**6. Discussion**

The outcome of the current research work, is a multi-sided, multi-device application, enabling the remote maintenance of complex machine tools under the framework of Industry 4.0. The term multi-sided is used in order to describe that multiple operators can connect to the platform simultaneously and discuss on a malfunction that has occurred unexpectedly. On the other hand, the term multi-device, is used in order to describe the compatibility of the developed framework in a variety of devices. It is stressed out that the most common devices used for AR implementations are HDMs and handheld devices, including smart-phones and tablets. However, since the scope of the current research was mainly focused on the provision of a totally marker-less AR solution, the use of Microsoft's HoloLens HMD is preferred, since the device supports spatial recognition, which, in turn, has facilitated the development of the framework. It is stressed out that the use of the HMD is initially intended only for the shop-floor technician as a means to capture their surrounding and, based on that, the expert engineer is capable of placing/registering the virtual content on the user's physical environment. The development of the framework was performed in a set of consecutive stages. The first stage was based on the conceptualization of the system's architecture, the discussion of the modules comprising the architecture, and the tools to be used for the final development. Then, in stage two, the basic modules were developed, utilizing the software described in Section 4. As soon as the alpha version of the framework was released, the framework has been internally tested in vitro, in the lab-based machine shop described in Section 5. During this stage, colleague engineers have been asked to test the functionalities of the developed application either from the side of the expert engineer or from the side of the shop-floor technician aimed at gathering insightful feedback. Then, based on the feedback gathered during the previous stage, the framework has been improved in certain areas, regarding the layout of the GUIs and the functioning of the communication services. Then the beta version of the developed framework has been prepared and tested with the industrial partner in their premises.

Throughout the development and validation stages, meaningful feedback has been gathered. The main strengths of the proposed framework can be summarized to the minimum development effort required by the expert engineer. More specifically, the remote maintenance frameworks are based on the development of AR scenes, which are then implemented on an AR-ready device. Therefore, this requires advanced development skills and a certain amount of preparation time. On the contrary with the proposed framework, an expert engineer is capable of creating AR content on the user's FoV with the use of simple Drag-and-Drop actions. As such, no special skills are required and the time for AR content preparation and implementation is eliminated. Another strength of this framework is the ability to create online sessions. The online sessions are based on the Unity 3D multiplayer API offering the remote connection of multiple users. As discussed in previous paragraphs, the developed application is compatible with most devices. Thus, there is no practical limitation on the use of equipment. In comparison with other approaches, presented in References [3–6], the functionality of remote connectivity combined with the broad compatibility of the app further extends mobility of the user, i.e., the shop-floor technician in an industrial environment. More specifically, in Reference [4], the technicians have to carry a laptop computer along with a USB web camera in order to establish communication with the expert engineer and proceed with the maintenance operations. The frameworks presented in the above-mentioned publications are using frame markers for the registration of the content on the physical environment. With that said, the performance of the system relies heavily on the quality of the camera used for the detection of the marker. The user's mobility is limited, as the camera has to maintain visibility of the marker at all times. Otherwise, the augmentations cannot be registered and updated as the user moves around. In addition to that, the proposed framework, since it utilizes newer equipment with increased computational power, the use of more vivid/realistic virtual objects is feasible. Not only that, but in previous approaches, the augmentations are limited to the display of simple schematics, e.g., arrows, text boxes, or indicate the component of interest via a wireframe model. On the contrary, the current approach can recognize the objects by the utilization of Microsoft HoloLens.
