Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment

: Remote diagnosis procedures are prone to communication errors due to varying levels of experience and knowledge between expert maintainers and technicians. These result in inefficiencies that delay the diagnosis process. The aim of the paper is to develop a Structured-Message Authoring framework for Augmented Reality (AR) Remote Communication (SMAARRC) and to evaluate its ability to enhance the efficiency of remote diagnosis services. The framework proposes a message structure and automatic AR content creation rules for it that enable data capture and sharing within a remote context. Laboratory experiments present an average time reduction of 56% for remote calls while maintaining same quality compared to traditional remote communication methods (phone calls and emails). Remote experts feedback evidence the usability and feasibility of this framework to work in real-life conditions.


Introduction
Engineering collaboration is a socially-mediated technical activity that involves multiple people working interdependently to achieve a greater goal than is possible for any individual to reach alone [1]. The progress in information-communication technology has made it possible for the collaboration to take place remotely over large distances, allowing globally dispersed businesses to operate around the clock. Many engineering processes, such as collaborative design [2] and remote maintenance [3], have been improving using remote interfaces. Remote engineering collaboration is increasingly becoming necessary in remote maintenance for various reasons such as reducing cost of travel and increasing efficient use of experts' time. A recent report on UK service and support industry valued the global market in 'service and support' across high value manufacturing sectors at £490 billion in 2017 [4]. The remote maintenance market is poised to grow at around 25.9% over the next decade to reach approximately $69.81 billion by 2025 [5].
Remote diagnosis is one of the biggest challenges in remote maintenance of complex equipment [5] (e.g. machine tools or aerospace machinery). Remote diagnosis refers to remote support involving procedures for finding failures, and validating and fixing components' faults that contribute to them. It is usually conducted with the guidance of expert remote maintainers and implemented by on-site technicians. The technicians can have varying levels of experience and knowledge, and require guidance from the remote expert. In these situations, enabling efficient communication is essential. Communication challenges such as task misunderstanding or component misplacement can cause delays and errors, and ad-hoc communication (e.g. phone calls) may lead to more confusion. Hence, communication should be structured in order to be efficient.
A very effective interface in remote maintenance contexts is Augmented Reality (AR).
Compared to conventional approaches (e.g. phone calls, emails), AR can enhance communication efficiency by its ability to overlay virtual information into real-life scenarios [6]. For example, it Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment 4 can enable the expert to point in technician's sight the exact location to which an instruction is referred to. One of the main advantages of using an AR application is that complex information can be represented in a more comprehensible format, making it easier for the operator to understand [7]. AR studies in maintenance show promising results for enhancing human performance when carrying out maintenance tasks [8] and show that AR can significantly improve maintenance efficiency in comparison to existing information-delivery methods (e.g. paper manuals, phone calls, etc.) [9]. However, several areas (e.g. authoring, context awareness, and interaction analysis) still remain challenging for efficient use of AR to support maintenance tasks [10]. A general concern within those is to identify the information that can be displayed for effective AR support of a maintenance task [5]. In AR for remote maintenance particularly, there is lack of research to characterise the language and the structure of communication for efficient information delivery [7]. Since no universal or standard language exists, a common approach is to define a set of agreed-upon symbols and then leave some space for free input.
This paper goes a step further and proposes a framework with a common message structure for AR-based communication in the remote diagnosis context. It tackles the challenges related to ambiguity and inaccuracy of communicating tasks and associated components by developing an AR-based solution for accurate and timely remote guidance. This research makes two major contributions to the AR and remote engineering collaboration literature in the context of improving diagnosis efficiency: • It develops a structure that determines elements of message codification to minimise the above-mentioned remote communication challenges.
• It proposes a new rule-based approach to automatically generate AR content from structured messages for efficient communication in remote diagnosis processes.
Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment

Remote collaboration: challenges and opportunities
Remote collaboration can play an important role in maintenance [11]. As an example, it is emphasised that equipment repairs and maintenance cost between 30% and 50% of the total operating costs of mine sites. Moreover, studies [12] also predicted that "every 1% improvement in equipment availability or productivity improves the overall productivity by up to 3.5%".
Remote collaboration can support such improvements by offering efficiency and effectiveness in delivering maintenance tasks [13].
Distance promotes the need for remote collaboration. The challenge with remote collaboration is not only related to distance that is driven by physical attributes, but also operational and affinitive. Affinitive distance involves the resultant gap in operational styles between team members, and its effect on synergy, camaraderie, and management. On the other hand, operational distance refers to that of communication between teams of various sizes, as well as the gap between the skill levels of different team members [15]. All of these factors result in affecting the efficiency of remote collaboration.
Remote diagnosis scenarios mostly involve "collaborative physical tasks". These are tasks in which various individuals work alongside to conduct activities in real-world objects [16]. Several observational studies [17][18][19] suggest that communication in collaborative physical tasks is directly related to target object identification, activities description, and successful performance confirmation. These elements can be described as the "situational awareness" [16]. Besides, it has Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment 6 also been pointed out [20] that situational awareness in remote collaboration is managed through the use of visual information. Hence, it can be said that the higher the visualisation of information, the higher the situational awareness; and so, the higher efficiency of remote collaboration.
A technology that can enable information visualisation for enhancing efficiency in remote collaboration is AR. The introduction of AR in remote collaboration processes, particularly for knowledge intensive works, is important for numerous reasons [21]: 1. reducing the costs of maintenance tasks 2. reducing the risk of accidents that may occur 3. improving time taken to complete a task

reducing experts' expenses for traveling to remote sites
The ability to overlay spatially meaningful information on the 3D space allows the AR technology to be a promising option to support knowledge-intensive work [22].

Uses of AR in remote collaboration
There have been numerous attempts to use AR for remote collaboration. Some have focused on improving the communication flow but involving very limited data. For example, 3D models can be created from only a picture [23]. In this case, a single picture taken by the technician allows a server to create a 3D shape of the object. The expert will be able to deploy holograms on this 3D model that overlays the physical object, in order to provide better instructions to the technician.
The e-maintenance for a photovoltaic power system [24] is another example of utilisation of AR to reduce time taken to complete maintenance tasks. During an assistance, the remote expert has the possibility to augment the vision of the worker by inserting 3D content in the video streaming. The objective here is to provide fast and easy access to relevant information to the worker in order to minimise the costs of maintenance tasks, reduce the risks of accidents that may occur in remote regions and to improve the communication between technicians and Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment 7 experts. Although, the content can take several forms (e.g. notes or 3D arrows), it is still completely up to the expert to decide the best format for sending messages.
Another example of this is shown by Rambach et al. [25], where 2D annotations were converted into 3D annotations to enhance the ability to easily locate the component those refer to. Other researches went further by providing different approaches according to the context. For example, Hadar et al. [26] proposed a solution with expert-and machine-assistance according to the technician's needs. They show different methods to assist technicians: 1. By pre-recorded information that will help in the tasks, through a digital step by step manual.

2.
By an autonomous cognitive system that can understand the situational context.

By instructions sent through a remote expert.
Although this proposal aimed at reducing verbal confusion with real-time instructions, there is no evidence of enabling confirmation of task completion or structuring of messages sent.

Approaches of AR to enable collaboration
Some applications of AR to Remote Assistance are not just improving communication flow between experts and technicians but enable a real collaboration between the two. This can involve the remote expert being able to interact with the environment of the technician.
Moreover, instead of basic instructions (e.g. drawing, deploying holograms), the expert can collaborate in real time with the technician by having hand gestures projected to the technician [27].
Another type of collaboration involved 3D referencing for remote task assistance [28] through AR. The aim of this was to improve accuracy and efficiency with which the remote expert can point a real physical object at a local site via a technician on another site. This application can track the hands of the expert to display them right in front of the technician. Thanks to this, the expert can achieve two things:

Research gaps
AR research in remote collaboration has focused in the following areas [29]: • Spatial problem solving: to enhance the ability to recognise objects and their virtual counterparts.
• Cognitive interaction: to adapt content to the cognitive workload of the technician.
• Interactive design: to investigate performance, behavioural and cognitive effects of virtual contents.
These areas tend to put the technology at the centre of delivering effective remote collaboration. Nevertheless, there is lack of research evidence focusing at methods that ensure appropriate levels of situational awareness regarding the messages being sent between experts and technicians in remote communications and the protocols to do so. A possible method could consider the creation of message structures for communication exchange in remote collaboration to enhance situational awareness of technicians through a better mutual understanding with experts. That could also have further impacts in remote diagnosis research as it can allow to record, and reuse data exchanged in previous communications for future remote diagnosis operations or other maintenance areas (e.g. failure prediction). Hence, this research aim is to propose an AR solution that includes structured messages and communication protocols to enhance the diagnosis efficiency by improving the situational awareness of the remote collaboration.
Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment 9

Methodology
The research aim identified in the previous section drove the selection of an appropriate research method to successfully design and validate an AR research solution. Inspired by similar works in the field [29][30][31] and well-established methodologies for design research [32], the method utilised by this research is as follows: 1. Identification of objectives: conduct a literature review [33] to find specific research gaps and justify the value of the research solution proposed. The results were presented in Section 2.

2.
Solution design: utilise the 5 W's method [16,32,34] to define the structure of remote messages and the communication protocol. Then, map [35] the resultant message elements against relevant AR content types to enhance situational awareness of remote communication. The proposed solution is presented in Section 4.

3.
Demonstration: conduct interviews [36] with experts in remote diagnosis to identify relevant scenarios and produce a case study for the research solution's validation. Then, implement the proposed solution in a prototype AR system for further experimentation. The case study and the system implementation are presented in Section 5.

4.
Validation: design an experiment [36] according to relevant criteria in remote collaboration [16,20] to validate the research proposed. Then, conduct the experiments, and analyse and assess the results. The experiment design is presented in Section 5. The experimental results and comparison with relevant literature are presented in Section 6.

Communication: disseminate the research's proposal, methods, and relevant results and
conclusions. This paper is the authors' proposal for the research's dissemination. Its conclusions are presented in Section 7.

Structured Message Authoring for AR-based Remote Communication (SMAARRC)
This research proposes an AR-based remote communication framework based on: (1) an innovative message structure (4.1), and (2) a rule-based authoring approach for automatic AR content creation (e.g. holograms, images, etc.) (4.2). Figure 1 presents this approach compared to

A message structure to comply with situational awareness needs of remote collaboration
A remote communication between an expert and a technician can be described as a 'call', which involves an exchange of 'messages'. A message, verbal or written, is made up of a set of 'message elements' strictly adhering to a 'message structure'. This offers a consistent language and structure for remote communication in the diagnosis context.
The Five W's method [34,36] was used to provide a mutually exclusive, collectively exhaustive set of message elements that complies with remote diagnosis challenges: • Ability to declare messages that can describe any procedure.
• Ability to record and replay a call based on the message logs.
• Ability to use 'message elements' for creating AR content. • Who? Messages can be sent by one or more experts and technicians that act as 'senders' or 'recipients'.
• What? Messages in remote diagnosis should define the 'type' of processes being described.
In a diagnosis procedure, this 'type' can be: action, confirmation, question, or response.
• Where? Messages should also indicate the 'place' where the process is occurring. This can be described in the form of a 'component' and/or a 'location' related to it.
• When? Messages have a specific order within a call. A message 'identifier' allows to reconstruct the order of the messages within a call.
• Why? Messages should clarify the context or 'category' of the process being described within a diagnosis call. In line with the definition above, a 'category' can be: failure definition or component's fault validation.
• How? Messages should describe the maintenance methods used to conduct the procedure being described. A method can be described in the form of an 'action' and a 'measure'.
• In case the previous elements cannot describe accurately a message, an 'object' could be added for further clarification.
These message elements and the values they can take to generate messages are presented in Table 1. The proposed message structure focuses on diagnosis and considers the full spectrum of the remote communication needs. It helps the diagnosis problem by ensuring the scope of diagnosis and associated sequence of steps is efficiently followed.
Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment time. This helps to ensure that the AR content for each message is not overlapped with content from other messages in the augmented scene. Figure 2 shows an example message and the AR contents for its message elements in both, technician's and expert's view. The former comprises the augmented scene. The latter includes a virtual environment to interact with the equipment's model and code messages, and live-streaming of technician's view. This message structure can also help to analyse remote diagnosis tasks. It allows to store and process messages in a structured way. The following section describes the rules and AR content types to augment message elements.

Rules-based authoring for message elements automatic AR content creation to enhance diagnosis efficiency
Augmentation refers to the methods deployed to enhance natural environments or situations with virtual information in order to offer perceptually enriched experiences [10]. awareness. Object awareness refers to the ability of messages to identify the real-world objects being referred [20]. Instead, procedure awareness refers to the ability of messages to clearly define the task being referred [16]. Depending on whether a 'message element' can affect any of these two aspects, an augmentation method is assigned and an authoring rule is created. These rules, presented in Table 2   Holograms are three-dimensional images used to overlay the real scene with digital artefacts such as arrows, circles or component's 3D models in order to mark a particular feature of the scene [8]. The expert can deploy holograms remotely to provide guidance to the local technician with the aim to increase object awareness. This is achieved by giving the expert the choice to allocate holograms with their preferred shape, position, scale, and rotation. These are then communicated through the message elements including 'measure', 'location', and 'component'.

3D
Measurements are used to make the messages more accurate. They provide more precise values for the message elements 'measure' and 'action' in real-time by using Bluetooth devices or the so called bare-hand interaction [6] gestures for metric measurements.
Textual values are used as an augmentation method for multiple purposes. For example, it overlays the predefined questions to derive diagnosis and it helps to ensure the recipient uses the same vocabulary as the sender. This increases procedure awareness for message elements including 'location', 'component', 'action', and 'measurement'.
Pictures can provide additional details to increase procedure and object awareness for 'location' and 'object' message elements. For example, when technicians send real-time pictures to experts for evaluating the condition of a component's surface.

System implementation
The proposed solution was implemented within a prototype AR system for experimentation.
This system comprises three components: (1) a desktop computer with a virtual interface (including technician's live stream) for the remote expert, (2) a cloud server to store the transferred data, and (3) an HMD (Head-Mounted Device) for the technician with an AR application using HoloLens. The system architecture, along with the software and hardware used to build it, is presented in Figure 4.
Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment

Experiment design
The validation experiments aim at evaluating the ability of the proposed solution to enhance efficiency of remote diagnosis operations by improving the situational awareness of remote messages. Their objectives are to collect data for demonstrating the following hypothesis: • There is no significant difference in errors results by solution (AR vs NOAR).
• There is a significant difference in time results by solution (AR vs NOAR).
• There is a significant correlation between solution and object awareness effects on time results.
• There is a significant correlation between solution and procedure awareness effects on time results.
• Real-life experts and testers consider the solution proposed useful to enhance efficiency of remote collaboration in diagnosis scenarios.
• Real-life experts consider the solution proposed feasible to be implemented in real-life conditions on remote diagnosis scenarios.
According to similar research [16,29,30], the following measures can be considered as appropriate for such evaluation: Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment • Quantitative: time and errors of remote diagnosis operations.
• Qualitative: usability and feasibility of AR methods that support remote diagnosis operations.
For these measures to be appropriate for evaluating efficiency enhancements, the following assumptions must hold true: • Time and errors can be a direct representation of efficiency if consistent quality of remote diagnosis operations is assumed. In order to ensure that, the study assumes a pre-determined operation, the quality of which does not depend on the technician performance.  1. Declare a remote diagnosis operation that includes common operations in the maintenance context.

2.
Declare remote messages that cover all operational steps and include all levels of complexity of situational awareness.

Conduct experiments with those messages to study the effect of alternative solutions (AR and
non-AR) on sending those messages by measuring time and errors.
If the assumptions above are correct, then it is reasonable to expect the following results: • Errors do not vary with the use of AR or non-AR solutions.
• Completion times are reduced with the use of AR solutions compared to non-AR solutions.
• Differences between completion times for AR and non-AR solutions increase when the level of complexity of situational awareness increases.
The study described above considers one variable to test assumptions (errors), a response variable (time), and three independent factor variables (AR usage, object awareness and procedure awareness). While the two variables have been defined above, the factor variables are defined in Table 3.

AR usage Utilisation of an AR solution to send and receive messages and confirm completion of associated tasks Object Awareness
Ability of a message to identify the real-world object to which tasks refer Procedure Awareness Ability of a message to identify the steps to successfully conduct the activity referred Each factor variable has different levels. Their definitions are presented in Table 4. The experiment aims at testing all levels of complexity of situational awareness. In order to do so, it is necessary to declare a remote collaborative diagnosis operation that includes remote messages with all those levels. These messages and their situational awareness complexity levels are presented in 5.3. Each experiment of the study will consist of a tester conducting the diagnosis tasks related to each message received. Time will be measured for each message, while errors will be measured for each tester. Errors measurements are taken to validate the assumption of maintained quality among solutions.

5.2.2.Usability and feasibility questionnaires
The usability and feasibility questionnaires aim at evaluating the perceived validity of the AR methods utilised to deliver information for remote diagnosis support. Usability and feasibility are criteria that refer, respectively, to the ability of the AR solution to deliver information in an appropriate manner from the expert to the technician from user and working environment perspectives. These criteria have been divided in different sub-criteria to cover independently each aspect of the AR methods used as well as the operation quality discussed above. The criterions utilised are based on Nielsen's usability criteria [37] already utilised in similar research [30,[38][39][40]. The criteria, the evaluation elements, and the evaluation methods are presented in Table 5.
Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment Table 5. Description of criterions, evaluation elements and methods for usability and feasibility questionnaires.

Criteria Sub-criteria Element Evaluation
Usability • Errors are not evaluated in quality terms as they may be dependent on user expertise, which can vary.
• It is assumed that the quality is of consistent level for the stop watch time study if the results of the questionnaire provide a similar result to the experiments.
The methods to collect data according to the tests and surveys above are explained in the following subsection.

Experimental and analysis protocol
The resources available for this research did not allow to conduct on-site experiments with real-life experts and technicians. However, it was possible to conduct sample demonstrations to them and record real-life experts' feedback. It was also possible to conduct experiments with university students (non-experts) as technicians and record their feedback. That is why the protocol declared below includes tester (students) experiments and questionnaires for testers This protocol considers some assumptions: • Testers are not questioned on solution's feasibility. This is because their inability to provide reasonably valid feedback as they are not subject matter experts.
• Feasibility is measured in a binary scale as there is no option for results comparison. So, the research question is whether the AR methods are feasible or not.
• Usability is measure on a Likert scale in order to compare results different interviewees according to their expertise in the subject. If experts and testers provide similar responses, then it can be said that content is usable in remote diagnosis environmental conditions.
The experimental protocol also includes the analysis of data collected. That is as follows: Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment 1. Errors effect: to ensure that "quality is maintained among experiments" is a valid assumption. Basic statistics and graphical evaluations will be utilised for this matter. if the tests assumptions are correct. Nevertheless, it will also be convenient to evaluate the assumptions of homogeneity, normality, linearity, and additivity.

Questionnaire results:
to study the different criteria utilise to validate from a user perspective the usability and feasibility of the solution proposed. Results quantify qualitative responses. Hence, basic statistics and graphical evaluations will be used to analyse them.
The data collected and analysed with the protocol above comes from a case study from the aerospace industry. Conclusions should be discussed within that context. The case study is presented in the subsection below to provide the necessary context for the analysis conducted in following sections.

Case study
The case study comprises a remote diagnosis operation, which, in turn, comprises the messages to be sent from experts to technicians, and the equipment with which the operation is conducted. The case has been selected based on the data given by an aerospace company. The company provided over 20 hours of interviews to identify current processes and solutions in remote diagnosis, which were the basis for the design of the case study.

Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment 23
The remote diagnosis operation is a combination of different remote-driven tasks which are common in the context of the company (e.g. visual inspection, photograph, repair, etc.). These tasks can be performed either with an AR solution or with current alternatives such as emails or phone-calls. The messages from which tasks have to be inferred by technicians are presented in Table 6. Table 7 presents the factor levels of these messages according to validation criteria (object (OA) and procedure (PA) awareness). Figure 5 presents an example of the proposed solution to conduct the case study's operation.

Message Description
A Expert asks to unscrew the screws of the front panel of the fuel hatch and open it B Expert asks to visually inspect the right and left sides of the hatch and to take a photograph of every defect found C [Two defects should be found by tester] Expert asks to repair by placing the patch D Expert asks to take a photograph of the previous reparation result and send it by email  Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment 24 The equipment used in the case study is a prototype of an aircraft's fuel hatch that was provided by the company (Figure 6). It includes components that are common in the aerospace industry such as rivets or panels. It also has some pre-made defects which are common for the aerospace context such as dents (Figure 6.d), scratches and breaks (Figure 6.c). So, it can be said that the case study proposed is a fair representation of a common remote diagnosis operation in the aerospace industry. The case study was utilised for both, laboratory experiments and expert demonstrations. The sample size for the experiment can be calculated "a priori". In order to do it, an F-test for a three-way ANOVA experiment was applied. The test considers the number of groups to be 8 The results of both, questionnaire as well as the stopwatch time and errors study, are presented in the following section.

Results and discussion
The aim of the experiments was to collect data to demonstrate presented in 5.2. The validity of these hypothesis is evaluated through the analysis of the experimental results presented in the following subsections.

Error results
Messages with different complexity levels were pre-defined for the experiments. Errors were defined as the number of tasks completed by a tester, which deviated in form or result of what is declared by the corresponding remote message. Hence, if the remote message were the same, then the number of errors should not be different between AR and NOAR solutions. As part of the Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment 26 analysis, it was necessary to test this assumption in order to ensure a direct correlation between experimental times and remote diagnosis efficiency.
Error results are presented in Figure 7 and Table 8. Figure 7 shows the number of errors made by each tester in their experiments. The number of errors made by testers is low, with only 4 testers out of 30 committing more than one error. Moreover, if we consider the total number of tasks that the messages comprised (10), then the overall number is also very low as the maximum number of errors (3) made by a tester do not involve more than a third of the total tasks (33%).
The results in

Time results
Time results aim at evaluating the effect of utilising the proposed solution in remote diagnosis scenarios and its effect at different levels of message complexity. As time is considered to be a direct measure of the diagnosis efficiency, a reduction of time produced by the use of the solution proposed can be understood as an enhancement of remote diagnosis efficiency. Besides, it is also interesting to analyse the variation of such time reduction by the use of the research solution, if it exists, for different levels of remote message complexity. Message complexity is classified in terms of object and procedure awareness at four different levels: 1. Simple-Simple (Message A ( Table 6)): message is simple in terms of object and procedure awareness.

Simple-Complex (Message D)
: message is simple in object awareness and complex in procedure awareness.

Complex-Simple (Message C)
: message is complex in object awareness and simple in procedure awareness.

Complex-Complex (Message B)
: message is complex in terms of object and procedure awareness.
Time results are presented in Figure 8 and Table 9-Table 11 Table 9.
Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment  The effects and interactions between factors on the time variable can be further analysed by evaluating the difference in time reduction between solutions for each message complexity level (Table 10). For level "simple-simple" (message A), the AR solution shows a time reduction of 13% on average. For the "simple-complex" level (message D), time reduction for the AR compared to the NOAR solution is calculated at 32%. For "complex-simple" (message C) and "complexcomplex" (message B) levels, time reductions are averaged at 54% and 59%, respectively.
Post hoc comparisons results from the Tukey HSD test can reveal significant differences between different factor groups.   When procedure and object awareness are both simple, differences cannot be considered significant. Hence, it can be said that the higher the complexity of messages the better improvements AR solutions provide in terms of time reduction; and so, enhanced efficiency.

Usability and feasibility results
Questionnaires aimed at quantifying testers' and experts' opinions on the proposed solution usability, and experts' opinions on its feasibility for real-life conditions of remote diagnosis operations. Graphical and numerical results of these questionnaires can be seen in Figure 9 and Table 12, respectively.
In terms of usability, experts and testers were questioned about the ease of use, ease to learn and overall satisfaction when using the proposed solution as technicians. According to results, both groups agree that the solution proposed can be easily used by technicians. They all scored above 4 (out of 5) for every criterion. Additional usability results are related to the usability of each tool, provided by the solution proposed. Experts and testers were also questioned about the utility of the proposed AR solution.
Although, both groups agree the solution proposed is usable, the scores are lower on average compared to the usability criterions. This correlates with some of the comments provided by experts and testers who mentioned the need to provide more "professional-looking" interfaces that can make the communication "smoother". From a quality perspective, it was found interesting to ask testers and experts their expectations on how much the proposed solution could improve current remote diagnosis performances. According to the results presented in Figure 9 and The importance of the feasibility questionnaire was to determine whether real-life experts might consider that the solution proposed could be used in real-life remote diagnosis scenarios.
In order to do so, they were asked about the ability of the AR solution to be used in such scenarios and also in the ability of the data recorded to be re-used in other contexts (data usage). All experts agreed that the four tools (button interaction, message log, picture interaction and user interface) could be used by real-life technicians. Besides, only two of them disagreed with the idea that the message-related data recorded (data usage) could be later reused in other areas (e.g. repair design) or in other remote diagnosis operations.

Discussion
The objective of the analysis results presented in the previous subsection is provide enough evidence to validate the hypotheses presented at the introduction of Section 6.
First, the error results offered sufficient evidence to say that within the sample there is no An important aspect to note regarding the validity of the results relates to the sample size.
The "a priori" F-test conducted to determine the sample size required for the three-way ANOVA determined a sample size of 37 people. Although the final number of testers was 30, the final sample size is close enough to the "a-priori" calculated sample size to still consider the results valid. Besides, similar researches reviewed [30,39,41] also utilised similar sample size. Hence, the results of the questionnaires can also be considered significant according to the sample size.
Structured authoring for AR-based communication to enhance efficiency in remote diagnosis for complex equipment

Conclusions
The paper 1) defines a structure that regulates message elements to ensure their situational awareness, and 2) develops a rule-based authoring approach that automatically creates AR content of message elements to enhance remote diagnosis efficiency. The proposed AR-based communication contributes to filling an important research gap in the remote diagnosis context.
The developed system offers structured and real time communication methods using automated augmentation of message elements. The validation results indicate strongly that AR based technology can improve the efficiency in remote communication in terms of time and errors reduction, whilst also being efficient and valid in working environments.

Future works
Future work will explore how remote communication can be enhanced beyond the diagnosis focus in this paper. It is also possible to extend this papers' focus from a one-to-one based communication, to a one to many. Currently only one 3D model (complex equipment representation) can be loaded on the AR system, which limits the possibility to select alternative objects for guidance. An additional area of interest for future research is voice-to-text implementation to improve the usability for technicians. Also, future research can be in data mining for inferring maintenance procedures from the collated structured messages.