positional deviations can range from tens of centimeters (several inches) to, in some cases, over 1 m (3.3 ft). While this may suffice for rough checks, it is not accurate enough for precise pile layout. On the other hand, combining with high-precision GNSS (RTK) can reduce errors to within a few centimeters (a few inches). In practice, using an RTK-enabled solution can keep horizontal position within a few centimeters (a few inches) and vertical position to around a few centimeters (a few inches) of error, allowing AR display of pile locations with construction-grade accuracy (centimeter-level accuracy (half-inch accuracy)).

Q. Is specialized knowledge or training required to use AR pile-driving guidance effectively on-site? A. No special CG creation skills or a surveying license are required. The basic procedure is simply selecting the model and point data on a smartphone and aligning them following the on-screen guidance. Even when using high-precision GNSS, the system automatically determines position and adjusts model placement, so users only need to confirm the on-screen prompts. With a simple, roughly half-day operational training for field staff, they should be able to make full use of it in their daily work.

Q. Do I need to install markers such as QR codes in advance or perform control point surveying on site? A. When using high-precision GNSS, special marker installation is generally not necessary. Because satellite positioning allows the smartphone itself to determine its position with high accuracy, alignment can be completed simply by displaying the model at the designated coordinates. However, in environments where GPS cannot be used, such as indoors, alternative measures are required (for example, AR markers or manual alignment to known points). For large outdoor sites, coordinate alignment using RTK positioning such as LRTK is the most efficient, allowing virtual models to be matched to real-world positions quickly and reliably.