The methodology proposed by Mark and his team incorporates guidelines covering both the selection and implementation of worker assistance systems. It is based on a high-level data catalog, which is a structured database providing a methodical listing of all assistive equipment available on the market.

The selection process itself consists of four steps:

This step allows to identify the constraints and limitations to which the assistive equipment will have to adapt. It allows a comparison of the characteristics of the different pieces of equipment in the catalog with the basic characteristics of the work environment. It thus becomes possible, early in the process, to eliminate equipment that is not suitable for the given environment, or to remedy a limitation unique to the environment.

This step identifies relevant parameters related to the task and to workers’ needs. To this end, Mark, Rauch, et Matt [6] propose a list of 23 parameters grouped into five categories, namely, relevant human senses (e.g., sight), physical ability, cognitive ability, personal qualities, and skills. The task is assessed based on these 23 parameters, which allows to establish a:

During this step, the worker’s needs are matched to the task requirements by performing a component-by-component product of vectors $\stackrel{\to }{\text{A}}$ and $\stackrel{\to }{\text{B}}$. The result is a vector $\stackrel{\to }{\text{C}}=\stackrel{\to }{\text{A}}*\stackrel{\to }{\text{B}}$, which identifies the parameters requiring assistance.

Each piece of equipment in the catalog (25 pieces) is rated according to the 23 parameters, on a scale ranging from 0 (no assistance) to 10 (maximum assistance). This generates a matrix M with 25 rows (equipment) and 23 columns (parameters). Matching is realized by performing the product $\stackrel{\to }{\text{S}}=\text{M}*\stackrel{\to }{\text{C}}$, where each row of $\stackrel{\to }{\text{S}}$ corresponds to a piece of equipment and indicates its suitability ranking. Equipment is ranked in descending order of relevance. Taking into consideration additional parameters, the top-ranking options are then examined in greater detail. Examples of these parameters include acquisition and implementation costs, technological maturity, organizational impact, etc.

The methodology proposed by Mark, Rauch, et Matt [6] stands out for its innovation and versatility, since it can be used in many different contexts and for different types of assistive equipment. It is not limited to cognitive assistive equipment, but also includes sensory and physical assistive equipment, and even allows to combine two or three of the categories to address a specific problem. The approach focuses on a detailed analysis of the task, the worker and the work environment. Furthermore, it takes into consideration the needs and constraints of all participants in a work system, while providing an exhaustive evaluation of available equipment thanks to its high-level data catalog.

It should nevertheless be noted that this methodology is primarily designed for a general manufacturing context. For complex assemblies realized in narrow workspaces, it cannot, by itself, precisely identify equipment that can provide optimum worker support.

Indeed, multiple studies have shown that the most suitable assistive equipment in such contexts are wearable devices. Although implementing Mark, Rauch, et Matt’s methodology [6] may lead to this conclusion, further studies are still needed to determine which wearable devices provide the best ergonomic and technological performance.

Syberfeldt et al. [11] propose a structured and simple selection process designed to help manufacturing companies efficiently identify the best augmented reality glasses from the many options available on the market. This process is based on a 12-stage sieve and takes as input a complete list of available glasses, after which it provides a single optimal option as output.

The initial list is built by listing all the augmented reality glasses available on the market. Each stage of the sieve corresponds to one of 12 comparison criteria: mass, visual field, battery life, optics, open programming interface, audio, control, processor, storage, processing memory and connectivity. Syberfeldt and his team also established an acceptability threshold for critical criteria such as battery life, mass and head-mounted display visual field.

The selection process at each stage follows a precise algorithm:

At each stage, glasses that do not meet the defined thresholds are eliminated, until, at the final stage, one is selected as the best among the initial 12.

However, this approach has several limitations. The first is related to the layout of the stages, which can lead to:

The second limitation of this approach is linked to a strict use of acceptability thresholds. By proceeding as such, the algorithm results in the direct elimination of smart glasses without taking into consideration the influence that some criteria may have on others.

Though complementary, these two studies highlight a crucial need: develop an assistive equipment selection approach that combines technical rigour with a consideration of human and contextual factors, while remaining pragmatic and operational for companies.

The present article adopts this approach by proposing a systematic selection process for visual and cognitive assistive equipment, adapted to the diverse needs of operators and the constraints of modern industrial environments. Based on a critical review of existing methodologies and an in-depth analysis of emerging needs, this approach aims to provide decision-makers with a practical framework for selecting the most suitable assistive equipment, while enhancing operational performance, safety, health and worker comfort.

Ergonomic criteria, including sub-criteria such as mass, interaction and the visual field, have a direct impact on worker comfort and performance. For example, an excessive weight or a poor load distribution of headsets can lead to increased muscle fatigue and neck pain, as has been demonstrated in work dealing with head-mounted devices [17] [18] . In addition, a fluid interaction and a wide visual field are essential for reducing visual distractions and improving efficiency in industrial environments [18] [19] . As a result, we consider ergonomic criteria to be far more important (8) than connectivity.

On the other hand, although connectivity does not contribute directly to physical comfort, it plays a crucial role in the integration of devices into connected industrial systems. It conditions the ability of HMD to provide real-time information, which is essential in modern production environments [19] . This justifies not neglecting this criterion, although its impact remains limited as compared to ergonomic criteria. We therefore assign a relative weight of 1/8 to connectivity with respect to ergonomics.

Technical criteria are mainly limited to the material performance of alternatives, such as durability, sensors, and computing capacity [19] . Although these criteria impact device reliability and longevity, their influence on users is indirect. Conversely, ergonomic criteria directly affect health and operator satisfaction, making them markedly more important. We therefore consider ergonomic criteria to be more important (5) than technical criteria.

While having only a limited impact on the performance of alternatives, the connectivity criterion continues to be essential for their integration into a connected system and for their ease of use. However, because of their direct impact on the functionality and performance of alternatives, technical criteria are considered more important (3) than connectivity.

Table 1 below presents a comparison matrix for the first-level criteria in a bid to select the most suitable head-mounted display for the work situation under study.

Calculations carried out in Excel using the XLSTAT software gave the relative weights of each group of criteria: we obtained 73.70%, 7.68% and 18.63%, respectively, for the ergonomic, connectivity and technical criteria groups.

A poor visual field has a significant impact both on working posture [17] and on the frequency and intensity of head and neck movements [18] . These limitations increase muscle fatigue and can reduce operator efficiency, particularly in demanding industrial environments. Furthermore, a low display resolution will directly impact the functional efficiency of HMD, in addition to degrading the user experience [18] .

Nichols [18] suggests that slightly heavier HMD are often preferable if they offer superior features in terms of the visual field and the display resolution. This underscores the importance of these two sub-criteria in the design and effective use of augmented reality devices.

A visual field of around 20˚ may be acceptable for applications such as sports or tourism. However, for industrial use, such a visual field is clearly not enough, making HMD practically unusable [11] . A wider visual field improves spatial perception and reduces the need for repetitive head movements to adjust the perspective. As a result, we ranked this sub-criterion at a much higher tier (7) than mass.

The viewing mode also significantly impacts the efficiency of HMD. An inappropriate visualization mode, such as video-based solutions with high latencies, can disrupt the synchronization between visual information and operator actions, which can have a negative impact on task accuracy. Conversely, optical or retinal projection-based visualization systems are better suited to industrial environments and provide immediate visual feedback and better integration of virtual and real information [11] . Owing to the foregoing, this sub-criterion is deemed more important (5) than mass.

Although mass is a determining factor for long-term comfort, its impact is relatively less critical when offset by a good ergonomic design and superior performance in the other sub-criteria. We therefore ranked mass at a lower tier than visual field and viewing mode.

Table 2 below shows the comparison matrix for the four ergonomic sub-criteria with respect to the first-level hierarchical criterion, ergonomic criteria. Developed using the Saaty scale [16] , this matrix ranks the sub-criteria according to their respective impacts on user comfort and efficiency in an industrial context.

Using XLSTAT in Excel, we calculated the relative weight of each ergonomic sub-criterion based on the data presented in the table. We thus got: 41.66% for the visual field, 6.78% for mass, 21.77% for viewing mode and 3.47% for interaction.

Technical sub-criteria such as processor, RAM, battery life, camera and storage play a crucial role in the operation and performance of augmented reality HMD. However, their relative importance varies according to their direct impact on user experience and industrial task requirements.

The processor and RAM are the driving forces behind HMD, defining their ability to handle complex processes in real time. Together, they impact essential functions such as:

Inadequacies in these two areas affect not only HMD performance, but also users’ physical comfort. According to Nichols [18] , slowness in the display or in controls can impair postural stability and cause fatigue [20] . Therefore, the processor is considered almost as important (2) as RAM, with both working in synergy to ensure optimal performance.

Autonomy is the third most important criterion. A long-lasting battery ensures continuous use of the HMD without frequent interruptions to recharge or replace the battery. This is particularly critical in industrial environments, where unplanned downtime can impact productivity. Autonomy is therefore deemed less important than processor and RAM (5) but remains a key priority.

The camera plays an important role in specific applications such as:

However, its impact is limited to scenarios where these functions are necessary. As a result, the camera is considered slightly less important (1/3) than autonomy.

Storage is the lowest priority sub-criterion in this hierarchy. Thanks to network connectivity, HMD can access corporate databases in real time, limiting the need for local storage. The priority given to storage is therefore much lower than that of the other sub-criteria (8 with respect to the processor).

Table 3 below compares the five technical sub-criteria (processor, RAM, autonomy, camera and storage) against the first-level hierarchical criterion, i.e., technical criteria.

XLSTAT was used in Excel to calculate the relative weight of each of these sub-criteria. We thus got: 2.69% for autonomy, 8.49% for the processor, 5.38% for memory, 1.45% for the camera, and 0.62% for storage.

Objective: Identify the technologies or types of equipment best suited to the context, taking into consideration the factors influencing their integration.

These factors, which were identified by Haase, Radde, Keller, Berndt, and Dick [21] , are summarized in a table of parameters designed to inform the assistive equipment choice. These factors notably include:

According to Petzoldt et al. [2] , incorporating workers’ needs into equipment design improves its acceptability and effectiveness. The taxonomy of assistance systems described by Pokorni and Constantinescu [22] is used for a preliminary sorting of the technical characteristics required for the equipment to be able to optimally provide the assistance sought.

Deliverable: A shortlist of equipment potentially suited to operational requirements.

After a specific type of equipment is identified, a multicriterion analysis is carried out to compare different alternatives. Given that we are handling both qualitative and quantitative data, the AHP method was chosen for decision support. By simultaneously considering numerous criteria, the AHP will allow to easily obtain the weight of each of the alternatives and thus determine which is to be preferred [14] .

The criteria used for the AHP are drawn from the scientific literature, including the works of Fraga-Lamas et al. [15] and of Syberfeldt, Danielsson and Gustavsson [11] . This method provides a rigorous framework to inform the decision-making process.

Deliverable: A priority alternative identified based on the relative weights calculated.

Objective: Check how practically compatible the selected equipment is with the company’s context.

This step assesses the practical acceptability of the alternatives selected in the previous step, taking into consideration the following:

Deliverable: A report detailing the practical strengths and weaknesses of the selected equipment.

Objective: Test equipment under real-life conditions to validate its usefulness and effectiveness.

In this final step, human participants are brought in to test the equipment under real-life conditions. The assessment covers:

This assessment is carried out in accordance with the recommendations of ISO 9241-11 [23] , which provides a framework for assessing the usability of equipment in terms of comfort, efficiency and satisfaction.

Deliverable: A final assessment to confirm or adjust the choice of equipment.

The industrial partner we worked with operates several workstations dedicated to gas turbine disassembly, inspection, maintenance and assembly operations. These operations require specially adapted tools and methods to ensure worker efficiency, health and safety.

Figure 4 illustrates the work system currently used for the assembly and disassembly of gas turbines.

Tasks are carried out by a team of two technicians working closely and who communicate regularly to ensure effective synchronization. One of the technicians plays the main role in carrying out the tasks, while the other provides the necessary support by performing various complementary actions, such as operating workstation elements using joysticks, providing work tools, and ensuring backup lighting using a torch.

Interactions between technicians, as well as with tools and equipment, are strongly influenced by the work environment. Notable factors include lighting intensity, noise, temperature and chemicals occurring in mechanical parts.

In the present study, we are particularly interested in the assembly of one of the gas turbine modules (see Figure 5 ).

The workstation consists of:

Moreover, the configuration of the workstation makes it difficult for the technician already installed in the turbine module to access work instructions and data located on a remote computer. Nevertheless, the technician must regularly:

This situation highlights the need for assistive equipment that meets these requirements, without requiring movement by the technician.

This section is dedicated to the first-level validation of the approach proposed in Section 5. The approach is applied to a real case study, centered on the assembly station of one of the gas turbine modules of an industrial partner, as described in the previous sections.

The first step in the process is to analyze the task, environment and workstation, based on the factors described by Haase et al. (2020). This analysis is crucial for the identification of the key points that will inform the preliminary choice of assistive technology or equipment. In the context of this case study, several significant observations were noted:

1) Mobility and flexibility of the assistance system:

As a result, the assistance system cannot be fixed; rather, it must be mobile or portable to adapt to workers’ frequent movements.

2) Hand occupation constraint:

These constraints underscore the importance of an assistance system that does not hinder operators’ movements, while providing them with real-time information.

3) Geometry and workspace:

Support equipment must therefore be compact, ergonomic and minimize excessive movement.

4) Difficulty accessing digital tools:

Because of to the workstation configuration, technicians cannot easily access the computer located outside the module. This computer is essential for:

These constraints underscore the need for a solution that provides immediate, hands-free access to data and instructions, while eliminating the interruptions associated with movement.

Figure 6 below shows an extract from the taxonomy of assistance systems proposed by Pokorni et Constantinescu [22] . This classification provides an overview of the different categories of available assistive technologies and their characteristics.

Taking into consideration the specific constraints and requirements identified in the previous analysis, we conducted a preliminary selection by excluding solutions deemed inapplicable or unnecessary for our case study. These exclusions, represented by red boxes, allow to narrow down the range of choices and focus only on technologies suited to the needs of the gas turbine assembly station.

A preliminary sorting was carried out, based on the taxonomy of assistance systems, to identify the types of assistive equipment likely to meet the industrial partner’s gas turbine assembly station’s specific requirements. This sorting, presented in Table 4 below, allows to narrow down the field of alternatives by eliminating unsuitable options.

This analysis isolates the head-mounted display as the most appropriate type of equipment for this case study.

Integrating a head-mounted display into the assembly workstation offers multiple advantages for improving efficiency, communication, and the precision of the tasks performed by technicians.

5) Reduce unnecessary movements

Thanks to its features, the head-mounted display eliminates the need for the technician to leave his workstation to consult work-flow updates on a fixed computer. This optimizes working time and reduces unnecessary interruptions.

6) Real-time remote assistance

Equipped with an integrated camera and audio system, the head-mounted display enables the technician to benefit from real-time remote assistance. This support, described by Aranda-García et al. [24] and Klinker et al. [25] , is particularly useful for quickly resolving problems encountered during task execution.

7) Simplified documentation

The integrated camera facilitates documentation of the work carried out and simplifies report writing [25] . This functionality is essential for detailed monitoring of operations and to ensure the traceability of actions.

8) Improved communication

Head-mounted displays represent a practical solution for communicating with other team members or line managers. This feature promotes smooth coordination and real-time monitoring of task progress, enhancing project collaboration.

9) Enhanced visibility in hard-to-reach areas

Connected via Bluetooth or Wi-Fi to a strategically positioned external camera, the head-mounted display overcomes the challenges of visually inaccessible areas, particularly during bolting operations. This connection provides effortless vision while reducing physical strain on the technician.

There are a wide variety of head-mounted displays available on the market, each offering different features and functions to meet various cognitive assistance needs. One cannot simply conclude that a specific head-mounted display is the right equipment, as that would be an incomplete and imprecise statement. It is essential to go a step further and identify which head-mounted display is best suited to the specific needs and constraints of the task and workstation under analysis.

A pre-selection of HMD was carried out to limit the number of comparisons in the AHP and reduce the risk of errors (see Figure 7 ). This approach made it possible to focus on HMD most widely used in the manufacturing sector, with proven technologies and positive feedback.

Newer head-mounted displays, whose reliability and effectiveness had not yet been validated, were excluded to ensure a selection based on equipment adapted to the industrial partner’s operational needs.

Finally, a process of elimination by dominance was used to select the best-performing models when the same manufacturer offered several alternatives. This approach ensures the availability of a shortlist of reliable options suited to the assembly station.

The problem of selecting the best head-mounted display alternative is structured as illustrated in Figure 8 below.

1) Establishment of comparison matrices of alternatives against criteria

With the relative weights assigned to each group of criteria and sub-criteria (see Section 4), this step consists in making binary comparisons of the alternatives on each sub-criterion, using the Saaty [16] scale.

The numerical performance values of the head-mounted displays evaluated for each criterion were used as the basis for these comparisons. However, for non-quantifiable criteria such as viewing mode and interactions, a qualitative approach was adopted.

For the viewing mode, based on the literature review, a higher priority was given to binocular or visor-type glasses with a transparent optics mode. This was followed by monocular glasses with transparent optics, while those using a transparent video mode were considered a lower priority. In terms of interaction mode, HMD offering greater diversity and flexibility of interaction were deemed superior.

The matrices obtained from the binary comparisons are shown in Appendix A.

Considering possible bias in judgments made during pairwise comparisons, it is essential to check the consistency of the matrices obtained. This step helps avoid assigning incorrect weights to the criteria, which could distort the study results.

Consistency is verified by calculating the consistency ratio (CR) and comparing it with the threshold value of 10%. Two possible scenarios may arise:

As all the calculated consistency ratios are below 10% (see Appendix B), the comparison matrices can be considered consistent and valid for analysis.

2) AHP results

Compiling all the calculations reveals a significant domination of the Magic Leap 2 (ML2) and Microsoft HoloLens 2 (MH2) HMD over all the other alternatives (see Figure 9 ).

In literature, various methods are used to assess the acceptability of technological equipment. These methods share common features, notably the assessment criteria and the involvement of human participants in the process. However, in the context of our study, the direct involvement of users constitutes a limitation. To overcome this constraint, the assessment of practical acceptability and usability of the ML2 and MH2 is based on two main approaches: the use of basic knowledge in ergonomics and the use of the results of similar studies carried out by other research teams.

The first assessment criteria pertains to user satisfaction. Studies carried out by Basoglu et al. [26] and Wang et al. [27] on HMD that are technologically inferior to the ML2 and MH2 (in terms of visual field, interaction mode, processing power, etc.) have shown that these devices, although less powerful, largely satisfy users’ needs.

By extrapolation, the advanced features of the ML2 and MH2 offer sufficient guarantees to meet user expectations in the present case study.

For the other criteria, a separate assessment will be made for each of the two HMD.

1) Magic Leap 2

Comfort and Design:

The ML2 distinguishes itself through a compact, lightweight, and discreet design, resembling standard eyeglasses. In contrast to the more cumbersome visor-like MH2, the ML2 provides enhanced ergonomic comfort, addressing typical pressure points identified by Kim et al. [28] , particularly around the nasal bridge and ears.

To optimize wearability, the ML2 includes an ergonomic adjustment kit: four interchangeable nose pads, two forehead pads, a flexible rear band for automated fit, and an optional top strap for added stability. This configuration ensures a balanced weight distribution, reducing fatigue even during extended use. Unlike other devices that suffer from asymmetrical weight distribution [29] - [31] , the ML2 maintains comfort and equilibrium.

Additionally, ML2 features dynamic dimming technology, as described by Hoffman, Stepien et Xiong [32] , which auto-adjusts display brightness based on ambient lighting and visual content. This improves visual clarity, reduces eye fatigue, and enhances usability in bright environments. However, a design limitation is the presence of bezel arms that may obstruct peripheral vision.

Ease of use and memorability:

The ML2 supports intuitive interaction through voice commands, gesture recognition, and a dedicated joystick. Voice input does not require strict syntax, and gesture-based control supports both near-field and far-field interaction modes. Real-time hand tracking and natural gestures enhance usability and memorability, even after prolonged periods of non-use.

Software compatibility:

Built on an open Android-based platform, ML2 supports custom application development via Android Studio and Unity. It is compatible with industry standards such as OpenGL, OpenXR, WebXR, and Vulkan, and integrates easily with enterprise systems through the Magic Leap Hub. Additionally, the Magic Leap App Store provides access to a variety of sector-specific applications.

2) Microsoft Hololens 2

Comfort and Design:

MH2 features a fold-down visor design with manual fit adjustment via a rear tightening ring. While it avoids direct pressure on sensitive areas like the nose and ears, its higher weight (556 g) may cause discomfort during prolonged use [17] - [19] . Its transparent visor allows easy transition out of AR mode and preserves peripheral vision, although outdoor use is limited due to poor screen visibility under bright light. Visual artefacts like reflections and rainbow effects can lead to eye fatigue.

Ease of use and memorability:

Like ML2, MH2 offers intuitive interaction via voice and gestures. It includes practical combinations such as eye-tracking with voice input. Environmental scanning is automatic and continuous, facilitating 3D mapping. MH2 supports multiple application windows simultaneously, whereas ML2 restricts to a single window, although the latter performs better with high-demand applications.

Software compatibility:

MH2 runs Windows Holographic, ensuring compatibility with most Windows 10 apps and tools such as Dynamics 365, Teams, and Remote Assist. While it enables custom development through Unity and Visual Studio, it is more closed than the ML2 in terms of integration. Its major strength lies in its autonomy, MH2 functions as a standalone unit, while ML2 relies on a cabled external processor.

The tables below show the pairwise comparison of the five selected HMDs according to the different sub-criteria.

Table A1. Comparison matrix of alternatives with respect to visual field sub-criterion.

Table A2. Comparison matrix of alternatives with respect to the mass sub-criterion.

Table A3. Comparison matrix of alternatives with respect to the viewing mode sub-criterion.

Table A4. Comparison matrix of alternatives with respect to the interactions sub-criterion.

Table A5. Comparison matrix of alternatives with respect to the connectivity criterion.

Table A6. Comparison matrix of alternatives with respect to the autonomy sub-criterion.

Table A7. Comparison matrix of alternatives with respect to the processor sub-criterion.

Table A8. Comparison matrix of alternatives with respect to the memory sub-criterion.

Table A9. Comparison matrix of alternatives with respect to the camera sub-criterion.

Table A10. Comparison matrix of alternatives with respect to the storage sub-criterion.

The consistency ratio (CR) was calculated for each pairwise comparison matrix to ensure the reliability of judgments made during the Analytic Hierarchy Process (AHP). According to Saaty [16] , a CR lower than 10% indicates an acceptable level of consistency in the matrix. The following table summarizes the consistency ratios obtained for each matrix used in this study.

Table B1. Comparison matrix consistency ratio.

All consistency ratios fall below the 10% threshold, confirming the validity and coherence of the pairwise judgments used in the AHP process.