1 Foundational Technology Layers

Interoperability in the SoS domain is a rising problem for cost-effective engineering and operation of systems of systems (see Figure F.23).

There is currently no industrial solution to this problem. Academia and industry are experimenting with approaches based on, for example, ontologies80, machine learning81 and open semantic frameworks82. Even if no clear winning approach can be identified based on current research results, growing interest can be noted for e.g ontology, data and model driven approaches. Automating considerable parts of interoperability engineering (design-time and run-time) will improve SoS operational quality.

To enhance EU leadership and sovereignty in the field of SoS, autonomous information translation for understanding is a necessity. Some integration platforms already focus on protocol and information interoperability (Derhamy, 201883). To enable the cost- and time-efficient engineering of solution integration and extension, their updates and upgrades over the lifecycle is crucial. Therefore, integration platforms have to provide mechanisms for dynamic and instant information translation across the ontologies and semantics used the individual constituent systems of the SoS.

To facilitate substantial cost reductions for SoS solutions, autonomous and dynamic mechanisms for information translation are required. Such mechanisms should cover:

• Translation between standardised data models (e.g. ISO 1030384, ISO 1592685, BIM86).
• Translation between different implementations of standardised data models.
• Automated data model translation.
• Autonomous data model translation.
• Efficient and flexible engineering procedures.
• Engineering tools that support the complete engineering process in both design- and run-time.
• Support for key automation requirements.
• Automated engineering.

Evolvability and composability is a multi-dimensional key aspect of SoS evolution, one that affects their architectures, properties, functionalities and behaviours from different perspectives (evolvability, trust, interoperability, scalability, availability, resilience to failures, etc.). Primarily, composability must ensure the persistence of the five major attributes that characterise a SoS (see Maier, 199887). Vertical (hierarchical) composability provides the most common way to build a SoS that is typically structured in a hierarchical stack composed of adjacent layers. Vertical composability has to deal with the different abstraction levels of the stack layers, adopting aggregation and de-aggregation solutions as references to compose the constituent systems of the SoS. Architectural composability, on the other hand, is fundamental for SoS design, specifically when critical requirements such as trust or safety must be satisfied (see Neumann 200488, for an extensive report on trustworthy composable architectures).

In the hierarchical structure of a SoS, the constituent systems that are at the same level typically compose horizontally (in parallel or serially), potentially generating competing chains of constituent systems. Serial composability represents a critical issue for all properties that are not automatically transitive, such as trust. Indeed, the inclusion of AI in embedded and cyber-physical systems increases the required level of trust, as well as the uncertainty of the results of the composition process (see, for example, Wagner, 201589).

When the constituent systems expose high-level services, service composability allows for the creation and provision of new added-value services at the SoS level, combining the resources, functionalities, information, etc., of the constituent systems. Eventually, the engineering process deals with composability, enabling it by design (already present from the constituent systems level) and/or managing it during the operations of the SoS, to address the dynamic nature of SoS in time (run-time composability associated with evolutionary development and potential emergent properties, behaviours, and functionalities).

The dynamic nature of SoS is based on the composition and integration of embedded and cyber-physical systems. The role of composability is to ensure that functional and extra-functional properties (scalability, quality of service (QoS), performance, reliability, flexibility, etc.), and the functionalities and behaviours of the constituent systems are preserved in the SoS or combine in a predictable and controlled way, even when the constituent systems recombine dynamically at run time. The lack of solutions to dynamically manage composability represents one of the limitations preventing the diffusion of SoS.

Composability should be conceived as a quality of SoS that makes them future proof: (i) the relationships between components that allow them to recombine and assemble in different and potentially unlimited architectural combinations, and ensure and exploit the re-use of components; (ii) the extension of components lifetime within the evolution of the SoS during its lifecycle; (iii) the possibility that SoS will easily evolve, adapting to new contexts, new requirements and new objectives; and (iv) the simple substitution of faulty, inadequate and/or new components with a minimal impact for the SoS, guaranteeing the survival and sustainable evolution of the SoS. Composability also have to consider cross sectorial requirement like e.g. security, safety, trust, evolution.

Ensuring composability at the SoS level represents a very challenging goal, potentially generating serious and critical consequences, and even preventing the integration of the SoS. Indeed, considering a property that characterises a constituent system with a certain attribute, it is not guaranteed that the same property will characterise it when the constituent system becomes part of a SoS. In addition, if the property is still present, it is not guaranteed that it will have the same attribute. The same applies to the constituent system’s functionalities, behaviours, etc.

As a consequence, one major effect of the composition, integration, evolution of the constituent systems is the evolution of the SoS, with emergent properties, functionalities and behaviours which generate uncertainty. E.g. when SoS evolution affects e.g. security, safety, trust, interoperability, scalability, availability, resilience to failures, etc., the impact of the uncertainty could potentially be extremely serious.

The inclusion of AI in SoS increases the importance of composability, because it may significantly increase the complexity, variability and fuzziness of composability results. AI enables a completely new category of applications for SoS. Therefore, the availability of specific solutions for the validation, verification and certification of SoS composed of AI-based systems is a critical requirement.

Predicting and controlling the effects of composability is also fundamental for the interaction of humans along the SoS lifecycle and the protection of human life should be ensured in SoS evolution. Uncontrolled and unmonitored composition could lead to deviations from expected behaviours or generate unknown emergent behaviours potentially dangerous for humans. The increasing level of automation introduced by SoS accentuates this criticality, and will require that humans still intervene in cases of emergency (for example, in automated driving).

The solutions proposed to manage composability will also have to support the multi-domain nature of SoS, the presence of different stakeholders in its lifecycle, and the different regulations and standards that apply to these domains. From an engineering perspective, emergent behaviours require that the development of SoS, applying composability, is evolutionary and adaptive over the SoS continuously evolving lifecycle, which potentially may never finish. In fact, SoS architectures and platforms, jointly with the proper engineering support, will have to provide solutions to control the uncertainty of evolvability and ensure adequate countermeasures.

Since the technology base, and the organisational and human needs are changing along the SoS lifecycle, SoS architecting will become an evolutionary process based on composability. This means: (i) components, structures, functions and purposes can be added; (ii) components, structures, functions and purposes can be removed; or (iii) components, structures, functions and purposes can be modified as owners of the SoS experience and use the system. In this sense, the dynamically changing environmental and operational conditions of SoS require new architectures that address the SoS goal(s), but thanks to composability will also evolve to new system architectures as the goal(s) change.

Evolution in SoS is still an open research topic requiring significant effort and the key areas of research and innovation include:

• Methods and tools for engineering evolvability of systems of embedded and cyber-physical systems.
• Evolutionary architectures in systems of embedded and cyber-physical systems.
• Evolvable solutions for trust, availability, scalability, and interoperability.
• Evolvable solutions capable for managing resulting uncertainty emergent properties, functionalities and behaviours, including resilience to failures.
• Evolvability in systems of cyber-physical systems supported by virtual engineering (e.g. digital twins).
• Methods and tools to manage emergencies in embedded and composable systems of cyber-physical systems.
• Service-based vertical and horizontal evolvability to enable high-level, and potentially cross-domain, interoperability of embedded and cyber-physical systems.

Europe is a world leader in the engineering of systems of systems. Major European companies such as Siemens, ABB, Schneider, Valmet, Bosch and Endress+Hauser, together with a number of large system integration companies (e.g. Afry, VPS and Midroc), offer complete engineered solutions, making Europe the leading global automation SoS provider.

Most solutions for embedded and cyber-physical systems engineering are based on highly experienced teams of engineers supported by a heterogeneous set of SoS engineering tools. For example, engineering practice and associated standards provide design-time solutions based on, for example, IEC 61512 (ISA 88)90, IEC 62264 (ISA95)91, IEC8134692, ISO 10303, ISO 15924, IEC 6289093. The proposed Industry 4.0 architectures, formally provided by the DIN specification 91345 RAMI 4.0, have not yet made it into industrialised engineering procedures, or associated tools and toolchains. Therefore, the industrial state of the art for SoS engineering still has its major base in legacy technology.

The current state of the art engineering of SoS remains more an art than a well-structured integration and engineering process. For example, the analysis of emergent behaviour of very large SoS is still at a foundational research level in academia.

The European leadership in application fields such as distributed automotive and industrial automation and digitalisation indicates some excellent skill sets in the art of SoS engineering. In the short to medium term, Europe has to transfer these skills into systematic and robust engineering procedures supported by integrated and efficient tools and toolchains.

This is expected to lead to engineering processes, tools and toolchains covering the whole life cycle that to significant extent can be automated while supporting integration between multiple stakeholders, multiple brand and multiple technologies. To support such integration and engineering efficiency, solution quality and sustainability concrete advancements like in (Figure F.2494 will become necessary. The advancement may include integration and engineering process capabilities like:

• Flexible integration and engineering procedures.
• Model-based engineering procedures and tool,
• Supported by interoperable and flexible toolchains.
• Integration of multi-stakeholder engineering processes.
• Automation of substantial parts of the integration and engineering process.

In support of EU leadership and sovereignty in the field of SoS engineering the ambition is to invest in a small number of integration platforms and their associated tools, toolchains and engineering processes. Strong European-based ecosystems should be created and provided with long-term governance also connected to open source. These engineering processes, methodologies, tools and toolchains shall provide, for example:

• Efficient and flexible engineering processes.
• Model based engineering.
• Engineering tools supporting the complete engineering process along the system's lifecycle.
• Support for key automation requirements.
• Automated engineering.
• SoS traceability and analytics interoperable with engineering tools and tool chains
• SoS evolution impact analysis
• Automated testing validation and verification (TV&V) along the life cycle.

In particular, SoS TV&V introduces a significant challenge, mainly due to complexity, to the effects of composition (not always known in advance) and to SoS dynamic evolution over time. For SoS, a full TV&V procedure prior to deployment is practically unrealistic. Typically, the TV&V of each constituent system is asynchronous and independent of SoS, challenging the SoS TV&V with feature and capability evolution. For this motivation, a structured framework methodology and tools is necessary to demonstrate an appropriate level of confidence that the feature under test is present in the SoS, and that no undesirable behaviours are also present. This implies a need for end-to-end system capabilities metrics and, according to the flow of data, control and functionalities across the SoS, additional test points, recurring tests and AI-empowered data collection. This analysis should be considered to address changes in the constituent systems and to receive feedback on anomaly behaviours.

In alignment with the traditional architectures in “Control and Automation”, automation, control and monitoring schemes in most SoS today are characterized by a hierarchical and centralised architecture, made up of layers of sensors and actuators, controllers, and associated computers that are distributed throughout the often complex, interconnected SoS. In terms of control, the atomic unit of a SoS can be found on the field level, where direct control of so-called agents takes place. This can be e.g. the control of a single generator in a power plant, which itself is part of a smart grid. The actions of field control are directed by higher levels such as plant supervisory control, scheduling control or plant-wide optimisation.

As complexity of SoS increases steadily, so does the number of systems involved in its control performance. As a result, when using state of the art control methodologies the communication effort would grow exponentially. Thus, balancing communication effort and control performance is a key factor. In this context and in view of limiting data traffic in a SoS, synchronisation of systems becomes a major control goal as it is directly linked to the stability of the control system.

In addition, control of complex cyber physical SoS must address other important aspects such as scalability i.e. to deal with a variable number and interconnection of systems and control loops and network phenomena (such as computation/communication latency, data loss). Looking at the aspect of data management, open SoS control platforms should ensure information security management, SoS scalability, SoS engineering efficiency and also SoS real-time performance.

For this Major Challenge we envision the following key focus areas:

• Tools for control system analysis of SoS.
• Considering humans, environment and the economy in the loop.
• Support in SoS control design.
• Reduction of communication effort, variable structure, variable number of systems in control loops.
• Control system testing, validation and verification (TV&V).

Current industrial state of the art for monitoring and management of SoS reflects back to monitoring and management of production automation, energy grid automation and similar. Looking closer we find a plethora of commercial application solutions tailored to specific applications. Many of these are very application and site specific and “home brewed”.

There is a wide set of different realms to be monitored and managed, ranging from modern production processes, smart grids, smart cities, automotive traffic networks, only to name some of them. Furthermore, for each of these realms their operation requires different competences and groups within an organisation, and it follows different guidelines. Some examples are:

• status of operation,
• safety,
• real time performance,
• real time monitoring of sensors and actuators, incl. fault detection and isolation,
• validation of signals (using redundancies created by the data network of the SoS),
• control,
• maintenance,
• assets,
• security.

These aspects have more or less known and understood relationships/dependencies which also will change in run-time. This provides a monitoring and management landscape which is very heterogeneous and dynamic.

A wide set of tools are available, each supporting one or a few of these dimensions. In most cases these tools mandate underlying information sources and data models, which sometimes correlates with current major industrial standards like ISA95, BIM, ISO 15926 and ISO 10303.

In summary a very complex and heterogeneous landscape of, to a large extent non-interoperable, tools and methodologies with no or little capacity to be integrated across SoS dimensions.

The emerging closer digital integration of industrial and societal functionalities and domains requires SoS integration and associated monitoring and management in very complex and heterogeneous environments. The current state of the art is far from efficiently enabling this. Such enabling will require closer cooperation and integration between several levels of the ECS domain stack. An example thereof is integration and functional interoperability between open and proprietary SoS architecture and implementation platforms, embedded software its tools and platforms and solution engineering its processes, tools and implementation platforms. Here solution requirements on lifecycle and evolution as well need to be considered.

To advance towards the vision technology and knowledge steps are required regarding:

• Monitoring and management strategies and architectural concepts in OT-IT environments.
• Methodologies and technologies for monitoring and management of multiple and interrelated SoS dimensions.
• Processes and technology for life cycle monitoring and management over SoS dimensions.