Smart glasses in industry: Matching controls to the task

What smart glasses display

Smart glasses are head-worn devices that display digital information in front of one or both eyes. How this information relates to the surroundings differs between devices and applications. If a device value is also received, integration through the IoT gateway must connect it with its source and the correct work step.

In assisted reality, a small display can show an inspection instruction, a measured value or a video image. The information is not spatially anchored to a real component. Spatial augmented reality, in contrast, can display a marker or digital object so that it appears assigned to a location in the surroundings. The application must establish and maintain this spatial relationship during use.

An AR headset is therefore one possible device class for industrial tasks. For a short instruction or remote guidance, a display without spatial anchoring may be sufficient. The decision starts with the information needed during the task.

Comparing the devices studied

The following overview distinguishes six devices or configurations from the cited publications. It describes the study setups at the time and the questions derived for a trial of your own. It does not state which devices 420+ supports or approves for a particular use. A transparent display alone does not mean that content is spatially anchored to a component.

Comparison table: scroll horizontally if needed

The last column contains editorially derived test questions, not demonstrated advantages of the devices. For the Navigator in the LEGO study, the bibliography refers to the Navigator 500, while the methods section names only the device family. Its results are therefore not attributed to the Navigator 520. The two HoloLens configurations also remain separate: a virtual manual at an assembly station and remote guidance with a hard hat in a moving work environment place different demands on the user.

An inspection point through smart glasses

A fictional example: Someone inspects an assembly for visible damage. They hold the part, change its position and examine a defined area. The smart glasses display the relevant inspection instruction. The person conducting the inspection judges the area observed.

Four elements must fit together in this step:

A photo can provide supplementary documentation of an unusual area. Whether it shows the relevant area and supports a finding must be assessed separately. Taking the photo alone does not complete the visual inspection.

If the person cannot clearly classify the area, the planned workflow must allow an appropriate response. This may include asking for clarification or repeating the inspection. With Human in the Loop, human judgement must be able to influence what happens next. A forced pass result would conceal the uncertainty.

Display, input, images and analysis serve different purposes

The display helps someone perform the work step. Input records an entry. An image capture provides image material. Subsequent analysis can examine that material or the recorded data. Whether an application provides all four functions and how they connect must be established for the specific use.

This distinction is particularly important for speech. Recognition of a spoken value initially says only what content the application captured. The person must be able to tell whether the value was captured correctly and how an incorrect entry can be corrected.

A development study on data capture with a Vuzix M400 in emergency care describes precisely these interaction issues. The prototype used voice commands and hand gestures. Participants reported, among other issues, the need to correct transcriptions of specialist abbreviations. The study offers design insights; effects on actual emergency care workflows, attention or workload were not measured. The research team also extended training between development phases.^[2]

This yields a concrete question for industrial trials: Can the person check and correct the captured entry while safely continuing their actual task? For subsequent use, the context model must connect the entry to the relevant task, inspected object and applicable instruction.

What different experimental results actually show

Research does not provide a consistent ranking of smart glasses, tablets and paper. Each study examines a particular setup with particular people and tasks. Three examples show why the comparison medium and statistical interpretation belong with the result.

Comparison table: scroll horizontally if needed

The LEGO participants had no prior experience with smart glasses and received a short introduction. In the computer component experiment, both instruction presentation and interaction differed alongside the devices. The repair study used a more extensive assistance system. These conditions cannot be combined into a common effectiveness measure.

The differences offer clues for selection. They do not, however, isolate how much of a result comes from the device, instructions, interaction or practice. In digital work instructions, too, the repair study's result concerns the entire workstation examined.

Your own trial should therefore consider completion time, errors and interaction effort separately. A workflow may produce fewer errors and take longer at the same time. Whether this improves the task depends on which errors need to be prevented and where the additional effort arises.

Spatial overlays or a small display

A spatial overlay can help locate an area on an object. At the same time, it requires the person to perceive both the display and the real surroundings sufficiently. A small display showing text or video places different demands on field of view and navigation. Both concepts must be tested in the intended movement space.

A study with 60 participants compared remote guidance through two headsets and a smartphone in a simulated ship environment. Certain head and knee contacts were recorded more frequently with the Trimble XR10 with HoloLens 2 than with the RealWear Navigator 520 and smartphone. No significant difference in these contacts was reported between the Navigator and smartphone. The cited work is an arXiv version from 2025.^[1]

This is not general evidence of safety for a device class. Device, display and interaction varied together in the experiment. The finding nevertheless shows that a display usable at a quiet workstation can place different demands on the user when moving around or passing through low openings.

Three questions help with selection: Must information appear spatially at the object? Must the person move while using the display? And what interaction becomes necessary when they change position? The answers determine which display concept should be tested in your own workflow.

A maintenance paper by Sara, Todde and Caria examined a Vuzix M400 through technical tests of QR codes, audio and video transmission, and voice commands. A step-by-step guide for inspecting a milking machine was also developed. Four operators tested 39 voice commands under different noise conditions. At 75 dB, recognition rates differed substantially between individuals; at 80 and 85 dB, only a few or no commands were recognised. This concerns the setup studied in 2022 and is not a general noise limit for present-day M400 applications. It did not demonstrate faster or less error-prone maintenance compared with paper.^[7]

Your own interaction trial must therefore account for noise peaks and differences between the intended operators. In Sara's measurements, the agricultural machinery room studied averaged 96.7 dB, well above the noise levels at which voice commands were still predominantly recognised in the experiment.^[7] A voice function that works in quiet surroundings must therefore also prove itself under the task's actual noise conditions.

What wearing the device entails

Interaction also includes wearing the device. Display position, visibility of the work area and compatibility with safety glasses or a hard hat must fit the task. A short test in a meeting room does not show how the combination behaves under workplace conditions.

In an occupational medicine review from 2020/2021 focusing on goods logistics and order picking, Holz and colleagues summarise studies in which smart glasses sometimes caused greater visual fatigue than conventional displays. They also describe the limited and inconsistent research base at that time. This supports neither a general assurance of harmlessness nor a blanket judgement that today's devices are unsuitable.^[6]

An industrial review by Solomashenko and colleagues adds an ergonomic perspective. For the Vuzix M4000, it describes difficulties adjusting the display, reaching buttons while wearing a hard hat, interfering cables and uneven weight distribution. The same review lists compatibility with personal protective equipment, connection of an external battery and a low weight of less than 200 grams as advantages. These reported advantages and drawbacks rely on a mixture of open sources, company material and user experience. They provide selection criteria, not representative frequencies or controlled evidence of reduced work performance.^[8]

A trial should therefore also include intended wearing duration, changes in gaze and the amount of information actually needed. What matters is whether the person can read the instruction and then assess the work area. In context-aware assistance, presentation must therefore also fit gaze direction, posture and the working environment.

What must be clear when confirming

A confirmation can complete a step or influence the subsequent workflow. The person must recognise the effect of their input, which content they are confirming and whether the application accepted it. Missing or ambiguous feedback can make even a short interaction difficult to manage.

For speech, recognition, ambient noise and specialist terms may matter.^[2] For a button-based interaction, one test question is: Could the person press the button accidentally or for the wrong display? The trial should therefore observe whether the intended action is unambiguous and whether an incorrect entry can be corrected in a controlled way.

Assignment to the person acting also belongs in the application's specification. The input method alone does not establish that assignment. For a confirmation to remain traceable in the audit trail, its relationship to the person acting and the relevant operation must be established.

Start the first trial with a concrete task

The cited studies examine neither 420+ nor a general standard for industrial smart glasses applications. They show the differences that can become visible when a specific workflow is tested with a particular display and interaction method.

A first trial can therefore focus on a clearly bounded step: the same inspection instruction, the same task, a defined input and an observable response to unclear or incorrect entries. Alongside completion time, what matters is whether the person perceives the work area sufficiently and can check their result without unnecessary steps.

Which work step currently requires both hands, a view of the instruction and a traceable entry at the same time — and which of these three demands must the smart glasses demonstrably meet better in the trial?