Simulation-Based Engineering of Heterogeneous Collaborative Systems—A Novel Conceptual Framework

Abstract

We discuss the collaboration support of loosely coupled Smart Systems through configurable hyper-frameworks. Based on the system-of-systems (SoS) paradigm, in this article, we propose the model of a novel extendible conceptual framework with domain-specific moderation support for model-based simulations and the engineering of complex heterogeneous systems. The domain knowledge meta-model and corresponding management enterprise architecture enable the creation of template-based specializations. The proposed SoS conceptual framework meta-model represents an initial framework prototype that supports modeling, simulation, analysis, and utilization of dynamic architecting of heterogeneous SoS configurations. A Smart-Habitat concept encapsulating Smart-Area, Smart-City, Smart-Lot, Smart-Building, and Smart-Unit abstractions illustrate the frameworks’ applicability. The proposed SoS conceptual framework represents the initial conceptual support for modeling, simulation, analysis, and dynamic architecting of heterogeneous SoS configurations. We plan to refine the component architecture meta-model, specify a language workbench with Domain-Specific Orchestration Language support, and verify the configuration-based simulation manifest creation. These actions lead to the framework’s next stage, an operational framework (OF) instance, as a transitional artifact to the aimed software framework (SwF) counterpart.

1. Introduction

Complex systems are composites that interact with each other over an arbitrary communication infrastructure (network) according to some behavioral (dynamic) patterns. The number and structure of system attributes (dimensions or parameters) used to specify the state of particular interacting systems, affected by the inherent nonlinearity, emergence, spontaneous order, and dynamically created feedback loops, exhibit nonlinear increases compared to the individual system size. The consequence is that understanding the entire system structure and behavior is not solely affected by the pure aggregation of participating components. Temporal configuration and context cause significant variations due to the emergent properties of swarm intelligence features. The intellectual effort, operational skills level, and sophisticated supporting tools are essential for collaborative context establishment and maintenance. The systems science, system engineering, and system-of-systems (SoS) methodologies favor system thinking and directly support the transformation of contemporary engineering approaches toward conceptual and operational framework utilization.

From the abstraction level aspect, the philosophy of system thinking is in [1] modeled as a refinement hierarchy composed of Systemic Sensibility, System Literacy, Systemic Capability, and System Roles. The authors particularly emphasize the crucial role of System Literacy in framing system-oriented research activities. The importance of behavioral theories in understanding the dynamics of complex systems and their state of the art, with the systematic literature review and cross-reference findings, is presented in [2]. Among the five suggested future research directions, from our point of view, the most challenging are data-driven and agent-based modeling approaches to future engineering activities. The engineering of next-generation complex systems is, at the same time, a transdisciplinary and cross-disciplinary activity that spans the entire life cycle [3]. It is a teamwork endeavor conducted in a virtual collaborative environment that utilizes the immersive simulation of an engineered system supported by a digitally operational framework that enables the automatic generation and optimization of candidate architectures and their components [4].

Addressing the operating context of collaborating systems through a system-of-systems (SoS) paradigm is fundamental in contemporary interoperability and integration approaches. The authors of [5] conclude that, although complex system management methodological approaches are broadly available, the main problem is that none possess firm success guarantees. Whether the particular methodology application would result in failure or success is not the consequence of proper selection. It dominantly depends on the operational context. Besides the well-known SoS obstacles (holism, uncertainty, ambiguity, context-dependency, emergence, statelessness, and monotonic characteristics), the importance of governance in regular operation and disaster recovery situations is crucial [5]. In [6], through the elaboration on Collective Adaptive Systems (CAS), the authors emphasized the importance of humans as unavoidable constituents of distributed cyber-physical systems. In the concluding part, they suggest that the inherent unpredictability of human behavior, the crucial constitutive role, and the active impact on systems dynamics, joined with human privacy, reliability, liability, usability, and controllability issues, represent the most challenging factors influencing contemporary and future CASs.

We find the research motivation in the orchestration support of distributed cyber-physical-systems that preserves individuality and encapsulates contextual, structural, and temporal variations through the entire life cycle of the overall SoS configurations and the configurations of any particular member. The life cycle encapsulates activities belonging to pre-design, design, implementation, deployment, exploitation, maintenance, upgrade, migration, and the retirement of component systems or the current configuration instance.

Based on the system-of-systems (SoS) paradigm, in this article, we propose the model of an extendible conceptual framework with domain-specific moderation support for model-based simulations and engineering complex heterogeneous systems. The domain knowledge meta-model and corresponding management enterprise architecture enable the creation of template-based specializations. The proposed SoSOpen Interoperable Conceptual Framework (OICF) meta-model represents an initial framework prototype that supports modeling, simulation, analysis, and utilization of dynamic architecting of heterogeneous SoS configurations.

The rest of the article is organized as follows:

Section 2 elaborates on problem domain aspects of conceptual framework specification and modeling and introduces the Smart City as a domain-specific example of heterogeneous collaborative systems.

Section 3 presents the foundations of the SoS Simulation Conceptual Framework (SCF) and domain-oriented specialization factory, the Smart Habitat abstraction ontology.

Section 4 discusses the SoS Simulation Conceptual Framework in the context of essential dimensions rating to the selected set of referenced frameworks.

Section 5 contains concluding remarks on SoS Simulation Conceptual Framework contributions and future research directions.

The last section lists the cited references.

2. SoS Engineering—A Problem Domain, Simulation, Modeling, and Smart City Influencers

Simulations may provide means for analyzing the complex dynamic behavior of engineered systems. They are applied in many critical engineering areas and enable one to address design and implementation issues before they become problems. Process simulation is a state-of-the-art technology to analyze process behavior, risks, and complex systems with inherent uncertainties. The systematic combination of simulation methods with empirical research has the potential to become a powerful tool in applied engineering research [7].

Data-driven and model-based simulations are the core mechanisms that drive the contemporary engineering or re-engineering of complex systems. The data-driven simulations of systems structure and behavior rely on two main issues: the quality and the role of the data used. They are highly dependent on the amount and stability of available data [8]. The model-driven approach builds new models and algorithms to reach the same goals. The main characteristic of data-driven modeling and computation is a framework-based integration of large multidimensional datasets with mathematical, statistical, and computational models, methods, and tools.

A model is a simplified and formally specified representation of a real-world entity or a system. It primarily serves for a better understanding of the system’s static and dynamic characteristics. Modeling appears as an unavoidable mechanism to cope with the complexity embedded in real-world entities, especially when preliminarily evaluating an arbitrary engineering endeavor. It is a creative task demanding problem-oriented thinking and a directed mindset built over the synergic interaction of modeling styles, methods, and guidelines. According to [9], contemporary modeling concerns address the model purposes (why), objects (what), stakeholders (to whom), formalisms (via), and tools (with). According to [10], the authors particularly emphasize the importance of inverse improvement in mathematical models as the reflection of model and data confrontation that helps in uncertainty qualification and improves inference mechanisms while selecting and reducing models. The data-driven model alternation aids in the comparative validation of simulation results obtained through physically based, knowledge-driven, or data-driven models.

Orchestrating heterogeneous systems introduces further obstacles, such astime-consuming modeling and simulation cycles, complex and changeable modeling fields, and complex domain knowledge. Complex system engineering activities balance two conflicting aspects: model abstraction level and its reusability potential. Raising the abstraction level introduces an order of magnitude amplification in top-down model transformation but lowers the overall reusability potential in concrete problem domains. To support the requirement analysis and initial design of the system through the entire life cycle in [11], the authors propose meta-model-based agile modeling and simulation technology concerning the Object Management Group’s (OMG) Meta-Object Facility (MOF) concept together with an agile heterogeneous systems modeling and simulation framework.

Simulation models act like virtual environments developed to test the hypotheses concerning the modeled system. They aid in obtaining the corrective policies far before their implementation in the real-world environment even starts. The usable simulation assumes that the underlying model and the corresponding data (physical or experimental) used to drive it accurately reflect the real world. The relevant data collecting is a challenging endeavor that favors the application of dimensionality reduction mechanisms that produces the essential dimensional set. The level of dimension essentiality is context-dependent and varies according to the simulation purpose. The selection of modeling techniques determines the way simulation models work. As elaborated in [12], the matter of scale (what aspects of the complex real-world system are relevant and sufficient) in simulation model specification is essential.

The state of the art of modeling and simulation inclines to the digital twin (DT) concept that couples real-world system structure and behavior with its virtual model. Although some authors claim that DT concepts originate from the National Aeronautics and Space Administration (NASA) projects in early 1960, in [13], the authors designate 2003 as the concept’s rebirth year and elaborate on definitions, misconceptions, application areas, the DT industry, enabling technologies, current research review, and future research challenges. Contemporary researches state that the DT concept helps span the system lifecycle through the real-time data acquisition infrastructure, model-based simulations, complexity reduction principles, and Artificial Intelligence mechanisms while raising the quality of systems’ inherent decision-making processes. The creation of a virtual digital twin helps with risk mitigation while engineering complex systems. A slight difference between the simulation of a real-world artifact and the creation of its digital twin exists concerning the scale. Simulations are dominantly oriented to single-process studies and support one-way information flow only. Digital twin (DT) utilizes a two-way information flow over multiple processes. Information flow in the DT environment starts with a real-time event initiated in the sensor network infrastructure and repeats according to the events’ time distribution. Therefore, simulations tend to be more suitable in the design phase of complex system engineering, while digital twin augments running systems’ behavior and usually exhibits higher process or product improvement potentials.

Depending on the granularity level at which we create the concrete digital twin of the real-world system component and the underlying problem domain, a wide variety of DT types exist. The DTs may be structural (element, part, component, collection, unit, system, and system of systems) or behavioral (process aspects that determine the timing schemas and direct the overall effectiveness).

From our experience, contemporary simulation models must be suitable for distributed, adaptable, and knowledge-based simulations. Simulation frameworks need agent-based (AB) mechanisms that support data distribution, process distribution, and combinations of data and process distribution of time and modal-dependent computation. Agent represents an autonomous entity that encapsulates a set of attributes (state), fundamental rules (behavior), and environmental awareness (context). Powered with the learning capability, it uses acquired knowledge and experience to adapt state, behavior, and context in a self-organized manner. The agent thereby constantly improves the decisions making process while staying on course of its overall mission. In [14], the authors elaborate on agent-based simulation challenges and propose the orchestration of Domain-Specific Modeling Languages (DSML) to achieve model-based simulation sustainability by fostering its robustness, efficiency, and maintainability. The approach to DSL graphical editor rapid development, based on the Sirius framework, is presented in [15]. The authors assumed that the Syntax-sensitive editor, whether textual or graphic, represents the core component of an arbitrary DSML framework.

Urban development and environmental sustainability are well covered by publications addressing particular domain-specific dimensions and technologies. Cities represent historically urbanized organizational systems that, from origination, growth, stagnation, and eventual collapse, exhibit a variety of obstacles affecting their sustainability and the sustainability of their internally embedded and externally affected ecosystems. Through particular epochs, the different technologies supporting urban ecosystem services have additionally introduced a variety of highly specialized impacts and obstacles affecting these services and the overall city development.

The essential role of architectural design (AD) and urban planning (UP) is to enable a forward-looking approach to building/facility creation. Construction Engineering (CE), although tightly coupled with AD and UP domains, expresses its routine mainly through the transition phase that transforms the ideas into usable urban artifacts and appears like a combination of backward and downward looking to the same process/product. All three domains are highly cooperative and run in the context of Environment Engineering (EE).

In [16], after the cross-reference analysis of relevant publications, the authors have concluded that, currently, there is no Transdisciplinary Collaboration Framework (TCF) supporting the collaboration between the Nature-Based Design (concerning the variety of natural influencers) and the Sustainable Built Environment methods. In [17], the authors elaborate on architectural, urban, and construction principles, methods, techniques, and tools that aid closely related disciplines’ collaboration through the extendible orchestration framework services support. Experimental evaluation of the proposed framework has enabled the creation of a general TCF prototype model. The envisioning of general TCF is the fundamental goal of our current research directions. They fulfill all future research suggestions of the previously referenced articles.

For a matured research domain, a question of an overall taxonomy is a fundamental precondition for consistency maintenance over different research directions and endeavors. In the recently published literature, sustainability and smartness are the most desirable features of future urban and ecology systems that demand a clear and generally acceptable set of measurable indicators. Formulation of a novel taxonomy of indicators that cross-correlates sustainability with smartness is at the core of [18]. The author proposes four main indicator categories: socio-cultural (11), economic (6), environmental (6), and governance (5), with a total number of 28 associated indicators.

Our experience shows that openness is the main success driver concerning the sustainability of used taxonomy. Within the SoS paradigm, heterogeneity is the main obstacle to reaching a consensus. Adaptation and mediation mechanisms of a supporting framework enable ontology and taxonomy mapping and foster risk mitigation.

The proliferation of contemporary information technologies in electronic service delivery support highly influences the digital transformation of arbitrary real-world organizational systems. The quality of delivered public services, concerning governing domain, appears as a relevant metric for sustainable living in human-centric social agglomerations. The most challenging aspect is the scope of available services. It directly influences the inherent cost/performance ratio through the entire lifecycle of the organizational system and the underlying information system supporting it. In [19], the authors elaborate on the emergence of urban computing support to provide intelligent services in Smart Cities. They have proposed a taxonomy concerning contemporary and future aspects of urban data, service design approaches, supporting applications and technologies, and resulting implications. The significance of cognitive cyber-security, data source diversity, complex distributed data persistency, and privacy concerns, among others, are our favorite challenging future research directions that have been pointed out.

The essential elementary component of an urban space defined as an arbitrary city is a building of any type or scale. In the SoS paradigm, a building is also a composite SoS. This approach facilitates naturally recursive algorithms based on simulation marshaling. The natural characteristic of recursivity assumes that the number of iterations, depth, and width of navigated SoS infrastructure is unknown before the simulation starts. The concept of Intelligent or Smart Home/Building (SH) is currently tightly coupled with the impact of living conditions of residential objects on human health status. The wide range of Smart Technologies embedded in the SH has enabled the efficient and economical usage of consumable resources and support for collecting a large amount of real-time data that is, afterward, dominantly used for monitoring and control purposes. They create a foundation for SH services roughly classified into three main intersecting service groups: energy efficiency, safety, and living. According to the concerning literature review presented in [20], the challenging future research directions belonging to the living group have to move the focus from pure technology aspects to their deeper incorporation into habitants’ daily activities. The other two groups need different scale network-based integration support on neighboring, district, town, regional, state, and even global levels. The SB attributes relevant to model-based simulation are elaborated on in [21].The dynamic generation of inner and outer residential house shapes based on parametric simulation and the spatial grid system, presented in [22], introduces an open multidimensional approach to estimate the level of smartness embedded in individual SB to gain the sustainable balance of internal and external influencers.

The possibility of creating virtual or augmented reality based on available software tools and integrated development environments becomes a challenge to domain experts and software designers. The inherent complexity embedded in real-world concepts promotes modeling and simulations as the unavoidable mechanisms for the preventive evaluation of engineering achievements. The particular challenges lie in modeling and parametric simulation of space, building, and urban blocks that enable the analysis of existing urban environments to gain potential revitalization or estimate future achievements concerning engineering Smart Cities.

Among several modeling languages considered suitable for Smart City modeling, Systems Modeling Language (SysML), Building Information Modeling (BIM), and City Geography Markup Language (CityGML) are unavoidable representatives.

SysML is a general-purpose graphic semi-formal modeling language developed for system architecture specification and modeling purposes. SysML is an Open-Source Project (initiative) initially formulated in 2006 by the Object Management Group (OMG). Concerning recent specification releases, it appears as the profile-based extension of Unified Modeling Language Two (UML2). Through nine basic diagrams and the traceability feature of linking different specifications, it facilitates a complex system’s static structure and dynamic behavior specification and modeling in compliance with the fundamental principles of the Model-Based System Engineering (MBSE) initiative.

According to the simplified definition of BIM, it is a digital representation of the facility’s physical and functional characteristics. It appears as a shared knowledge resource holding the information about a facility and forming a reliable base for decision-making process support during the facility’s entire lifecycle. A basic premise of the model is the collaboration of different stakeholders at different phases of the facility life cycle aimed to: insert, extract, update, or modify model information reflecting the role/roles of each stakeholder.

The CityGML represents an open conceptual model standard that enables the integration of urban geo-data with applications supporting Smart Cities and Urban Digital Twin [23]. Based on a core semantic information data model and XML-based format, CityGML supports storing and exchanging virtual 3D city and landscape models. In contrast to other 3D vector formats, CityGML uses a rich, general-purpose information model that extends the basic geometry and appearance of modeling objects. The fundamental characteristics of CityGML are modularization (one core and three extension modules), multi-scale modeling (with five levels of detail(LODs): LOD0—highly generalized terrain model, LOD1—block model, LOD2—realistic model with differentiated surfaces, LOD3—highly detailed architectural model, and LOD4—interior structure of LOD3 3D objects), semantic geometry modeling (geometric features are relevant only at a spatial level), external reference (access to external resources), appearances (any entity at an arbitrary level of details may possess a different view), prototyping (support for evolution prototype creation via surrogate objects linking), and runtime extension support (extending application services by General Objects and Attributes, and data model via Application Domain Extensions) [24].

Although some resources suggest that general-purpose modeling languages and their diagrams are still complex, confusing, and too abstract for effective use [25], we consider them the ultimate candidates for the extendible language library of any sustainable model-based simulation framework. Considering Smart City or Smart Building as an example of a Socio-cyber-physical system, another challenging direction emerges from the elaboration of [26] on goal-oriented language integration with formal and semi-formal modeling languages of an arbitrary type [27].

The generic problem domain and its specific DTs are the main topics of current digital transformation research and development. According to [28], Urban Digital Twin (UDT) as a final product is not still commercially available. Building it through an interoperable orchestration framework appears as the rational approach. The available single-point solutions such as Building Information Modeling (BIM), Computer-Aided Design (CAD), Geographic Information System (GIS), Energy Management System (EMS), and Building Management System (BMS), even when armed by third-party Artificial Intelligence (AI), Machine Learning (ML), or other plug-ins, do not offer scaling, flexibility, and interoperability that UDT demands to support the entire life cycle of individual constituents and their dynamic, context-dependent configurations. Comprehensive cross-reference research, presented in [29], systematizes and visualizes the bibliography concerning the Internet of Things (IoT), BIM, and DT’s role in the contemporary construction domain. The authors conclude that IoT, BIM, and DT have promising potential impacts in Construction 4.0. In [30,31], the authors review challenging aspects of building and construction activities that foster Smart City support mechanisms through the unifying framework model incorporating Industry 4.0 and Construction 4.0 technologies in the context of DTs.

Table 1. Additional Smart City Research Cross-reference classification with framework area impact.

Topic	Subtopics	Individual Reference Impacts	Impacting Framework Areas
Model-based simulations of building/city attributed	Ecology and ecosystem	Urban ecological network simulation at the city-scale resolution has been elaborated in [32]. The importance of parametric models in helping designers to test residential artifacts over space and time [33] and ecosystem services [34] round up the contemporary ecology and ecosystem drivers.	Parametric-based simulations of course-grained habitats.
	Human behavior and wellbeing	In [35], the authors elaborate on the importance of recognizing and monitoring the dynamic activities of humans in residential settings. One of the crucial dimensions of contemporary habitats is healthy living and related concepts such as the integration of Neuro-architecture in architectural and urban design [36].	Human presence demands frameworks to support features that enable nondeterministic discrete systems.
	Building geometry, models, and parameters	In [37], the authors elaborate on extending CityGML 3D building models with multi-scale data management features. In [38], the authors elaborate on the extended reality technologies fostering buildings’ attributes state monitoring in an Urban Digital Twin (UDT) context, while [39] extends UDTs to the overall urban management mission.	The framework’s ability to run data-driven parametric simulations, arbitrary data sources, and support for the extendible set of modeling languages.
	Urban growth and flow management	Urban planning and growth estimation effects on urban service delivery represent the mainstream of urban growth simulations [40]. Joined with urban flow management [41,42], the dimensionality of the simulation model emerges.
	Energy simulations and management	In [43], the authors elaborate on the toolbox enabling extensive search over large CityGML data sets for energy simulations on a district level. The evaluation of Smart-Grid technology to predict energy consumption [44] and energy efficiency monitoring and control [45,46] illustrate possible approaches in energy simulations and management.	Simulation model generation based on large-scale public data sets clustering.
Nonfunctional requirements	Security, safety, and access control	General aspects of safety and security in intelligent habitats [47,48], indoor and outdoor early fire prediction [49], detection, and monitoring [50].	Incident monitoring services and safety verification.
	Other abilities	One of the challenging approaches to MBSE is the proper handling of nonfunctional requirements. In [51], the authors elaborate on integrating the system’s reliability, availability, and maintainability features in the early engineering phases of the Model-Based System Engineering paradigm.	Framework extending with constraints-driven simulation support.
Industry 5.0 challenges	Concepts, education, and sustainability issues	The need to overcome the limits of the classical engineering paradigm is emphasized in the context of sustainability and education for sustainability [52]. The sustainability issue in systems development emerges from the unachieved goals of previous industrial revolutions (4.0) and the projection of future ones (5.0), moving the focus to the cooperation of human and automated intelligent systems [53].	Intelligent framework development represents a crucial future challenge.

systematizes the additional Smart City research cross-reference with the impact on the proposed frameworks’ areas.

The former fosters our motivation to promote the contemporary Smart City as a suitable candidate for problem domain stakeholders and influencers elicitation. The corresponding UDTs are usually 3D virtual representations accompanied by large-scale multimedia data sets and software tools supporting a limited number of underlying system dimensions.

Incorporating real-time data generated by the variety of single responsibility subsystems within an interoperability framework enables cross-simulations of virtually any combination of single participating constituents of contemporary and future Urban or Environmental structures. That is why they are the challenging candidates for framework-based integration over a system-of-systems paradigm.

3. The Foundations of SoS Simulation Conceptual Framework (SCF)

The life cycle of Complex system engineering characterizes a continuous delivery approach based on flexible architecture development and incremental components deployment. The risk mitigation in the pre-implementation stages relies on a platform-independent, simulation-based top-down approach to analysis, specification, and modeling in solution and design domains. On the other hand, the implementation and operational phases are platform dependent and currently follow the bottom-up agile approach that, in the current context, tends to deliver the maximum possible value to the end users. With the inherent variations in context and technology throughout the entire life cycle, the sustainability of the engineered system depends on the clarity and stability of a global picture. Orchestrating heterogeneous systems introduces further obstacles, such as time-consuming modeling and simulation cycles, complex and changeable modeling fields, and complex domain knowledge. Complex system engineering activities balance two conflicting aspects: model abstraction level and its reusability potential. Raising the abstraction level introduces an order of magnitude amplification in top-down model transformation but lowers the overall reusability potential in concrete problem domains.

The conceptual framework, software framework, or prototyped operational development-first framework is always a challenging dilemma in the digital transformation of distributed cyber-physical-systems.

The role of a conceptual framework (CF) as a coherent mental representation that structures and generalizes knowledge, skills, and experiences in problem-solving endeavors are incomprehensible without conceptual framing of a specific type. As such, CFs serve as recipes for creating real-world artifacts. They are dominantly problem domain-oriented, platform-independent, and buster model-based development and model-driven simulations that aid in a better understanding of problem domain static structure and behavior. In the interoperable SoS context, CFs need to preserve individuality and encapsulate contextual, structural, and temporal variations through the entire life cycle of the overall SoS configurations and the configurations of any particular member.

On the other hand, the creation of operational frameworks (OFs) relies on the corresponding conceptual ones through template-based code generation or strategy-based orchestration of micro-services. The recent trends in Domain-Specific Language (DSL) development have established a challenging target for current Model Driven Engineering research. DSLs are dedicated languages that better suit domain experts and support them while specifying concrete domain solutions. The model-based generation of the initial manifest script, composed of the highest possible common and standard abstract concepts from the particular problem domain libraries, and its dynamic context-based runtime alteration represent a typical form of the OF’s dynamics.

Software frameworks are abstract, course-grained code templates (meta-code composites) providing generic functionality. They gain complete control over object flow and extend mechanisms for the efficient and effective creation of scalable and resilient support for domain-specific specializations. The process of a platform-dependent model to code transformation depends on the programming language used, available software libraries, reusable components, and integrated development environment. Software libraries are composed of reusable pre-coded and pre-tested dedicated pieces of code that buster software development by direct insertion into the architecture of applications’ code. In component-based architecting, the software development process transforms into the component’s interconnecting with or without interface adaptation or component mediation. The component’s internal ingredients are inaccessible to the outer world, thereby busting the reusability only on the functional level. The resulting code-building process supports the code inclusion, with or without modifications, according to explicitly supported specialization enablers defined within each library component.

Contemporary Industry 5.0 directions address sustainability challenges and the intelligent support for complex system engineering and transpose the creation of integrated development environments into framework-based interoperable tools orchestration supporting human collaboration with intelligent systems, demanding inter-domain and cross-domain cooperation of different experts. To raise the level of generality, we define a system as recursively dynamically complex if its internal structure is defined and managed by a data-driven configuration manifest object whose content is interpretable by an open set of different strategies that may vary in compliance with associated systems’ context information.

That is why we decided to specify the conceptual framework as an entry-level artifact and gradually transform it into an operational or software counterpart (or even both). Our choice to select the conceptual framework approach is contextually justified by [54], where the author discusses four templates for conceptual article structuring and their inherent potential exposure (theory synthesis, theory analysis, typology, and model). The latter two: the typology (through the categorization of dimensions) and the model (dynamic relationships predicted through model-based simulation supported by the script manifest), are dominant in preparing the rest of this article.

The founding mainstreams unleashed in published research and development and standardization in industrial practices promote language-based modeling approaches. The Meta-Object Facility (MOF) specification, as four levels abstraction model (instance, concept, meta-concept, and meta-meta-concept, respectively), upraises language (meta-meta-concept) as the highest abstraction level with unique self-specifying formalisms (See Figure 1).

According to the SoS problem domain simulation, modeling in the context of Smart City influencers (Section 2), we specified the main characteristics of a proposed simulation framework as follows:

• A transdisciplinary conceptual framework based on general meta-model foundations with domain-specific template-based meta-model extensions;
• Multidimensional and multilevel dynamic reconfigurability of independent heterogeneous participants (SoS paradigm support) based on secured master broker architecture with component-hosted services;
• Manifest-based interpreting—language-driven simulations;
• The explicit support for a hybrid repository of information resources;
• Two-way information flow over multiple processes (digital twin paradigm support) with simulation only, the operation only, and combined regimes;
• Data-driven, model-driven, and agent-based simulations support with an open set of simulation techniques covering deterministic, stochastic, static, dynamic, continuous, event-driven, quantitative, qualitative, single, and hybrid;
• Integrated Stakeholder support.

The development of Computer-Supported Cooperative Work Tools superimposes software engineering and Problem Domain Designing competencies. It requires the ultimate understanding of what a particular stakeholder wants to achieve and keeps in mind that a sustainable solution emerges when the expectations, support, and actual behavior of created artifacts are cross compliant and well aligned. The fundamental challenge is the level of domain knowledge gained to formulate an operationally usable framework. Figure 2 shows the domain-specific composite model of generalized framework requirements.

A comprehensive explanation of the particular domain packages is as follows:

• The problem domain requirements (PDR) package encapsulates entirely domain-specific expert stakeholders’ requirements and fosters the solution domain separation of concerns, thereby raising the reusability level of a particular solution in different problem domains;
• The operational domain requirements (ODR) package encapsulates problem domain requirements acquired from operational stakeholders belonging to the problem domain;
• The solution domain requirements (SDR) package refines PDR and ODR packages in the context of the concrete technology solution constraints;
• The execution domain requirements package is derived from the SDR package and is responsible for framework context (context) management in the execution environment that implements the operational context interface and supports manifest-driven (manifest) dynamic marshaling;
• The implementation domain requirements (IDR) package refines the SDR and the EDR in the implementation constraints context of concrete technologies.

The analysis of complex problem domains depends on the availability and accessibility of domain knowledge where inherent individual complexity often discourages software engineering efforts. The clue in gaining a higher framework usability level is supporting the actor’s goals (what a particular actor wants to accomplish through framework services orchestration) rather than the way to achieve them (how to orchestrate services). Figure 3. illustrates the fundamental ontology and architecture aspects of the domain meta-model (upper) and domain management enterprise architecture model (lower) interdependences. The existence of domain-specific mental models fosters an understanding of actors’motivations and the emotional and philosophical context in which they operate (Figure 3 (lower)).

The modeled concepts, with no obvious semantics, are briefly explained below:

• Domain knowledge (DK) is a set of concepts and terminology understood by practitioners involved in the area(s) of expertise that usually encapsulates technical literature, engineering implementations, customer surveys, expert advice, and requirement base;
• Domain knowledge representation (DKR) is a general foundation model guiding domain experts through domain engineering/re-engineering activities. It possesses a suitable domain ontology and supports knowledge capture and representation mechanisms. The ontology of a particular domain encapsulates its terminology, concepts, taxonomy, and relations, joined by the possible set of domain axioms or rules. The DKR supports the Domain Knowledge Manager Set of Services that encapsulate domain knowledge and enables domain analysts to utilize domain analysis methods;
• Domain analysis (DA) is a process that supports identifying and analyzing the domain knowledge representations. TheDA is derived from existing system studies, underlying theory, emerging technology, and development histories within the domain of interest and enables the development and maintenance of the Domain Model Base(DMB);
• Domain model representation (DMR) aspects of domain-specific modeling activities are currently dominantly focused on linguistic approaches raising the importance of the appropriate Domain-Specific Languages (DSLs) formulation and design.

The separation of domain-specific aspects of SoS architecture and the simulation frameworks’ universal kernel bridges the gap between different problem domain applications and the generality of the solution domain. This generality enables the separation of domain-specific specializations and the framework’s kernel architecture abstractions based on the delegating principle.

Considering software support for distributed systems composed of decoupled software components, the well-aligned conceptual architecture model is a broker architectural pattern. The broker acts like a communication moderator that hides the internal architecture of service requestors and service providers and encapsulates heterogeneous components’ communication. The Broker Pattern Ontology Model, used in the context of our approach, utilizes the delegation principle to decouple responsibilities over well-defined classes and the inheritance mechanism for hierarchy building (Figure 4).

In the specified model, the Broker is an abstract concept using a delegation principle in responsibility sharing. The delegation mechanism consists of the Broker Service Delegate (responsible for hiding structural and operational details behind the extendible set of interfaces that decouple Exchange, Utility, and Administration services), the Broker API Delegate (responsible for the Broker’s Application Programming Interface support), and the Broker Security Delegate (responsible for Broker’s Security Policy Handling) meta-classes.

Based on the related work analysis and SoS supporting mechanisms, we defined a Master Broker Architecture Model (MBAM) that enables dynamic configuring of hyper-framework instances. The role of the Master Broker is to broker the participating Brokers in compliance with the marshaling manifest script formed for particular session management.

Requestor, Provider, MasterBroker, and NodeBroker are structural and behavioral specializations of the abstract Broker concept. This hierarchy enables the dynamic creation of hyper-structured instances and is one of the main challenges for resource discovery and marshaling.

The Requestor delegates request to the RequestorProxy while the Provider delegates its requests and services to the ProviderProxy, thereby supporting their surrogate instances handling when appropriate. These features buster navigation when concrete instances are large objects whose content is irrelevant to the particular marshaling session.

Action, Protocol, Repository, Resource, Service, and State are meta-concepts that support the multidimensionality of structural and behavioral characteristics of the particular higher granularity meta-concepts in a dynamically extendible manner.

Figure 5 contains a master broker architecture model that forms a hyper-graph structure derived from the discussed ontology meta-model. Master Broker is a broker that orchestrates NodeBrokerInstances and potentially other Master Broker Instances and supports two delegated responsibilities: SoS Security Handling (the SecurityPolicyBroker) and SoS Services Handling (the ServiceBroker) and hosts a persistent object repository (the Broker: Database), based on service discovery mechanisms and concrete manifest script Master Broker orchestrates services and Information Resources that are marshaling objects under its jurisdiction.

In Figure 6, the SoS Framework Node Broker Enterprise Architecture Model represents the conceptual template model of the proposed frameworks’ kernel internal architecture.

The Node Broker consists of the following set of dynamically configurable components:

• The Framework Broker is a conceptual framework orchestrator that enables the dynamic creation of a manifest script through the Configurator component services;
• The Modeler Component encapsulates modeling activities in compliance with the extendible set of domain-specific modeling formalisms, methods, and tools;
• The Simulator Component encapsulates simulating activities in compliance with the extendible set of domain-specific simulating formalisms, methods, and tools;
• The DataAnalyzer Component encapsulates data science analytic formalisms, methods, and tools closely related to The ComplexityHandlerComponent’s responsibilities for complexity handling based on an extendible and configurable set of software tools supporting different complexity reduction mechanisms;
• The RealTimeDataAccess Component supports the IoT sensors’ data generated over a configurable network of sensors supplying the virtual Digital Twin with real-time data obtained from the real-world systems through the embedded IoT sensor networks;
• The DataStructureHandler Component supports the separation of concerns regarding The FrameworkBroker state management, while The ServiceHandlerComponent supports an internal set of coarse-grained services (FrameworkServices) that run over the frameworks’ overall data structure;
• The Visualizer Component consists of the opened set of Viewers that support presentation and animation services running over data structure elementsencapsulated by the DataStructureHandler;
• The NetworkHandler Component is responsible for connectivity tasks according to the manifest script object generated for a particular service or data walks;
• The InteroperabilityModerator Component is responsible for possible adaptations and mediations while fulfilling the marshaling plan;
• The StakeholdersHandlingComponent is responsible for human actors management and dominantly performs User Experience and User Interaction services;
• The PersistentObjectHandler Component is responsible for hiding individual repository component objects, and storing and retrieving operations regarding the particular data and storage model types. It is dynamically extendible with an arbitrary number of repository handler instances that implement concrete get() and put() operations on the abstract information resource. Under the abstract information resource, we assume an arbitrarily complex data, information, or knowledge abstraction that exists in the Broker Node persistency layer domain;
• The SecurityPolicyBroker is an instance of application runtime that implements security policy on a particular Broker Node level.

A concrete instance of SoS Framework Broker Node component architecture is a specialization of the presented conceptual framework Broker Node architecture, according to the Domain-Specific Meta-Model and Domain Management instances (Figure 3).

In the related work section of this article, Smart Cities and Smart Buildings emerged as suitable candidates for building the foundations of the SoS conceptual framework architecture model. A Smart Habitat concept’s role is to raise the level of abstraction and allow the creation of Bridge Specialization of the SoS Framework’s conceptual model (Figure 7).

SmartHabitat inherits structural and behavioral characteristics of the BrokerNode composed of interrelated Components under the Configuration directed associative collection.

The specialization of a Bridge Pattern decouples SoS abstractions and SoS implementations via two distinct interconnected hierarchies, allowing the abstraction and implementation parts to grow or shrink independently without impacting each other. This concept busts the extendibility of the proposed framework.

Depending on the granularity level, SmartHabitat abstraction is currently specialized by the SmartArea, SmartCity, SmatrLot, SmartBuilding, and SmartUnit abstract meta-concepts. Following the Bridge Design Pattern, SmartHabitat specifies the extendible set of Delegation interfaces that hide the implementation details related to BrokerNode component architecture implementations hidden behind eight interface specializations: Modeling, Tuning, Simulating, ComplexityReducing, Visualizing, Data Analyzing, Engineering, and Operating. SubDomainImplementation meta-class moderates domain-specific inheritance-based concrete implementations: Architecture (architectural design activities), Urban (responsible for urban design activities), Construction (construction engineering activities), Communication (responsible for arbitrary flow design activities), Energy (energy infrastructure design activities), RealTime (IoT sensor network design activities), Management (control and management activities), and Constraints (constraints management activities).

Management generalization is further specialized by supporting the following control and management activities: SocialAndHuman, Ecology, Supplies, Risks, and Growth. Constraints generalization is responsible for the extendible set of nonfunctional requirement specializations (Safety, Security, Reliability, Availability, and Maintainability).

4. Discussion and Simulation Frameworks Comparison

Balancing the abstraction level and degree of reusability in model-driven system engineering determines the approach to specification, modeling, and development of a supportive SoS simulation and engineering. Concerning the global picture stability, we propose the model of an extendible conceptual framework with domain-specific moderation support for model-based simulations and engineering complex heterogeneous systems. The domain knowledge meta-model and corresponding management enterprise architecture enable the creation of template-based specialization. Besides the initially focused general dimensions discussed and originated from the Introductory section, the Smart Habitat abstraction mainly originates from Section 2, Smart City reference analysis. The lack of Transdisciplinary Collaboration Frameworks (TCFs) supporting Nature-Based Design (concerning the variety of natural influencers) and collaborative methods in Sustainable Built Environments founded the domain-specific transformations embedded in the SCF formal specifications. The configuration-driven dimensionality, fostering relevant data collecting only, introduces the application of dimensionality reduction mechanisms that dynamically create the instances’ relative essential dimensional sets. The level of dimension essentiality is context-dependent and varies according to the simulation purpose.

Our experience shows that openness is the main success driver concerning the sustainability of used taxonomy. Within the SoS paradigm, heterogeneity is the main obstacle to reaching a consensus. Adaptation and mediation mechanisms of a supporting framework enable ontology and taxonomy mapping and foster risk mitigation.

Table 2. The SCF main hyper-dimensions list.

Dimension	Description
Heterogeneous SoS Collaboration	The main issue in problem domain addresses the collaboration of heterogeneous independent Systems (SoS paradigm).
Configuration Scaled Extendibility	Configuration 1-based scalability and runtime extensibility.
Digital Twins Support	Full support for DT-based collaboration supporting model-based simulations and engineering coupled with real-time sensor data.
Modeling	Explicit support for model-based simulations and engineering
Simulation	the orchestrated extendible set of simulation methods and tools.
Multi-Layered	Layered architecture support enables complexity hiding.
Hyper-Dimensional	Any particular dimension-related instance may be a dimension.
Dimensionality Reduction Support	Explicit support for opened set of dimensionality reduction methods and tools.
Linguistic Foundation	The explicit support for the extendible set of language workbenches orchestrated through manifest scripting.
Role-based Configurability	The explicit support for role-based configurations specialization fostering stakeholder-dependent customizations.
Reverse Design Support	Forward chaining and backward tracking traceability built in particular configurations.
Framework Type	Conceptual, Software, Operational, Experimental, and Hybrid.

¹ Structural, behavioral, and time-dependent.

systematizes the selected set of SCF dimensions used in concluding the comparative analysis.

Table 3. The cross-dimensional comparison of related frameworks impact on SCF.

		Compared References 1
Figure 8, Figure 9 and Figure 10 Markers		a	b	c	d	e	f	g	h	i	j	k
Dimension	No 2	[55]	[56]	[57]	[58]	[59]	[60]	[61]	[62]	[63]	[64]	[65]	SCF
Heterogeneous SoS Collaboration	I	1	1	2	7	10	6	2	0	5	0	0	10
Configuration Scaled Extendibility	II	0	0	5	6	10	5	6	5	5	0	0	10
Digital Twin Support	III	0	0	0	0	0	10	0	0	0	5	0	10
Modeling	IV	0	2	7	5	7	5	7	10	8	9	6	10
Simulation	V	2	7	5	8	8	7	8	8	9	9	9	10
Multi-Layered	VI	0	4	7	5	8	6	7	8	6	9	5	10
Hyper-Dimensional	VII	3	0	7	5	8	6	7	8	6	5	5	10
Dimensionality Reduction Support	VIII	0	0	0	0	5	0	0	0	5	0	0	10
Linguistic Foundation	IX	0	0	8	5	0	0	9	0	0	4	0	10
Role-based Configurability	X	0	0	0	0	0	0	0	0	0	0	0	10
Reverse Design Support	XI	0	0	0	0	0	0	0	5	0	0	0	10
Conceptual	XII	0	6	8	5	0	7	5	10	9	10	0	10
Software	XIII	2	0	2	7	10	0	7	0	0	0	10	0
Operational	XIV	0	0	0	5	5	5	0	0	0	0	2	5
Experimental	XV	6	0	4	2	0	4	7	0	0	0	5	0

¹ Impact scale (0–10) and ² dimension indicator in diagrams (Figure 8, Figure 9 and Figure 10).

contains a detailed specification of relevant orchestration frameworks published in the available and analyzed literature that highly correlates to the SCF over a defined set of SCF-supported dimensions (Table 2) and their relative impact on the proposed framework characteristics scaled from 0 to 10.

Figure 8 represents the radar diagram correlation of selected references rated to the SCF on all 15 dimensions. In Figure 9, a selected subset of closely related generic dimensions’ comparative visual representation over the first eight representatives expresses the relative rating for highly rated SCF dimensions.

Figure 10 contains the cross-relation rating radar diagram representations of the remaining seven dimensions. In our opinion, Figure 9 and Figure 10 aid the representation clarity of the compared dimension ratings.

The additional simulation frameworks related to comparative analysis show that simulations are dominantly oriented to single-process studies and support one-way information flow only. On the other hand, the digital twin (DT) utilizes a two-way information flow over multiple processes. Therefore, simulations tend to be more suitable in the design phase of complex system engineering, while the DTs augment running systems’ behaviors and usually exhibit higher process or product improvement potentials. Combined, they foster the novelty of the proposed SCF. The interoperability challenges and the embedded support for model-based heterogeneous SoS framework-based simulation support are crucial for sustainability assessment.

Network-based systems generating challenging issues when transforming the SCF into an operational or executive framework instance consider the Configuration Scaled Extendibility dimension, the IoT, and pervasive and mobile systems with dynamic configuration and reconfiguration ability. The Smart Collaborative Environments [66], generated through marshaling script interpretation, may serve as the encapsulating mechanism for either process or chain-oriented composition of dynamic SCF service configurations.

Along with standard language support features (model editing, merging, splitting, and consistency checking), the integrated MBSE environments (tools) offer a variety of extending features (simulation and formal verification, model-based automatic or semi-automatic code, and test scenario generation). Through collaborative or cooperative interoperability scripting, these tools may form dynamic toolchains, methodology chains, or technology chains.

Support for the interoperability issues among arbitrary abstraction levels (tools, methodologies, and technologies) introduces the extendible tool library dimension of any sustainable model-based simulation framework as a must.

5. Conclusions and Future Work

In the introductory section, following the stepwise approach, the foundations, building the mainstream challenges concerning the SoS sustainable conceptual framework model, are elaborated. The number and structure of systems’ attributes (dimensions or parameters) embedded in complex systems are strongly affected by inherent nonlinearity, emergence, spontaneous order, and dynamically created feedback loops on different granularity levels. Complex system engineering activities balance two main conflicting aspects: model abstraction level and its reusability potentials. Raising the abstraction level introduces an order of magnitude amplification in top-down model transformation but decreases the overall reusability potentials in particular problem domains. Modeling, as a creative task based on problem-oriented thinking directed by a designer’s mindset, depends on the synergic interaction of modeling styles, methods, and guidelines intended to cope with the inherent complexity of the modeled system. Simulations, on the other hand, provide means for analyzing the complex dynamic behavior of modeled systems before launching the construction phase, thereby mitigating the impact of the potential risks.

We have concluded that the sustainable modeling and simulation framework has to at least address the following main issues: a dynamically configurable System of Systems (SoS) infrastructure, domain-specific specialization support, a digital twin real-time sensor network as feedback for continuous improvements in referent architecture, a real-time accessible input/output hybrid information resources repository, an open set of simulation techniques, an executable, multidimensional, dynamically configurable, traceable, and formally verifiable model, armed with domain-specific modeling languages orchestrator, and powerful visualization and presentation mechanisms.

In the related work section’s cross-referent analysis of a selected set of representative references, we have refined an initial set of requirements concerning the conceptual SoS framework’s features derived from human habitats as building objects of sustainable, intelligent environments raised to the higher abstraction level via the conceptual framework’s meta-model specification.

The importance of domain knowledge representation, modeling, and management helped derive the domain meta-model and corresponding supportive management architecture. The distributed nature of the internal architecture of systems under consideration has influenced the generic building block architecture formulation based on the master broker approach to hyper-architectural pattern specification. The BrokerNode enterprise architecture model has covered all formulated characteristics of sustainable conceptual framework propositions.

With a Smart Habitat specialization, based on the Bridge Design Pattern object-oriented meta-model, the overall proposed approach’s applicability reached the initially estimated verification and validation.

Regarding the former elaborations, we believe the proposed SoS-based conceptual framework represents more than just a promising framework for modeling, simulation, analysis, and utilization of dynamic architecting of heterogeneous SoS configurations. As a conceptual framework, the SCF represents a reference architecture that, under the same umbrella, integrates a relevant set of dimensions that were, independently or partially integrated, applied in other research and engineering endeavors.

The highlighted support for digital twin technologies involvement directs the mainstream of future research endeavors related to transforming a conceptual one to the software development framework instance.

The dynamic marshaling of model-based simulations, engineering, control, and management activities, directed by manifest scripts, demands the specification and development of suitable domain-specific orchestrating language and the related language workbench supporting tools.

Contemporary software and system engineering development leads to the heterogeneous framework-based collaborative interoperability support direction, incorporating Artificial Intelligence and Machine Learning mechanisms, low-coding or no-coding development, and multi-platform integration, regardless of some potential software frameworks’ drawbacks that demand designers’ proper attention and solutions.

We plan to refine the component architecture meta-model, specify a language workbench with Domain-Specific Orchestration Language support, and verify the configuration-based simulation manifest creation.

These actions lead to the framework’s next stage, an operational framework (OF) instance, as a transitional artifact to the targeted software framework (SwF) counterpart. Through SwF, we intend to address the main software framework challenges related to the proper handling of the following issues:

More demanding learning curve and potential frameworks’ rigidity, which are moderated or even eliminated through domain-specific language formalisms, scenario-based learning, context-based extendibility, and customization support;

Hidden dependencies and compatibility issues are usually the consequence of software frameworks’ strong relations to the particular languages, code libraries, and reusable component pools and may be moderated or eliminated by suitable wrapping support with low or no-coding support; and

Frameworks’ maintainability, which crucially depends on quality software development and the scalability of frameworks’ meta-model transformation support.