The entire value chain is linked across applications as part of the digital transformation. Internet of Things (IoT) technologies are being used to combine traditional operational technology (OT) with modern information and communication technology (ICT), provided that this results in economic or social benefits. A driving motivation behind this convergence is the desire to be able to operate an infrastructure based on manufacturer-independent standards, which would have the advantage of increased flexibility.

Digital networking

Digital networking in industrial applications is relevant in this context. Fieldbus systems and their Ethernet-based successors have played a key role in production for around 40 years. However, we are currently on the verge of a breakthrough with a series of manufacturer-independent standards that are being adopted in the Time-Sensitive Networking working group of the Institute of Electrical and Electronics Engineers (IEEE). Building on this, machines will communicate in future using Open Platform Communications Unified Architecture (OPC UA), which has been standardized in the 62541 series of standards of the International Electrotechnical Commission (IEC). The foundation for the introduction of the fifth generation of mobile communications (5G) into the factory has also already been laid, as users will be able to purchase local frequencies from this year.

However, the areas of connectivity (TSN) and communication (OPC UA) only cover part of the classic automation pyramid, which is known to extend from input/output (I/O), programmable logic controllers (PLCs), supervisory control and data acquisition (SCADA) systems and manufacturing execution systems (MES) through to enterprise resource planning (ERP). In line with the Industry 4.0 vision, this pyramid is to be replaced in the long term by the use of autonomous cyber-physical systems (CPS). Intelligent, autonomous units interact directly with each other to enable production to be as flexible as possible.

Cloud computing

In the meantime, however, the cloud computing paradigm is finding its way into production. Devices, data and services are networked via central servers on the internet in order to optimize processes and gain insights from data. When connecting machines and their data to the cloud and possibly also machine control from the cloud, a number of questions arise. These can be roughly divided into two areas: The processing of production data on external servers in times of the General Data Protection Regulation (GDPR) and global cyberattacks, and the fulfillment of network requirements, especially with regard to deterministic communication.

To answer these questions, it is helpful to take a look at the past: for around 70 years, we have been able to recognize an alternating trend in distributed computer systems. While the first mainframe computers could already be used centrally in the 1950s/60s, the trend changed in the 1980s/90s in favor of distributed client-server systems. Since the beginning of the new millennium, data and services have once again been stored and hosted centrally under the term cloud computing. However, as outlined above, not all data can and should be processed outside the company's own administrative domain and outside defined geopolitical boundaries. The reasons for processing data locally or centrally are manifold and are known as 'the Vs of Big Data':

• Volume - challenge of unfiltered centralized storage.
• Velocity - Challenge: communication of all data via a network.
• Variety - challenge: central interpretation of heterogeneous data.
• Vulnerability - challenge: storage of data outside the administrative domain.
• Datavolatility - challenge: heterogeneous persistence requirements for data.
• Data validity - challenge: heterogeneous relevance of data.

Edge computing

The continuum of different edge computing terms.

© Fraunhofer FOKUS

This means that there are situations in which data cannot or should not be transferred to a central cloud instance. The edge computing paradigm refers to a distributed cloud computing approach that enables data to be pre-processed locally. The entire spectrum of possible distributions can be referred to as fog computing. However, the terms themselves are not uniformly defined in the literature and are sometimes used differently depending on the application domain.

There are a number of fields of application for the edge computing paradigm in automation. Primarily, existing OT infrastructures can either be supplemented or partially replaced. In the first case, edge nodes can be used, for example, to dynamically install and execute applications at the data source. In particular, approaches from the field of artificial intelligence (AI) can be used here. An obvious example would be an augmented reality (AR)-based human-machine interface (HMI), which requires very short communication latencies and places high demands on object recognition. Analogous to the OT/ICT convergence of connectivity and communication technologies and following the trend of 'softwareization', partial aspects can also be replaced by ICT approaches. For example, the boundaries between physical PLCs and traditional industrial PCs (IPCs) could become even more blurred through the use of edge-based, virtualized PLCs. However, open questions, such as guaranteed compliance with deadline-driven, hard real-time requirements on multi-purpose hardware using virtualization environments, are challenges that have not yet been conclusively resolved.

The standardization

A necessary prerequisite for these developments: The use of open, independent standards. This is the only way to break the dependency on individual manufacturers. The advantage is that software-based infrastructures will be possible in the future, which can be dynamically adapted to current requirements using management systems. However, various technologies for connectivity, communication, virtualization, orchestration and data description must be taken into account. The Multi-Access Edge Computing (MEC) standard, which comes from mobile communications and is specified by the European Telecommunications Standards Institute (ETSI), is often mentioned here. However, this does not go far enough and does not meet the needs and requirements of automation technology in particular.

A reference architecture

Preliminary reference architecture model Edge Computing 4.0 (RAMEC4.0).

© Fraunhofer FOKUS

The complexity of the whole process can be seen in the preliminary version of the Edge Computing 4.0 reference architecture model (RAMEC 4.0). This was developed based on the Industry 4.0 reference architecture model (RAMI 4.0). Requirements such as security, real-time capability, acceleration (e.g. for the efficient use of neural networks) and management must be addressed differently at different levels, which can have different characteristics depending on the topological position. This results in the need for different edge solutions consisting of a combination of domain-specific standards. With their limited resources, it is almost impossible for small and medium-sized enterprises (SMEs) in particular to make a suitable selection and combine the selected solutions. Although there are already some proprietary solutions from various providers, current developments are somewhat reminiscent of the early days of the 'fieldbus war', which will soon be over.

This is where the industry-driven Edge Computing Consortium Europe (ECCE) comes in, which is to be founded later this year as a non-profit European organization. The aim is to reduce research and development times as well as manufacturer dependencies and to develop an open edge computing ecosystem that also takes into account European characteristics in particular, such as the GDPR or specific standards. In order to achieve this goal, edge nodes are to be developed that are based on open standards and consist of coordinated, open software components. The primary aim is not to design new standards or drive forward implementations, but rather to make practical recommendations on the following points:

• Domain reference: Description of specific requirements by industry partners of focused domains (such as discrete manufacturing);
• Reference model: Delimitation of the specific problem space (for example: Gateway-Edge in RAMEC4.0);
• Standards: Recommendation of and contributions to relevant standards (IEEE TSN, IEC 62541 or IEC 23360-x:2006);
• Initiatives: Exchange and coordination with consortia (Plattform Industrie 4.0, Industrial Internet Consortium - IIC, OPC Foundation);
• Implementation: Combination of and contributions to open source implementations (EdgeXFoundry, Akraino or Kubernetes);
• Blueprints: evaluation of specific edge nodes in realistic environments and feedback of the results.

In November 2018, more than a dozen partners welcomed the founding of the consortium at the Edge Computing Forum. Other companies are invited to play an active role in the development process.

However, the application of the edge computing paradigm does not end in production. Rather, the OT/ICT convergence described above can be observed in all areas of the Industrial Internet of Things (IIoT), which is not limited exclusively to connectivity. The networking of the entire value chain within and across application domains requires a large number of standardized solutions. One of the most important key technologies here will be distributed cloud computing approaches - regardless of the specific terminology used.

Author:
Dr. Alexander Willner is Head of the Industrial IoT Center at the Fraunhofer Institute for Open Communication Systems (FOKUS).