A Critical Analysis and Roadmap for the Development of Industry 4-Oriented Facilities for Education, Training, and Research in Academia

Abstract

The development of Industry 4-oriented facilities in academia for training and research purposes is playing a significant role in pushing forward the Fourth Industrial Revolution. This study can serve academic staff who are intending to build their Industry 4 facilities, to better understand the key features, constraints, and opportunities. This paper presents a systematic literature review of 145 peer-reviewed studies published between 2011 and 2023, which are identified across Scopus, SpringerLink, and Web of Science. As a result, we emphasise the significance of developing Industry 4 learning facilities in academia and outline the main design principles of the Industry 4 ecosystems. We also investigate and discuss the key Industry 4-related technologies that have been extensively used and represented in the reviewed literature, and summarise the challenges and roadblocks that current participants are facing. From these insights, we identify research gaps, outline technology mapping and maturity level, and propose a strategic roadmap for future implementation of Industry 4 facilities. The results of the research are expected to support current and future participants in increasing their awareness of the significance of the development, clarifying the research scope and objectives, and preparing them to deal with inherent complexity and skills issues.

1. Introduction

The manufacturing industry is facing intense challenges of competition, energy, labour supply, and instability in the current world [1,2]. Higher flexibility is required to enhance the efficiency of manufacturing processes and reduce the batch size of production to enable more personalised products [3,4,5,6,7,8]. Stronger adaptability is expected to shorten the time-to-market and withstand shockwaves from the market [2,9,10,11,12,13,14]. More reliable and secure remote access is desired to improve the maintenance performance and to lower costs [15]. Higher safety requirements are needed to take care of the human workforce’s well-being [15,16], and more.

To meet those expectations, the concept of Industrie 4.0 (or Industry 4.0) was introduced in 2011 in Germany [8,15,17]. The original intention of introducing Industry 4.0 was to enhance the competitiveness of the manufacturing industry in Germany by integrating several critical cutting-edge technologies, such as cyber–physical systems (CPSs), digital twins (DTs), Internet-of-Things (IoT), etc., within their manufacturing systems [18]. The presentation of the concept then caused a rapidly growing awareness of a reality that industries were evolving into the era of the fourth industrial revolution, or Industry 4 for short [12,19].

Although the definition is still not settled between academia and industry [20,21,22], the true philosophy of Industry 4 has been outlined by Schwab [1]. One commonly agreed-upon definition is that Industry 4 is a digital transformation by integrating CPSs into the whole value chain in the manufacturing industry [23]. It is believed that the boundaries between physical, digital, and biological domains, not only in manufacturing, but across the entire world too, will be removed in the era of Industry 4 [1]. We will use this definition for Industry 4 in this paper.

1.1. The Philosophy of Industry 4

The hallmarks of the past three industrial revolutions were the innovation and applications of certain technologies. Although technological leaps are still pushing the current revolution, it has been discussed that the philosophy of Industry 4 goes far beyond merely employing a series of Industry 4-related technologies [1,20]. Industry 4 is not only about implementing cutting-edge technologies but is also concerned with the development of the skills and competencies of the workforce [24,25]. In addition, Schallock et al. [26] also pointed out that novel social infrastructures are also a key focus of Industry 4 ecosystems from a global perspective.

The implementation of Industry 4 implies a digital transformation [27,28]. This can be achieved by demolishing the barriers between physical assets, geographical locations, and independent stakeholders alongside the whole value chain [29]. This in turn leads to an expected fusion of resources and expertise to shorten the product development periods, enable product individualisation, increase production flexibility, establish decentralised production models, and achieve industrial sustainability [19,30,31,32]. From a technological perspective, such fusion can be reached by integrating cyber–physical production systems (CPPSs), developing industrial internet-of-things (IIoT), employing ubiquitous computing units and sensors, embedding artificial intelligence (AI) algorithms, deploying advanced robotics, and more [33].

There are four main pillars in building an Industry 4 ecosystem in a manufacturing context. They include vertically integrated manufacturing systems, horizontally integrated global value chains, end-to-end engineering integration [34], and deployment of Industry 4-related cutting-edge technologies [35].

A new paradigm is being developed in the Industry 4 framework for designing factories and value creation structures [27], which requires practitioners to have interdisciplinary expertise [36,37]. Factories with Industry 4 implementations are known as smart factories [38,39,40,41], or factories of the future [7,42].

1.2. Research Aims

When considering the unstoppable and enduring trends that will see many viable industries forced into the era of Industry 4, manufacturing is the originator of Industry 4. Still, it has the greatest needs for its Industry 4 readiness of the upcoming digital transformation [19]. Thus, it is believed that building Industry 4 facilities in academia as education and training facilitators is a booster and a key enabler to facilitate the new paradigm transition, for academia as a producer of skilled workforce and industry that requires the workforce.

To the best of our knowledge, no research has been found focusing on reviewing the current development of Industry 4 facilities in academia. This paper aims to present a structured and critical analysis of the global development of Industry 4 facilities in academia, with a focus on their roles in education, training, and research.

There are five research questions that have been set in this research as below:

• RQ1: Why is it important to build Industry 4 facilities in academia?
• RQ2: What features and characteristics of the Industry 4 system shall be considered in the development?
• RQ3: What Industry 4 technologies can be used in the development?
• RQ4: What challenges are expected during the development?
• RQ5: What are the current research gaps and solutions?

1.3. Research Contribution and Novelty

This paper provides a multi-dimensional framework for current and future users of Industry 4 with meaningful insights for building an Industry 4 facility in academia. The insights include the significance, design principles, core technologies, and potential roadblocks in the development. This will help the participants attain an increased awareness of the new paradigm, while outlining the core framework, understanding the available technologies, and being prepared for the inherent challenges of transitioning to a new way of doing business.

In addition to the review of existing development in the selected literature, this paper, in Section 8, also uniquely synthesises the literature using an analytical lens focused on the maturity and deployment complexity of core Industry 4 technologies, implementation bottlenecks, and underexplored areas. Moreover, a five-stage roadmap is proposed for the development of Industry 4-oriented facilities in academia. This high-level strategic plan can guide the academic participants to incrementally develop the expected institutional readiness, stakeholder engagement, and technical capabilities, which make it a valuable tool for the participants to build an Industry 4 ecosystem in an academic context.

This paper is structured as follows: Section 2 introduces the research methodologies; Section 3 discusses the main types of Industry 4 facilities and their significance, and Section 4 reviews the design principles of building an Industry 4 ecosystem. This is followed by Section 5, which provides a review of the main Industry 4-related technologies that have been used in the current literature. Then, Section 6 highlights the challenges that have been faced in building Industry 4 facilities. Section 7 presents critical discussions on current research gaps, technological maturity, and a proposed implementation roadmap in Section 8. The research limitations are discussed in Section 9. Finally, Section 10 concludes the paper.

2. Review Methodology

2.1. Selection of Databases

University library databases, including SpringerLink, Scopus, and Web of Science, were used as the major sources for the articles analysed, as they cover a broad range of peer-reviewed journal papers and conference papers in the field of industrial engineering and engineering manufacturing.

2.2. Keyword Selection and Boolean Strings

There were two groups of keywords used in searching the articles. One group included the keywords related to the development of an academic facility, including learning factory, learning laboratory, and Testlab. The other group of keywords related to Industry 4, which was used for filtering out the works that are not relevant to Industry 4. The keywords included the term “Industry 4” (Industry 4.0/Industrie 4.0), and some derived terms relevant for the research, including “digital transformation”, “smart factory”, “intelligent manufacturing”, “DT”, “CPS”, “IoT”, “intelligent robotics”, “machine-to-machine” (M2M), “AI”, “cloud computing”, “machine vision” (MV), “virtual reality” (VR), and “5G”.

The search strings used were the keywords in group 1 and group 2. One keyword from each group was used at a time. So, there were 52 combined search strings for the initial search. The Boolean string operator was used for effectively conducting the search. In addition, the range of publication years was selected from 2011, which is the year in which the concept of Industry 4 in Germany was named, to 2023, which is the last year with full publication statistics. Also, the types of articles which were considered included journal papers, conference proceedings, and book chapters, published in the Engineering discipline and in English only. The search strings and initial search results are summarised in

Table 1. Search Boolean strings/keywords used and the initial numbers of identified articles.

Database	Boolean Strings	No. of Articles
SpringerLink	(“Learning Factory” OR “Testlab” OR “Learning Laboratory”) AND (“Industry 4” OR “Digital Twin” OR “Smart Factory” OR “CPS” OR “CPPS” OR “IoT” OR “IIoT” OR “Intelligent Manufacturing” OR “AI” OR “Machine Vision” OR AR OR VR OR 5G)	325
Scopus	TITLE-ABS-KEY ((“Learning Factory” OR “Testlab” OR “Learning Laboratory”) AND (“Industry 4” OR “Digital Transformation” OR “Digitisation” OR “Smart Factory” OR “Intelligent Manufacturing” OR “Digital Twin” OR “CPS” OR “CPPS” OR “IoT” OR “IIoT” OR “Intelligent Robotics” OR “Machine-to-Machine” OR “AI” OR “Cloud Computing”))	218
Web of Science	TS = ((“Learning Factory” OR “Testlab” OR “Learning Laboratory”) AND (“Industry 4” OR “Digital Transformation” OR “Digitisation” OR “Smart Factory” OR “Intelligent Manufacturing” OR “Digital Twin” OR “CPS” OR “CPPS” OR “IoT” OR “IIoT” OR “Intelligent Robotics” OR “Machine-to-Machine” OR “AI” OR “Cloud Computing”))	120

.

2.3. Article Filtration and Selection

There were 663 articles in total that were identified after the initial database search by using the Boolean strings as outlined above. Then, some filtering strategies were applied to narrow down the search results. Firstly, duplicated results from different searches were removed to lower the number to 375. Secondly, the title and abstract of the articles were screened to filter out further irrelevant articles, which reduced the total number to 193. In this stage, articles that were not related to the educational and research environment were excluded. Finally, the full content of the articles that remained from the previous stage was investigated to finalise the list of articles for review. In full-text screening, articles that discussed both Industry 4 integration and educational facilities development were considered. As a result, 145 articles that studied the development of Industry 4 facilities in the academic teaching field were selected and reviewed in this research. Another set of 22 articles that discussed Industry 4 systems generally and the related technologies was included in the review. A PRISMA-style flow diagram for the screening process is demonstrated in Figure 1.

A grouped summary of the 145 reviewed articles that were organised by technology, design principles, and key challenges is provided in Appendix A.

3. The Significance of Developing Industry 4 Facilities in Academia

Building Industry 4 facilities in academia has great significance for both academia and industry. It is a key enabler for the introduction of the new paradigm, as education and industry have strong connections and are correlated with the development of the economy and society at large [43,44]. The establishment of such Industry 4 facilities in academia can close the gap between the educational strategies and industrial demands [35,45,46] and push the technological leap to the next stage [47,48], which is of importance for all stakeholders [11].

Industry 4-oriented facilities are designed to offer a platform for training, education, and research purposes in the Industry 4 era [31,49], which is important for science, industry, and society [50]. A well-established Industry 4 facility can also attract better graduate students and prepare better researchers, which in turn can foster the academic reputation and influence of universities [51], in a virtuous cycle.

3.1. Types of Industry 4 Facilities

There are two main types of Industry 4 facilities in academia: learning factories and Testlabs. However, it can be seen through the statistics in this literature review that the development of Industry 4 Learning Factories is shown in significantly greater numbers than the Testlab concept implementation.

Most of the development of learning factories considered the integration of the Industry 4 philosophy as soon as the concept of Industry 4 was proposed in Germany [16]. The framework of learning factories can perfectly fit into the Industry 4 philosophy to build a pilot-scale smart factory. They are used for learning and research purposes, as well as for industrial practices, which gives learning factories great potential to demonstrate the processes of integrating the digital transformation [52,53]. That is the main reason why Industry 4 learning factories, also known as smart learning factories [54], have become the main type of Industry 4 facility in academia and in industry.

The Industry 4 learning factories are targeted at facilitating education, research, and relevant vocational training with the aim of demonstrating, developing, and integrating the digital transformation [55,56]. In particular, Industry 4 learning factories can contribute to academia with a physical environment for undergraduates to conduct interactive workshops, a testbed for higher degree research students to work out novel concepts for their research and theses, an incubator for industrial participants to conduct collaborative and pilot projects, and a platform for local communities to hold technical events and demonstrations [52,57,58,59]. In addition, some challenges in the traditional teaching paradigm of learning factories, such as unaffordable maintenance costs and rigid teaching strategies, can be mitigated by implementing Industry 4 principles [15].

In contrast, publications reporting on the development of Industry 4 Testlabs in academia are significantly fewer than learning factories. Nevertheless, it is true that the development of Industry 4 Testlabs has the same goals of pushing Industry 4 transformation in both academia and industries, but with a slightly different focus. Schlette et al. [60] developed an Industry 4 Testlab for developing smart robotic assembly strategies. Similarly, Protic et al. [61,62] demonstrated their preliminary works in developing an Industry 4 laboratory to enable robotic flexible assembly strategies by employing collaborative robots and a digital twin system.

Both Industry 4 learning factories and Industry 4 Testlabs can offer participants project-based activities for teaching, research, and training purposes. However, Industry 4 learning factories emphasise more on integrating the whole infrastructure of Industry 4 ecosystems to facilitate educational and research activities. In contrast, Industry 4 Testlabs focus more on showcasing the concept and individual technologies at this stage.

3.2. Facilitating Academic Teaching and Research

Due to the complexity and interdependence of different operations of manufacturing processes at present, it is challenging for undergraduate students to attain the necessary skills in a traditional classroom [63]. It is believed that action-oriented and experiential learning methods have more significance and likelihood of advancement compared to regular teaching methods [43,52,64,65]. Thus, activity-based learning paradigms, with variable and increasing complexity, that can be developed and implemented in a learning factory, have been created for engineering students. The intention was to provide a platform for interdisciplinary engineering students to carry out capstone projects [38]. Louw and Droomer [55] also pointed out that undergraduate students can gain valuable competencies through learning factory activities, such as problem solving, project management, prototyping, etc.

Building Industry 4 systems needs interdisciplinary knowledge, which is not accessible via traditional classroom activities [6,66]. Students can gain valuable Industry 4 required competencies through the activities in a dedicated facility, in a co-evolutionary approach, where project-based teaching methods are employed [11]. It allows students to not only learn Industry 4-specific technology knowledge, but also to practise the design of the organisational structure of a manufacturing context undergoing digital transformation [2,6,67], which makes students career-ready once graduated. Ideally, students will not need to learn from a future workplace, but contribute to the knowledge of the company as new employees.

Leal, Fleury, and Zancul [18] pointed out that current Industry 4 learning factories focus more on a research perspective than undergraduate learning projects. The developed Industry 4 facilities can be used as testbeds for developing new technologies and the related implementation in a real industrial context [4,10,16,37].

Furthermore, students can gain experience in solving real-life problems in collaboration with local industries [66,68], which benefits both graduate students and local industry. The vertically and horizontally integrated manufacturing functional layers and modules with digitised processes enable remote accessibility for all participants to seamlessly collaborate [66].

In return, the facilities can also be continuously developed via the project activities, as developing and maintaining the facility is also a part of the work and experiences of the students [24,39,42,69]. Also, challenges and limitations in transdisciplinary study can be mitigated by implementing the principles of Industry 4 [70].

3.3. Boosting Industrial Collaboration

In recent years, industrial practitioners have shown increased interest in the ongoing digital transformation [71]. This is particularly true in Europe [26,41,72,73]. However, it is challenging for practitioners to find a way to acquire and integrate the related resources for the upcoming digital transition [26,45,74]. Special skills, knowledge, and competences are required [25].

The situation is particularly challenging in small and medium-sized enterprises (SMEs), which are suffering a shortage of qualified personnel to help themselves to facilitate the digital transformation [10,72,75,76,77,78,79]. Also, they are not capable of building their testbed from scratch due to high entry barriers [6,32,80]. Thus, low-cost digitisation transformation solutions are desired to facilitate the development of digital competencies for SMEs [24,79,81].

There are a few significant and multidimensional roadblocks for companies to implement the digital transformation towards Industry 4, such as a lack of clear scope, pathway and methodologies [32,81], uncertain return on investment [12,68,82], lack of competencies of employees [45,50], and low new-paradigm awareness and acceptance [41,56,83]. Thus, companies are eager to seek adequate training solutions for their employees to gain digital competency to achieve their expected digital transformation [84].

The training for industrial participants in learning factories focuses on both technological and organisational aspects [73]. Industry 4 learning factories can provide a sandbox for each scenario to outline the specific business and organisational framework for training and research purposes [32,66,85]. Centea, Singh, and Elbestawi [41] have showcased a number of collaborative projects with local industrial participants in their Industry 4 learning factory. An Industry 4 learning factory is also a good place to practise IT security strategies in a relevant environment [81].

Moreover, it is costly and inefficient for companies to provide physical on-site training for the employees [84]. Equipped with VR technologies, an Industry 4 learning factory can offer virtual working environments to facilitate training activities and overcome the obstacles in traditional training methods [86,87]. The configuration of a virtual training context requires much less effort than the traditional processes [71]. One virtual learning factory can be simultaneously configured with multiple products for training different participants [87]. Such facilities can also act as an industrial solution incubator to pilot Industry 4 projects for local industries [45].

Although the evolution of digital transformation heavily depends on cutting-edge technologies, the human workforce still holds the core competence of the industry [5,88]. The learning factory is also a perfect place for analysing human behaviour-related problems [89]. Such sensitive data is hard to ethically and safely collect in an active manufacturing environment. But a properly designed learning factory can help to provide conclusive data, and that can be used to enhance the performance of human operators on the factory floor.

4. The Design Principles

The design principles of Industry 4 offer a fundamental framework for establishing an Industry 4 ecosystem. Although the evaluation of Industry 4 maturity has been discussed for a decade, there is no conclusive consensus on what principles and features a true Industry 4 system must include [28]. Nevertheless, four high-level design principles have been outlined for building Industry 4 ecosystems in the manufacturing industry [21]. They include interconnection, information transparency, decentralised decisions, and technical assistance. Studies by Prinz et al. [88] and Canas et al. [90] have also emphasised the same functional structure for building Industry 4 ecosystems. The principles are mainly applicable to the manufacturing industry but can also be applied to the logistics industry [28].

4.1. Interconnection

In the Industry 4 paradigm, machines, devices, and systems are interconnected and capable of exchanging data with each other within the manufacturing context, which is the cornerstone of Industry 4 ecosystems [91,92]. CPSs facilitate seamless M2M communication and data exchange within smart manufacturing environments [17,93]. Building CPSs with cross-domain networks in smart manufacturing contexts is a consequential solution for gaining interconnectability in compliance with Industry 4 visions [28,43].

The design principle of interconnection can be interpreted as building interoperability in smart manufacturing systems, which is a way to enhance the performance of system integration, scalability, flexibility, and security level of the systems [90,91,94,95]. Achieving the desired bi-directional M2M and machine-to-human (M2H) communications heavily relies on enabling data sharing in real time, which is the central principle of interoperability.

Industrial communication protocols play significant roles in supporting instant data exchange for building interoperability in manufacturing systems [96]. In a broad range of applications, the Industrial Ethernet Transmission Control Protocol/Internet Protocol (TCP/IP) is used for building the networks in smart manufacturing contexts [7,16]. On top of that, OPC Unified Architecture (OPC UA) and Message Queuing Telemetry Transport (MQTT) are the most commonly used industrial communication protocols for enabling communications for both hardware and software platforms [40,97,98,99].

From a different perspective, the interconnection can also be presented as the foundation for enabling systems integration alongside value chains in smart manufacturing contexts [83]. The horizontal, vertical, and end-to-end integrations are the facts that distinguish the Industry 4 paradigm from its three predecessors [83]. Well-established integrations of Information Technology and Operational Technology (IT/OT) components in manufacturing contexts are an essential Industry 4 expectation [59].

With the support of advanced IT infrastructure, horizontal and vertical integration enable flexible production models across the value chain and organisational levels [46]. Horizontal connection links all field-level devices, vertical connection permeates through all organisational levels, and there is also cross-linkage between physical assets and their virtual representation [88].

From a local perspective, horizontal integration targets the linking of machines and devices at the shop floor level to enable communication between machines over various field bus technologies, e.g., Modbus, ProfiBus, and CANopen [6,24]. From a global point of view, horizontal integration is achieved through the whole value chain [35].

Vertical integration starts from the Enterprise Resource Planning (ERP) level to the field level via Industrial Ethernet systems to enhance the real-time data transfer and acquisition [6,43]. Fieldbus and industrial communication techniques are commonly used to enable vertical integration, crossing all levels of the manufacturing system, to establish a backbone for Industry 4 systems [76]. Vertical integration can also enable the digitisation of manufacturing processes [17].

In smart manufacturing contexts, end-to-end integration refers to the fusion of all production processes alongside the entire value chain, from the raw material suppliers to the end customers [100]. The connection enables seamless data acquisition through all stages of the product lifecycle, achieving greater efficiency, flexibility, and customisation [7]. For example, an online product configurator is developed in the Textile Learning Factory 4.0 to allow choice and ordering of personalised products [12].

4.2. Information Transparency

Information transparency is the second key design principle in building Industry 4 systems, which can be achieved through the establishment of interconnection. The interconnected networks and fully digitised systems allow data to be instantly collected from different machines, devices, and platforms in manufacturing systems. Making use of smart sensors, such as Radio Frequency Identification (RFID), is a common method to retrieve data and track products in Industry 4 systems [26,93].

Raw data collected will then need to be interpreted into high-value information [5,21]. The interpreted data can reflect the actual production status and can be used for making decisions in production planning and control strategies [2,16]. Also, Meissner et al. [85] have discussed the potential value of collecting real-time data for the benefit of shop floor management. In addition, Vogt et al. [101] and Zarte and Pechmann [16] have discussed the applications of using real-time shared data in managing energy consumption.

In the manufacturing context, a DT is a consequential application of achieving information transparency, which can be developed for both discrete systems [45,62] and process-based systems [97,102,103]. DT provides production insights by projecting a physical working context into a virtualised virtual replica for monitoring the production processes [45,104]. In various DT applications, acquired data can be visualised onto physical dashboards or implemented into a virtual context to summarise the status of machines and tools and resource allocations [38,79].

Data management is playing a significant role in achieving information transparency in Industry 4 systems. The methods for storing data can be as easy as using tabular forms [2,3,65,105] and Microsoft Access [2] for small projects, or using SQL-based software (Microsoft SQL Server) for complex systems [98].

4.3. Decentralised Decisions

After establishing the interconnections required to gain information transparency, achieving decentralised decision-making capability is the next expectation in building an Industry 4 system [21]. This, in turn, enables the manufacturing systems aspiring to be or become autonomous or smart. Cohen et al. [106] have also outlined a pyramid to support the inherent progression of Industry 4 principles. However, the work related to the top tier in the pyramid received the least contributions in the literature identified, accessed, and reviewed in this research.

Highly intelligent and flexible manufacturing processes are intended to satisfy fast-changing demands in the market, which in turn require enablers of batch-size-of-one production models for mass individualisation and customisation of the products [3,4,26]. Simons, Abe, and Neser [15] discussed an attempt at employing RFID and data matrix codes in the intralogistics system to individually identify products. However, Schuhmacher and Hummel [4] believed that the material flows of CPPSs in a smart factory are highly dynamic and changeable, which require decentralised planning and control strategies to be involved.

It is believed that establishing decentralised decision-making capacity requires the use of certain technologies and enhancing related software and hardware components in manufacturing contexts. Single-board computing units can be employed across the factory floor to enhance the infrastructure for supporting decentralised decision-making strategies [7]. Also, IoT and IIoT are considered to have great potential to contribute to building decentralised control systems in learning factories [4]. Blockchain technology can be used for secure sharing of information along the value chain and product life cycle, to enhance decentralised decision capacity [44]. Kumar et al. [37] discussed an application of machine vision in their learning factory to enable a decentralised decision-making strategy for quality control purposes. Also, Erdmann et al. [107] developed an enhanced manufacturing execution system (MES) to enable a decentralised control strategy for the intralogistics system of their learning factory. Advanced AI algorithms are essential components to boost the performance of decentralised decision making [37,107].

4.4. Technical Assistance

Human participants are expected to be assisted by the smart systems in operational and decision-making processes. The benefits of technical assistance are mainly directed in five directions, including production process optimisation, quality management, asset management, energy management, and training.

With Industry 4 architecture deployed, the production errors can be identified and rectified faster in smart manufacturing contexts. This is due to the traceability of products and parts that can be significantly enhanced in the digitised processes [108,109].

The visualised digital data can be a facilitator for optimising the value stream in the assembly cells [78]. For example, it can be achieved by analysing bottlenecks between the assembly stations and then shifting the assembly task to idle assembly stations [79]. Also, most of the simulation software has integrated analytic tools for evaluating and optimising business and manufacturing processes, which helps practitioners with making data-driven decisions [38].

Data-driven strategies in quality management have been applied extensively in current manufacturing contexts [110]. The process of product quality inspection can be facilitated by applying 3D scanning in a VR environment [97]. In-line quality control can be achieved by integrating MV into MESs [56].

The health condition of assets is a critical aspect for any industry. Predictive maintenance, as a subgroup of proactive maintenance, has great potential in managing assets in the manufacturing industry. One of the key methods to enable predictive maintenance is to monitor and analyse the data that is acquired and streamed from physical assets [111]. Monitoring performance can be significantly enhanced by deploying machine learning algorithms to analyse the data acquired from production lines [112]. Augmented reality (AR) and 3D printing technologies can also enable and enhance regular maintenance work [48,88].

Energy efficiency management can be visualised for people to better understand and fine-tune the systems, to achieve optimal strategies [6,42,76,113]. The data acquired for energy consumption can be used for production planning and control [16].

Visualising the learning context is a good way for learners to obtain a good idea of the system at first contact [64]. The integration of AR has become a common feature in the manufacturing process to help human operators quickly pinpoint the issues within the systems or sub-systems via tablets or smart glasses [88]. On the other hand, virtualised learning environments have stronger resilience capabilities and are capable of helping to mitigate the impact of a pandemic, such as COVID-19 [114]. A virtual learning context can provide a hazard-free environment for all participants [31]. Also, in the virtual environment, the factory layout can be easily reconfigured to fit various organisational structures and business processes for different training scenarios [38]. Immersive 3D virtual context can significantly enhance the training experience for students and industrial practitioners [38]. Wank et al. [17] have proposed an assistance system for assembly workers to enhance their work performance by providing individually tailored assistance.

5. The Technologies

Since Industry 4 is still a technology-driven industrial revolution, technology employment is a crucial factor. The Reference Architectural Model Industry 4.0 (RAMI 4.0) has outlined the foundational technologies that drive the new Industry 4 paradigm [8,109]. Thus, listed below are eight core technologies that were represented in the reviewed articles, which are summarised in this section.

5.1. Cyber–Physical Production Systems

A CPS is a foundational technology in Industry 4 ecosystems, which can be implemented in a broad range of industries [2,18]. CPPSs are the application of CPSs in a manufacturing context [2,37,41], a key enabler for achieving smart manufacturing [16].

The formation of a CPPS can demonstrate the established interconnections crossing all levels of manufacturing systems [7]. Information transparency can then be achieved by collecting and visualising the production data via wired and/or wireless networks [2,97]. Finally, the analysed and interpreted data can be used for enabling decentralised decision-making strategies to achieve the intelligence of the system [7,115].

The development of CPPSs can significantly enhance the flexibility and adaptability of the systems, which in turn enables the desired batch-size-of-one production model [3,21]. For example, Schulz et al. [64] discussed that scalability and modularity are the main considerations while building the Proto Learning Factory to enhance the changeability of the facility. Similarly, Rokoss and Schmidt [116] pointed out that changeability can also be facilitated by digitising the physical processes. A digitised production system can help the deployment of the desired scalability and modularisation [103,117].

CPPSs can be applied to core manufacturing operations as well as supporting systems such as storage [101,118], and material handling and intralogistics systems [2]. The structure of a CPPS includes the manipulators, control stations, material handling systems, and data management systems [39,42,119].

There are three main types of machines that are employed, as shown in the papers reviewed for this research, namely CNC machines [2,3,7,11,63,120], collaborative robots [7,9,11,63], and 3D Printers [7,11,41]. A centralised database can be used to interconnect manufacturing processes and business processes [38].

Communication techniques play a significant role in a CPPS. In the reviewed papers, there are a number of communication systems/protocols that were considered in building Industry 4 facilities. MTConnect and OPC UA are the most commonly used communication standards for building CPPSs [61]. Alternatively, Singh, Centea, and Elbestawi [42] discussed the use of a virtual network computing (VNC) connection and Modbus in the CPPS of the SEPT learning factory. The use of the TCP and user datagram protocol (UDP) and a comparison between them were discussed in building a CPPS learning factory [7].

Lastly, ERP software, as the top tier in the vertical integration of a CPPS, has also been discussed in the reviewed articles, as it is used when people design CPPSs in Industry 4 facilities. Faller and Feldmuller [6] used SAP, a top-level ERP solution, to establish the vertical integration in their learning factory. However, due to budget constraints, low-cost ERP software, called CANIAS, was considered in building the Ostfalia learning and innovation factory [80]. Similarly, Mladineo et al. [40] have employed MS SQL-based ERP software (called Venio ERP) as the top layer of the vertical integration in their learning factory.

5.2. Advanced Robotics

Industrial robots were one of the iconic hallmarks in the era of automation and mass production [98]. Robots can operate in tougher working conditions than humans, such as areas with limited space, high temperatures and high humidity, and exposure to hazards [101]. Nevertheless, their rigid manipulators could only be employed for some highly repeatable tasks at the early development stage (e.g., welding, painting, pick and place) [11]. However, it is believed that the significant development of advanced and smart robotics in recent years is a key solution to relieve the difficulty of hiring human labourers for low-skilled, repetitive tasks in the era of Industry 4 [1,11,53,121].

It is required for industrial robots in the Industry 4 era to become smarter and more adaptive, which can be achieved by integrating more effective control and programming methods for industrial robots [47,122]. Ogbemhe et al. [53] have discussed a strategy to offer a novel offline programming method. Jin, Marian, and Chahl [3] have proposed a novel control strategy to enable batch-size-of-one production in robotic flexible assembly cells. Furthermore, another work by Jin, Marian, and Chahl [13] has verified a new redundancy strategy that can be implemented in robotic flexible assembly cells to mitigate the impact of individual robot failures.

Collaborative industrial robots, also known as cobots, have been broadly deployed in a smart manufacturing context in recent years [123]. In the reviewed articles, cobots are one of the major manipulator classes deployed in academic Industry 4 facilities. There are two types of application scenarios for using collaborative robots in manufacturing environments. One scenario is to configure cobots to work alone with limited or no interaction with human operators. Another one is to achieve seamless integration and collaboration between humans and cobots in manufacturing.

The employment of collaborative robots has the potential to significantly reduce the total assembly processing time [7,9] and enable cobots to handle more complicated tasks [3,11,49]. End-effector changes can be automated using predefined algorithms integrated into robotic control systems [3]. In addition, with the integration of AI, machine vision, and mixed reality (AR and VR) in robotics, there is even greater potential for robots to contribute to manufacturing processes [11]. For example, Sievers, Neumann, and Tracht [86] have investigated an application of using cobots in a mixed-reality learning environment.

In addition to cobots, mobile robots are the other type of robotics. Automated guided vehicles (AGVs) are typical examples in this category, which can offer flexible and autonomous material handling services on the factory floor [26,39,41,67,97,107]. In some cases, collaborative robots can be mounted onto an AGV to perform specific tasks [49,107].

5.3. Additive Manufacturing

Additive manufacturing (AM) technology, also known as 3D printing, is one of the most advanced technologies deployed in current manufacturing contexts. Due to their low cost-effectiveness, the additive manufacturing technologies have not been broadly deployed on factory floors for mass production [124]. AM allows for the fabrication of geometrically complex parts in a single production step [51], which makes AM a perfect technology for applications like rapid prototyping in the aerospace industry [111], medical implants [125], 3D scanning [126], etc.

Also, new materials have been unlocked to be used for 3D printing technology, which in turn fosters the research in new materials development [127,128]. Moreover, the deployed 3D printing machines can be networked with other equipment to test and develop the desired interoperability in Industry 4-oriented facilities [51].

5.4. Digital Twin

A DT is a virtual replica of the whole or partial physical system [129]. It is a key enabler of an Industry 4 ecosystem [130,131]. In smart manufacturing contexts, DT systems can be used for simulating assembly strategies [62], simulating and managing intralogistics systems [66], monitoring and optimising real-time production processes [131], managing energy consumption [17], facilitating training programmes in Industry 4 learning factories [131], and more.

DT systems are capable of bi-directionally transmitting data between the physical assets and the virtual representations, which is the key distinction when compared to the digital model and digital shadow of the system, respectively [131]. There are three major considerations in designing a digital twin system in academic Industry 4 facilities: 3D virtual environment, communication protocols, and control methods.

Firstly, a 3D virtual environment is a platform for creating digital representations of the physical assets in Industry 4 facilities [82,129]. Most of the users choose plug-and-play 3D simulation software to establish the digital context, including Tecnomatix Plant Simulation [33,71,107], Siemens NX MCD [62], Siemens CIROS [132], Unity3D [32,114,133], fastSUITE [98], AnyLogic^® [65], Emulate3D [58], ExtendSim9 [134], VENIO ERP/PLM [135], FlexSim [38], Delmia [53], VEROSIM [60,136], 3DEXPERIENCE [67], Robot DK [11,129], and ROS [137].

On the other hand, some practitioners prefer to make long-term investments for developing their virtual platforms, which offer the teams greater freedoms and potential to broaden and deepen their research [102,138].

Secondly, there are two common instant communication protocols that are used in the current development of Industry 4 facilities for building digital twin systems, which are OPC UA [5,6,60,129,132] and MQTT [132,133]. Despite its long development period, ModBus still plays a key role in the Industry 4 digitised context [6,129]. For example, recent research by Savchenko et al. [132] demonstrated the use of CIROS, Node.Red, and OPC UA to create a digital twin system in their learning factory for business and engineering students.

Lastly, on top of communication protocols, a suitable control method is also a key enabler for building an effective digital twin system [138]. Middleware software can be placed between the physical assets and the digital replica to stream the bi-directional data transmission [139]. For example, Al-Geddawy [129] discussed the application and the advancement of using CODESYS as control software. Schlette et al. [60] employed the Robot Operating System (ROS) as the middleware for controlling their robotic assembly cell. Protic et al. [62] developed a customised control agent in Python (3.9.18) to coordinate the communication between the OPC UA server and client modules. Also, RFID and QR Code technology can be used for providing tracking capabilities in digital twin systems [67,140].

5.5. Mixed-Reality Solution

VR and AR are two key solutions for building digitised manufacturing environments in the transition to the Industry 4 paradigm [141,142]. VR is a computerised virtual context for people to immersively interact with real-world objects [132,143]. On the other hand, AR is a technology for enabling the enrichment of physical objects by embedding digital information that can be visible to the users via various smart mobile devices [14,131].

The employment of AR and VR is called a mixed-reality solution, which paves the way to offer a new paradigm of learning and training methods in Industry 4 facilities [131,144,145]. With the employment of AR and VR technologies, new collaborative processes can be achieved through the entire product lifecycle [66,146,147]. For example, AR is mainly integrated in production processes to highlight the system’s statistical information, and in assembly and asset management processes to provide dynamic and step-by-step instructions of operation [14]. On the other hand, VR can significantly enhance the training performance in a data-driven context [148,149].

5.6. Big Data Analytics

The paradigm shift of digital transformation towards Industry 4 implies that all processes are data-driven [4]. The volume of data collected from manufacturing processes can be huge, which is the reason why the term “big data” is used [2]. The intention of employing big data analytics is to interpret the massive raw data into meaningful information, which in turn enables information transparency in line with the Industry 4 design principles.

The interpreted data can be used to help participants make data-driven decisions. In a smart manufacturing context, the use of AI can provide new contributions to decision-making processes by analysing and interpreting massive amounts of data collected from the whole product lifecycle [63]. Advanced AI algorithms (e.g., machine learning, neural networks) can facilitate the analysis of large volumes of data to identify the trends, patterns, and correlations at high speed [150]. Thus, AI-powered data analysis methods are broadly used and are an important contributor and differentiator in the development of Industry 4 facilities.

There is a broad range of applications that are related to big data analytics in academic Industry 4 facilities. In the reviewed articles, there are seven main data-driven applications that have been discussed.

The application of production process optimisation received the most contributions in the reviewed papers. Yang et al. [112] applied machine learning algorithms to monitor the working status of machine tools, which helps with optimising production processes. Louw [47] discussed the application of AI-empowered data analytics to reduce waste in production processes. Vogt et al. [101] proposed a method for optimising the performance of cooling systems by continuously monitoring the temperature and humidity of the workpieces. Lang et al. [58] demonstrated the implementation of Dijkstra’s algorithm to optimise production processes. Bao, Zhang, and Jia [99] introduced the application of production monitoring by analysing real-time production data in their 5G learning factory. Buergin et al. [30] proposed a strategy to enhance the system’s adaptability by monitoring the real-time system performance against the pre-set KPIs for each assembly station. Perdana et al. [54] proposed a method for enhancing production process automation by analysing ultrasonic testing data to position jigs. Erdmann et al. [107] employed genetic algorithms to optimise intralogistics performance. Similarly, Zhang et al. [63] introduced an application using an AI algorithm to enhance the performance of drone navigation in the intralogistics system. Seitz and Nyhuis [2] discussed the application of big data analytics in the developed CPPS of an IFA learning factory for enhancing the performance of production planning and control.

Quality management is also a fundamental discipline that can significantly benefit from employing AI-powered big data analytics. Yang et al. [110] proposed a methodology to monitor in real time the assembly quality by analysing collected data on the torque, angle, and speed of the screwdriver from an assembly station. Kumar et al. [37] and Louw and Droomer [55] discussed applications of online quality control by employing a machine vision system. The developed TensorFlow model training algorithm can effectively distinguish defective products.

From the energy management perspective, Wang, Zhang, and Jia [151] proposed a method to monitor energy consumption in real time in an MR context. Linear regression and neural network algorithms were used to predict and optimise the strategy of energy management. Other than the manufacturing industry, an innovative energy management method was discussed in a learning factory for the building industry [152]. The intention was achieved by collecting and analysing the real-time data of carbon dioxide levels in the HVAC systems of buildings.

Predictive maintenance is a key feature in smart manufacturing, which has not been widely achieved yet. Daniyan et al. [111] and Daniyan et al. [153] demonstrated the performance of using artificial neural networks and Levenberg–Marquardt algorithms in a data-driven analysis method to enable predictive maintenance. Similarly, Centea et al. [154] introduced the application of using a machine learning method with a neural network algorithm to detect asset faults.

Big data analytics can also enhance the performance of human–robot collaboration. Aqlan et al. [148] integrated an AI-powered analytic method into the immersive virtual environment of their learning factory to enhance the eye-tracking efficiency of the VR system. Centea et al. [154] also discussed the use of neural network methods to enable cobots to conduct motion-related arm movement.

There was only one research paper examining the implementation of big data analytics in supply chain management. Darun et al. [44] discussed a proposed strategy to enhance the resilience of the supply chain by employing blockchain and data analytics.

Lastly, the process of digitisation and data visualisation also requires effective and efficient data collection and process strategies [126]. Savchenko et al. [132] elaborated on how data was processed in the learning factory for building the digital twin. Fink et al. [79] discussed the application of visualising collected data for enabling dynamic value stream optimisation in their CPPS. Singh, Centea, and Elbestawi [42] discussed data management strategies to enable real-time energy monitoring in their SEPT learning factory.

5.7. Industrial Internet-of-Things

IIoT, as the application of IoT in the industrial sector, is the foundation of the ongoing digital transformation in the manufacturing industry [155,156]. In the Industry 4 paradigm, all components of hardware and software are expected to be connected to achieve seamless communication in a smart manufacturing environment [21].

Analysing the technical aspects, there are two considerations while building an IIoT system in the development of Industry 4 facilities, including data collection and data transfer [41]. Data collection relies on the employment of smart sensors embedded in manufacturing systems [111]. Data transfer requires the establishment of industrial communication with integrated data management strategies [78].

RFID is well-known as the first IIoT device, which is designed for capturing data at the factory floor layer [39,40,42,155]. RFID technology has been well-developed to fit into ERP and MESs to make factories smarter [157]. Also, RFID systems should be able to withstand the tough conditions in manufacturing environments, such as high temperature and humidity, and exposure to chemicals [59].

With the advantages of low costs and simple configuration, RFID tags have been used extensively in the intra-logistic system of manufacturing settings to track production progress [107,158]. Captured data can then be sent to relevant systems, such as ERP, MESs, Product Data Management (PDM), and Computer-Aided Process Planning (CAPP), via cabled or wireless networks [40,157]. However, RFID has limitations in metal- and liquid-enclosed environments, such as machine tooling systems [105]. Also, the performance of RFID devices can vary, depending on the brands and models [157]. Thus, special considerations need to be made when building the system, including radio frequency deconfliction, read/write time, memory type, capacity [157,158], etc.

In terms of building industrial communication, IIoT systems support the majority of common industrial communication protocols such as OPC UA [88,107] and MQTT [42,99,132]. The Advanced Message Queuing Protocol (AMQP) and Open DDS protocol are also options [42]. OPC UA is the most promising industrial communication protocol to build the interoperability in smart manufacturing systems [97,98]. The OPC UA protocol is designed to support multiple operating systems and hardware platforms. Also, it is suitable for both simple and complex systems while supporting a broad range of devices.

In addition, 5G technology has a massive capability to boost the development of IIoT in smart manufacturing environments [63]. It enables high security and low latency in massive data transmission, which lays the foundation for the edge cloud computing services [63,99]. The employment of 5G technology can also enhance the wireless connectivity of the systems and has shown its advantages in large data transmission situations [150,159,160].

Moreover, Mukku, Lang, and Reggelin [134] discussed the potential of using Light Fidelity (LiFi) to replace Wi-Fi on the factory floor. The advantages are clear, such as higher frequency and bandwidth, enhanced data security, less interference, etc.

Despite the advantages of the aforementioned technologies, the sockets of IEEE 488 and RS 232 are still often used for field devices to establish the interconnection in current industrial sites [161]. Connection via USB and RJ45 ports is much more commonly established for simple projects in the development of academic Industry 4 facilities [105]. Moreover, QR codes and bar codes still have critical positions in the systems of Industry 4 learning factories [39,42,56].

5.8. Machine Vision

The adoption of MV systems has significantly increased in recent years for enabling smart manufacturing strategies [37,56]. MV systems are a key element of CPPSs in a smart factory [37,55]. With the integration with MESs, there is likely to be much more use of MV in the future [140].

In smart manufacturing contexts, machine vision is a key technical method for facilitating the productivity and quality management on the shop floor by eliminating human bias [108,160]. In the reviewed articles, the main applications of machine vision systems include achieving high automation [37,55], parts handling and actuation assistance [56,140], conducting quality control [11,37,55,56,98], providing safety precautions [37], positioning parts [56,60], and enabling real-time monitoring [37]. The majority of the applications require the use of AI algorithms to achieve these intentions [63].

Despite the advantages presented above, the implementation of machine vision systems in smart manufacturing systems can suffer from high costs [56]. Thus, Kumar et al. [37] and Louw and Droomer [55] discussed the employment of PiCamera to build low-cost machine vision systems in their learning environment.

6. Challenges

It has been discussed that the development of Industry 4 facilities in academia has great significance and potential for both academic and industrial participants. However, the work of developing Industry 4 facilities faces major challenges and limitations, which have been widely discussed in the reviewed articles. Those highlighted challenges can be grouped into four classes, detailed below.

6.1. Lack of Awareness

Although it has been a decade since the concept of Industry 4 was proposed, the paradigm shift to the Industry 4 era is still at an early stage. Industry 4 maturity is still being discussed as a possibility across academic fields [23]. In some circumstances, the persistence of traditional values can act as a barrier to building Industry 4 learning factories [162].

Building Industry 4 facilities in academia is a typical engineering project [36]. Thus, it is crucial to very clearly define the project scope of the Industry 4 facility in the academic context [9,39]. However, the scope and objectives are often unclear at the senior management level [81,161]. What explicit elements of Industry 4 systems need to be included in the development can be a difficult decision. Some projects tend to add as many technologies as possible at the early stages of the development, which makes the learning factories prone to becoming extremely complex [58]. Unclear project definitions tend to lead to never-finished construction and integration, as the required work exceeds any reasonable workloads and budgets.

The development of Industry 4 facilities relies on the use and integration of various hardware and software components, as the current digital transition towards Industry 4 is also a technology push process [163]. A proper selection of the required technologies, equipment, or devices is subject to the knowledge of the technologies and the awareness of relevant opportunities and risks. For example, the choice of the tags for RFID technology needs to consider the working environment, as some tags can be heavily and irreversibly affected by metal or water barriers [59]. Similarly, Mladineo et al. [40] have argued that RFID systems can vary in operating frequency, radio range, and data format. RFID usage may be constrained by interference issues, tag memory limitations, and integration challenges within broader system architectures [59].

6.2. Technical Difficulties

In manufacturing, the core philosophy of the ongoing Industry 4 paradigm shift is to enable smarter decision-making strategies and more efficient production processes by analysing the collected data through the digitised manufacturing processes [21]. However, building adequate interconnections and industrial communications to serve the expected functional development is a challenge [134]. It is challenging to properly manage the vast amount of data through the interconnected components of manufacturing systems. Henning et al. [50] discussed the difficulties of managing the data collection and data transfer in building the interoperability in their learning factory. Similarly, Daniyan et al. [111] outlined that the challenge of enabling predictive maintenance is the real-time capacity of the system in handling the massive amount of data. It is also challenging to use the acquired data for different sub-systems within a manufacturing system [105].

Despite various advantages of working in a digitised context, the workload to establish the expected digitisation in Industry 4 systems is huge [87]. Umeda et al. [98] pointed out that 3D modelling can include large workloads to build a digital twin system for a manufacturing environment. Also, Ralph, Schwarz, and Stockinger [36] observed that virtualising and simulating metal forming processes are extremely challenging as they require finite element analysis methods integrated to define the parameters of the microstructure of the material. It is also a challenge to seamlessly integrate the existing physical systems into the virtual context [47,71,114]. Achieving the integration can be tricky as devices in the existing systems may have different interfaces, system parameters, and performance [28,47].

Software agents are playing a significant role in the ongoing digitisation processes in manufacturing. However, it can be difficult to find a suitable software package to build the digital world of the learning facilities, as there are many options in the market [98]. Open-source software is the preferred option for practitioners as it has great potential in building scalability and interoperability [129]. However, there are still certain limitations that need to be considered, including accessibility, license availability, user friendliness, maintenance costs, documentation, help system, upgrades, etc.

Many emerging technologies pose usability challenges, particularly for new users or educational settings [51,129]. For example, participants took significant time to familiarise themselves with using the devices of the developed AR assistance system in their first try in the development of the learning context, which results in a worse outcome in comparison to the traditional operating method [14]. Similarly, Aljinovic et al. [9] emphasised that advanced robotics is the main manipulator in Industry 4 facilities. Thus, more efficient and dynamic ways to program the robots are greatly desired.

The immaturity of the latest technologies is an inherent limitation in the implementation in academic training facilities [162]. This is particularly true for those technologies that need to directly interact with human operators. For example, a wearable device for AR technology implementation may cause visual discomfort after practising for a long time [2]. Also, health concerns have been raised for human operators who are exposed long-term to high-intensity RFID radiation [40]. All those limitations and concerns may cause distraction for the participants, which in turn will affect the performance of the designed activities.

Technologies are quickly evolving to serve the demands of the digital transformation. New competitive equipment and technologies can become available every 2 to 3 years [52]. However, industrial standards development is far behind the requirements of new technology implementation. Process integration should follow standardised practices to ensure system interoperability and maintainability [23]. For example, Eder et al. [14] pointed out that there is no reference standard that can be used for integrating AR technologies in smart manufacturing contexts. Also, standardising the sockets is another common basic problem in developing Industry 4 facilities in academic facilities [58]. This is because the equipment and devices might be purchased from different suppliers at different stages of the project.

Cybersecurity is a significant concern in the digital era. Although many cutting-edge technologies have been applied in industry, some of them are still immature. The data security in those deployed technologies may not be transparent enough for the users [41,51]. Mourtzis, Angelopoulos, and Kimitrakopoulos [11] pointed out that 80% of the data in current manufacturing environments is unstructured, which leaves significant safety concerns for the participants. The uncertainty at the cybersecurity level can also be a barrier for the stakeholders to move forward in the digital transformation [23].

6.3. Process Complexity

It is not uncommon to find that many academic learning facilities were built before the Industry 4 concept was defined, resulting in demands on transitioning existing systems into Industry 4-oriented frameworks [7,41,45,88]. Pedagogical framework design must be in line with Industry 4 principles to ensure that students are able to acquire adequate skills and competences [45]. Teaching and training activities need to be carefully designed to cover the complex Industry 4 framework [32]. Also, the staff in existing facilities need to acquire adequate competence to push the transition [45]. Moreover, advanced IT infrastructure and cutting-edge technologies are needed to showcase the desired Industry 4 features [14]. Thus, Marmier et al. [104] believe that it is hard to develop a suitable project-based learning context by redesigning the current training and research facilities to demonstrate Industry 4 characteristics.

Manufacturing processes can be highly complex, which makes it difficult to establish a pilot-scale learning context in academia that captures this aspect [32,38,87]. The well-established Industry 4 learning facilities can only focus on limited research topics in the Industry 4 context [25,103], such as flexible assembly [40,56,93], quality management [56], lean production and energy efficiency [10], intralogistics [107,158], metal forming [36], etc.

Flexible assembly received the most contributions in the reviewed literature in such a hybrid educational and research environment [2,3,4,26,56,88,98,123]. However, it is challenging in virtual training environments to create high-quality dimensional product replicas, which in turn causes limitations on collecting data from the product and simulating collaboration [31,164]. As a result, participants must simplify the process complexity as a trade-off solution [26,71]. Pick-and-place movement by robotic arms is the easiest manufacturing process to be digitised and simulated for discussing the flexibility of assembly processes [3,13,24].

6.4. Resource Limitations

Resource limitation is also a major constraint in the development of Industry 4 learning facilities in academia. The sources of the constraints include institutional policies, human resources availability, academic support, and industrial partnership.

Lack of expertise is a common challenge in organisations that are building Industry 4 facilities [52,68,140]. This is because wide multidisciplinary knowledge is needed for the team to establish a representative Industry 4 facility [68,107].

The development of Industry 4 facilities requires the employment of cutting-edge technologies and the construction of advanced infrastructure. The different selection of technologies causes huge differences in budget requirements [28]. Some of the required technologies and equipment could be extremely expensive, or even inaccessible to a research team [18,83]. However, financial limitations are always a constraint for engineering projects [68,140,152,164]. Thus, practitioners must prioritise the budget consideration before the performance consideration while choosing the equipment and technologies for building the facility [126]. Thus, low-cost, robust, and cost-effective solutions are desired for most Industry 4 facility development projects [37,55].

Safety concerns are another major consideration while building Industry 4 facilities in academia. When compared with a real manufacturing site, learning factories or Testlabs in universities often have a smaller physical room to build them [139]. However, the management team will want to employ more equipment and establish more processes to increase the efficiency and utilisation of space. Centea et al. [51] have expressed their concerns about the hazard exposure from 3D printers that are deployed in their learning factory. It is also the reason why collaborative robots are the first choice to be used in the laboratory environment [11].

Building a partnership with local industries is crucial for the success of the developed Industry 4 learning factories [140,165]. Centea et al. [51] underlined the hassle in recycling the leftover materials from the additive manufacturing training processes. Also, Kumar et al. [37] point out that training for AI models consumes huge amounts of resources, including time, energy, and money. The collaboration with industry participants has great potential for solving these problems, which in turn facilitates the development of Industry 4 learning facilities.

7. Discussion

7.1. Implementation Gaps

The reviewed literature reveals that academic institutions worldwide are pushing the transition towards the new paradigm by establishing dedicated Industry 4 learning factories and Testlabs. Despite the growing interest in academic contexts globally, there remain critical implementation gaps that limit their educational, research, and industrial impacts. As our descriptive analysis demonstrates that institutions in Europe contribute 60.8% of the reviewed publications, followed by Asia (12.3%), North America (10.8%), Africa (6.9%), Latin America (5.4%), and Oceania (3.8%), which made the least contribution in this regard. Figure 2 illustrates the contribution statistics from different regions.

The gaps fall into a wide range of fundamental aspects, including technological, pedagogical, and policy-level constraints.

Lack of infrastructural intelligence: A deeper analysis suggests that significant parts of the effort remain focused on establishing physical infrastructure, such as intelligent robots, IIoT platforms, and 3D printers, rather than achieving more advanced features, such as decentralised decision making, predictive intelligence, and bi-directional system integration. It is understandable that it is easy for academic participants to simply employ visible and mature technologies to begin the development projects from scratch. But this also highlights a bottleneck that must be addressed to make academic Industry 4 facilities real catalysts for the ongoing industrial revolution, not just a demonstration.

Lack of system integration: Another common implementation shortcoming in the reviewed literature is the disjoints between sub-systems. In many reviewed cases, collaborative robots are manually programmed with minimal data-driven feedback loops. The development of system middleware and standardised architectures is far behind the needs for building the expected scalability and interoperability of the facilities. The majority of the developments are still at the sub-system level, which shows a promising research trend to develop holistic Industry 4 systems in academic contexts.

Lack of supportive analytics for educational purposes: Although one of the primary goals of Industry 4 facilities in academia is to enhance educational and training activities, very few implementations have focused on the outcomes of supporting student learning activities. How to use real-time data streams, adaptive feedback systems, and DT-guided decision making to facilitate teaching activities in universities is underexplored.

Lack of long-term engagement with local industries: While the establishment of Industry 4 facilities is often started and pushed forward by collaborating with local industries, the actual depth and breadth of the engagement vary. It is not rare to see that some partnerships are limited to one-off projects, rather than long-term collaboration in curriculum development, research design, and workforce training. This may need more commitment from the high-level management of the universities or even the local government.

Lack of strategic perspective: There is only a small number of reviewed articles that have shown their long-term strategic roadmap in the facility development, such as the STEP Learning Factory [35,41,43,51,59,68,154,166,167]. The majority of the cases are developed with a short-term vision, such as a PhD degree programme or a pilot project with industries, which limits the maturity of these facilities. Staged development plans are desired for academic participants to align their efforts in line with the institutional or national digital transformation agenda.

7.2. Bridging Design Principles and Key Technologies

As discussed in Section 6, there are eight key Industry 4-related technologies that have been summarised in the reviewed articles. The technologies can be grouped into four categories. Each technology group can be considered as a main contributor to implementing one of the four design principles that have been discussed in Section 5. A synopsis of mapping the specific technologies to the Industry 4 design principles is summarised in

Table 2. Industry 4 design principles and the related technologies.

Industry 4 Design Principles	Industry 4 Technologies	Intentions
Interconnection	CPPS	Establishing the integrations and interoperability by interconnecting hardware and software components.
	IIoT	Capturing raw production data in real time from smart sensors across the interconnected networks.
Information transparency	DT	Visualising physical assets in a digitised context. Simulating production workflows and equipment behaviour.
	MR	Visualising physical assets in a virtualised context. Providing valuable information in manufacturing processes for human operators.
Decentralised decisions	Big data analytics	Providing optimal decisions in production processes.
Technical assistance	Advanced robotics	Advanced manipulators to replace human operators. Fulfilling autonomous intralogistics.
	AM	Fabricating products with an extremely complex structure in one go. Achieving a prompt prototype in the product development process.
	MV	Monitoring production processes in real time.

.

7.3. Technology Maturity and Deployment Difficulty

A further qualitative analysis was conducted to evaluate the eight key technologies based on the deployment maturity and technical difficulty. The maturity level was determined by assessing the depth, breadth, and frequency with which a given technology was discussed in the reviewed literature. High maturity technologies were implemented in multiple studies across diverse regions. In contrast, low maturity technologies were only presented with conceptual or experimental engagement. In regard to deployment difficulty, integration complexity, technical skill requirements, and costs and infrastructure needs are considered as the key indicators. A matrix of the outcome has been developed in

Table 3. Industry 4 technology maturity and deployment difficulty.

Technology	Maturity Level	Deployment Difficulty	Remarks
Cyber–physical systems (CPS)	High	High	Critical integration demands; software–hardware complexity.
Industrial IoT (IIoT)	High	Medium	Wide use, but requires sensor interfacing and network configuration.
Digital twin (DT)	Medium	Medium-High	Bi-directional integration remains immature.
Mixed reality (AR/VR)	Medium-Low	Low	Strong training usage, but the costs of employment can be challenging.
Big data analytics	Low	Medium-Low	Toolsets are maturing, but the talent gap persists.
Advanced robotics	Medium-High	Medium	Widely employed, especially the cobots.
Additive manufacturing	Low	Medium-Low	Ideal for prototyping and student projects.
Machine vision	Low	Medium-Low	Requires AI model training; resource-intensive.

. An operational approach to rating the maturity level and scoring the deployment difficulty is detailed in Appendix B.

8. Strategic Roadmap and Future Trends for Industry 4 Facilities Development

8.1. Strategic Roadmap for Academic Implementation

To further support academic institutes, we propose a five-phase roadmap to outline a staged plan for developing and deploying Industry 4 facilities. The first stage is called Foundation and vision, which focuses on creating awareness and developing a strategic vision for the expected Industry 4 development. At this stage, awareness workshops and stakeholder consultations should be considered to underpin the long-term goals and short-term objectives of the development and identify the timeline and key milestones that align internal and external expectations. For example, a clear scope of building the learning factory at Lernfabrick Fallenbrunnen was outlined before its development [69]. Similarly, a framework design in detail is well discussed for building the Proto Learning Factory [64].

The next stage is called Pilot deployment. This stage aims to pilot low-cost infrastructure that includes key modular Industry 4 components. In this phase, some low-cost technologies can be commonly considered, such as an RFID system, Raspberry Pi-based IIoT tool kits, and OPC UA connectivity. In the reviewed articles, the majority of the studies are at this stage of demonstrating solo or limited technologies only. For instance, Rokoss and Schmidt [116] discussed an implementation of a digitisation strategy in their learning factory. Similarly, Sievers, Neumann, and Tracht [86] only discussed an MR strategy implementation that can be applied to their digital learning factory.

The third stage is called Curriculum integration, where the laboratory setup and academic curricula can be linked. After the Industry 4 modular components have been piloted, early student engagement can be considered in this phase to integrate real-world technologies into coursework, which in turn will build a structured educational and research environment. Marmier et al. [104] briefly discussed how the framework of their Industry 4 learning factory matches the pedagogy at the University of Strasbourg. However, in the reviewed articles, there is a very limited number of studies that have considered curriculum integration in their Industry 4 learning facilities.

The fourth stage is called Advanced capabilities. The purpose of this stage is to elevate the facilities into a fully digitised and semi-autonomous environment to support high-level research and real industrial practices. In the reviewed articles, the SEPT learning factory in Canada has successfully demonstrated multiple stages of implementation of advanced technologies, achieving a live training and research environment [35,41,43,51,59,68,154,166,167]. However, the development at this stage requires huge funding support, which is a main challenge for both academic and industrial participation [164].

The final stage is called Regional collaboration and scaling, which aims to enlarge the scope of collaboration with other institutions and industrial partners to build an open-access platform. The intended outcome in the last stage is to build a cooperative and scalable academic ecosystem with true Industry 4 competencies, which benefits not only the institution but also all potential academic and industrial partnerships from the shared infrastructure. Although no practices have been implemented, the participants at the Proto Learning Factory have taken into consideration integrating the learning factory into real-life industries to maximise the benefits [64]. The proposed five-stage strategy and key milestones in each stage are demonstrated in Figure 3.

There are some key characteristics and challenges that can be identified and expected in the implementation processes. As summarised in Figure 4, the first key characteristic and challenge are technology engagement. It requires relatively low demands at stages 1, 3, and 5. But the requirements sharply increase at stages 2 and 4. In terms of cost and investment level, it shows a steady increase in the first four stages but a slight drop at the last stage. This is because the near-mature collaborative platform could bring back benefits. The other key feature is the institutional complexity of the ecosystem, which has a steady growth all the way up. Similarly, industrial engagement also has a smooth curve from beginning to end. The last identified characteristic is the risk and failure exposure. There are two peaks that can be seen at stage 2 and stage 4, where technologies are heavily employed, followed by a slight drop once the technologies and equipment are successfully deployed. The relative-scale values in Figure 4 were obtained via the qualitative synthesis by reviewing the 145 articles.

8.2. Future Trend of the Industry 4 Facility Development

In terms of the future trend, the development of Industry 4-oriented facilities in academia is expected to focus on building more modular, scalable, and hybrid platforms, which improve the flexibility of the learning environment and reduce the risk and failure exposure during the implementation. Also, cloud-based infrastructure and AI-driven algorithms will be broadly employed to lower the cost and investment level while maintaining the system performance and productivity. Moreover, competency-based frameworks will be heavily adopted into the ecosystem to support the cultivation of industry-ready graduates, which is one of the fundamental goals of the Industry 4 facility development in academia.

9. Research Limitations

Although this study attempts to make contributions to the field of the development of Industry 4 facilities in academia by summarising the significance, outlining the design principles, discussing the challenges, and proposing a roadmap for the development, there are certain limitations that must be acknowledged to ensure the research transparency, contextual clarity, and future research.

Firstly, while the process of searching and screening articles was instructed by a PRISMA guideline, the process did not include an inter-rater reliability assessment for scoring and selecting the articles for the review. Although the article selection was performed with well-considered criteria, the lack of a reliability check process may lead to subjective judgement on the technological classification.

Additionally, the search process was limited to only three main databases, namely SpringerLink, Scopus, and Web of Science, and the publication language was limited to English only. While we are confident that the scope of the research has covered a majority of the current research outcomes of the topic in academic fields, findings from non-English sources and regional databases could be, unintentionally, excluded.

Finally, due to an inherent constraint that the research topic emerged only in the 2010s, there are relatively few peer-reviewed studies that have been published. As a result, many of the reviewed articles were from specific publication sources, such as the Conference of Learning Factory and Procedia Manufacturing. Although the articles from those sources have a well-matched framework and content within the research scope, this could still raise a concern about bibliographic balance. To address this, three summary tables are developed in Appendix C to clarify the bibliographic distribution against key information, which provides transparency with respect to the diversity of the literature sources.

Table A5. Bibliographic distribution against publication type and the influence metrics.

Publication Type	Percentage
Peer-reviewed journal	Impact factor > 2:	6.21%
	Other:	48.96%
Peer-reviewed conference proceedings	44.14%
Book/book chapter	0.69%
Total	100%

summarises the distribution against publication types and the metric.

Table A6. Bibliographic distribution against country-wise situations.

Publication Country	Percentage
Germany	33%
Austria	9%
Canada	8%
Other country	50%
Total	100%

presents the distribution along the country-wise situations.

Table A7. Bibliographic distribution against primary publication sources.

Publication Source	Percentage
Procedia Manufacturing	35.2%
Conference on Learning Factories	26.9%
Other sources	37.9%
Total	100%

outlines the distribution between primary publication sources used and others.

10. Conclusions

This paper provided a critical insight into, and a roadmap for the development of, Industry 4 learning and research facilities in academia. The research outcomes are expected to help current and future participants by increasing their awareness of the significance of the development, clarifying the research scope and objectives’ uncertainties, and preparing them to deal with the real scale of complexity and skills issues.

An Industry 4 design principles pyramid was provided in Section 4.3, which outlined the true intentions of the ongoing digital transformation in the manufacturing industry. In the reviewed articles, the establishment of interconnection in Industry 4 systems has been widely discussed. This includes the development of CPPSs and IIoT. However, only a few have demonstrated aspects relating to information transparency. And even fewer papers discussed decentralised decision-making strategies, which are the top tier of the pyramid. Thus, it is clear that the current development of Industry 4 facilities is still at the early stages, even in industry.

There are still many true Industry 4 characteristics that have not been well addressed in the literature. These include bi-directional data transmission in digital twin systems, high adaptiveness and resilience of manufacturing systems, fully automated batch-size-of-one manufacturing strategies, effective predictive maintenance application, and so on.

Regarding future work, the team of authors is working towards the development of an Industry 4 laboratory at the University of South Australia. The development is intended to close the gaps that have been identified above to achieve an improved Industry 4 learning and research context.