Keywords

1 Introduction

Current research contributions identify possible synergies from digitization and sustainability [1, 2]. In contrast to typical definitions of sustainability, which are mainly characterized by environmental, social and economic factors, there are also domain-specific factors. In the industrial context, sustainability also refers to areas such as emissions, products and services, transportation, appropriate working conditions, community relations, economic performance, and market presence [3]. These areas of industrial sustainability are enhanced by utilizing modern concepts of urban production (UP). UP is defined as a production system that creates value for real goods and services based on transformative processes of input flows in urban space [4, p. 28]. UP of sustainable products is enabled by city-friendly factories and production systems on behalf of all partners [5]. This paper presents a technology stack as an immersive tool for UP concepts with its direct contribution to sustainability. In addition, the reliability and performance of the system under the constraints of UP requirements and their indirect effects towards sustainability are discussed [6,7,8].

2 Related Work

The symbiosis of the UP provides a sustainable contribution to the optimization of the ecosystem and the minimization of negative effects in several aspects [5]. Workplace and residential areas can merge and the social benefits can offer a more flexible work-life balance [9]. Energy and exhaust heat recovered from industrial processes can be made available for heating private housing due to sector coupling for example [10]. A main strategy of UP rests in the efficient use of already existing infrastructure and its sealed areas for sustainable industrial use [11]. To improve the competitive market position and to enhance the regional economy, the local, decentralized potentials of manufacturers, customers, services and lean supply chains can be used efficiently [12].

One challenge is the ‘space dilemma’, which is the lack of industrially available space versus the increasing need for production space to ensure growth and value creation [11 p. 3]. UP fulfils the requirements for the sustainable repurposing of vacant buildings in the so-called ‘Greyfields’ to avoid unnecessary sealing of surfaces. The challenges of sustainable industrial parks in urban environments are not only in the appropriate site selection, but also in the implementation, considering constraints such as space and cost issues, regulatory issues, emissions, pollution, odors, noise, vibrations, planning, construction, logistics and traffic [12].The realignment of production to be designed with the involvement of all stakeholders, taking into account life-friendly conditions in terms of optimization, growth, customer integration and resource sharing [12].

For the assessment of potential locations and variants of the UP concepts, different methods are suggested. Augmented Reality (AR) can support the five stages of evaluation (analysis, scenario development, location selection, scenario evaluation and planning implementation) [13] and improve planning processes in general [14]. Stakeholders and developers of UP planning concepts can be displayed relevant, contextual information straight ahead in the field of view at the place of fulfilment [15].

3 Concept

The conclusion of the related work indicates that the objectives of the UP concepts can be sustainably supported by the use of augmented reality technology in decision-making and concept development. Based on the ‘ultra-efficiency approach’ [16], an immersive implementation is created. This allows the relevant interrelationships of the five areas, including material (white lines), energy (orange lines), people (yellow lines) and emissions (violet lines), to be displayed and examined in an AR simulation at the point of performance (Fig. 1) [16].

Fig. 1.
figure 1

Mockup of an AR UP scene with production and distribution close to the customers. In the scenario, both interconnection (highlighted purple emissions, orange energy, yellow people, white material) and value stream planning combined with sector coupling (utilisation of process heat and excessive energy for private residences) are based on [16]

Full size image

3.1 Research Demand and Research Question

In this context, the research demand focuses on the challenge of a resilient, usable AR planning application that fulfils the requirements of the UP dimensions including the necessary rendering performance. These challenges are met in a mixed reality application comparable to the mobile AR field [17]. The present research design is designed to meet the following research question: What is the maximum amount of augmented volume or how many 3D models can be rendered in one AR scene as a function of the level of detail (LoD) using ‘On-Device-Rendering’?

3.2 Implementation of the Application

The application (Fig. 2) represents an implementation of an AR system for visualization of complex, spatial planning scenarios. The goal is an AR implementation approach for UP application that can be provided for a lean, SME-suitable process to masters the trade-off of sufficient detail and quality of immersive information (3D models and contextual annotations) with sufficient system robustness. Different native formats CAD, BIM or 3D models have to be manually processed as mesh-based, textured 3D models according to a defined asset pipeline. The models are uploaded in the mixed reality application at runtime via a web-based AR content management system. Due to the specified requirements, such as a low-budget solution, low complexity and independence of technology providers, the approach of cloud rendering was not preferred. The following approach was implemented:

Fig. 2.
figure 2

System architecture of the implementation

Full size image

To render the geometric information in the HMD hardware, the first step is to load a 3D model onto the client at runtime of the application. For this, an asynchronous HTTP request is sent to the webservice by an user input. This serves as an abstraction layer for the provision of the file-based data storage of the 3D models. After the file has been successfully downloaded, an asset template is created, which functions as a placeholder for the geometric information. This template provides the application logic for interactions, which remains the same for all instances, and provides the adjustments to the visualization of the exchangeable geometric data. The loaded mesh and the respective textures are inserted into this placeholder and stored on the device for later instantiations. Afterwards, a concrete instance of the complete asset (e.g. 3D model of the UP planning scene) is created and added with meta information, including an optional scaling factor, an asset name and a dynamically generated instance ID. This information is listed via user interface for the identification of all instances loaded in the scene. Subsequently, the animation data optionally available in the loaded file is converted into animation objects and an animation dialogue is made available to them on the user interface. Finally, the instance is inserted as a node in the scene graph of the 3D engine and is located relative to the real environment using a spatial anchor.

4 Methodology

To explore the overall system capability (performance) of the aforementioned implementation, the following DOE is developed:

4.1 Design of Experiment (DOE)

The objective of the experiment is to examine the maximum number of 3D models considering predefined Level of Detail (LoD) in the industrial context. The number of simultaneously rendered 3D models and the LoD are defined as independent variables. As dependent variable the system stability is evaluated. Starting from zero, the number of 3D models is continuously increased. Four different LoD are defined by manipulating the number of polygons and resolution of the mono-texture (Table 1). As an indirect metric, the polygon density in the unit polygons/m3 was derived (Eq. 1) [18]:

$$\text{polygonal density = }\frac{\text{total polygone count}}{\text{augmented volume}}$$
(1)

The LoD is selected based on prior experience in the industrial context and the subjective quality assessment depending on the available model for the experiment. Starting from a high-quality model by the asset pipeline, the number of polygons and the texture resolution are successively reduced in percentage. The structural complexity of the object hierarchy is reduced to a singular level.

Table 1. Predefined LoD as function of number of polygons, density and texture
Full size table

To ensure the system stability, the displayable frames per seconds (FPS) are validated against the threshold 20 FPS by the system performance dialogue (current RAM load and FPS). Above 20 FPS, the usability of the system is assumed. According to the manufacturer's specifications of MS HoloLens 2, the system performance is affected by several influencing factors on 3D models [19]. Thus, the following parameters are defined to the 3D models: hidden and unused data are removed, drawing calls are reduced, inverted surface normals are reversed and conflicting tangent bases are solved.

4.2 Experimental Setup

The experiment is realised in a de-cored building (previously used for industrial purposes) on the first floor (cf. Fig. 3). The dimensions of the vacant area are 41.58 m in length and 31.60 m in width, which represents a total area of 1,314 m2. The available room height is about 3.05 m. In total, the room offers a volume of 4,007.48 m3. The 3D model represents a CNC milling machine (Index GFG 250). The test is conducted on a Microsoft HoloLens 2 (Windows Holographic V21H2, Build 20348-1432). The application was developed in Unity (Ver. 2021.1.9). For each LoD, a series of experiments is performed in which the number of rendered models is increased cyclically. After each additional model, the system stability (functionality) of the application is validated. The test ends with the performance-related crash of the application or dropping below the threshold of 20 FPS.

Fig. 3.
figure 3

Immersive screenshot of the experiment with the rendered 3D models

Full size image

5 Results

Twelve 3D models (corresponding to an immersive volume of 327.38 m3) of the LoD-100% in a total of 1.284 million polygons can be rendered stably. With the LoD-50%, 35 3D models in a total of 1.876 million polygons (corresponding to an immersive volume of 954.70 m3) can be reliably displayed. By the LoD-35%, 70 instances in a total of 2.625 million polygons and an immersive volume of 1,909.09 m3 can be resiliently depicted. The resulting number of 3D models by LoD-10% can not be determined in the given setting due to the lack of available space. The results suggest that the maximum total count of 3D models does not only dependent on the number of polygons, but also on the polygonal density, among other things.

6 Discussion

One of the paradigms of sustainable software engineering is the ‘limitations imposed by the state of technology (…) on the environment's ability (…) to meet present and future needs’ [7]. This recurs to the comparison of the study result to the specification of the hardware, in that the present implementation exceeds the number of reliably representable polygons by a factor of 8 compared to the reference implementation (‘low-scene complexity’, three models à 100,000 polygons) [20]. The direct effects on the software footprint, such as the energy consumption of the system can not be considered [6]. However, the implementation serves sustainability, as larger scenes can be mapped via On-Device-Rendering without relying on the resource-intensive cloud rendering. Basically, the AR system can be used for the rendering of large scale industry scenarios to support UP (planning). According to the rebound effect, a robust application leads to a positive user experience [6]. In this case the software contributes to effective UP concepts, due to its resilience. Especially towards the UP purpose, an immersive application supports users of different skill levels and roles in a sustainable way [7]. By providing relevant information at the place of fulfilment [8], avoidable misinterpretations in the evaluation and development of UP concepts can be reduced in early planning stages and communication barriers can be lowered between the stakeholders. Furthermore, it becomes apparent that the effective performance depends, among other things, on the build of the operating system in the industrial metaverse. In most cases, the approach is not standardized and is therefore described as precisely as possible in the interest of reproducibility. Further research with to be developed, standardized methods is recommended.

7 Conclusion

With the presented results, the research question can be answered precisely. If a comparable AR system such as the introduced approach is to be used in an UP planning task, analogous study results can facilitate the right choice of the LoD and the evaluation of a priori capabilities in order to ensure the successful application in terms of system stability. The presented solution as an immersive tool of the ‘Ultra-Efficiency Approach’ [16] can strengthen the technical UP methods with sufficient rendering performance. Thus, such a software system can be used sustainably and offers technological approaches towards the design of a powerful industrial metaverse. For further investigation of the technical sustainability, it is necessary to investigate the influencing factors and effects of specific industrial applications in terms of effectiveness, efforts, acceptance and other stimuli of human-machine interactions.