Supporting Smart Cities Quality Evaluation Exploiting Model-Driven Engineering

The concept of Smart City was coined in 2011 to define an idealized city characterized by automation and connection. As a consequence of the digital revolution, it further evolved detailing the multiple aspects characterizing smart cities themselves (e.g., sustainable mobility, environmental management, citizens inclusion). The European Commission published an agenda containing several objectives, called Sustainable Development Goals (SDGs), to reach in 2030 to face the crisis and promote smart, sustainable, and inclusive growth of European cities. On top of the SDGs, international projects targeted the definition of smart cities’ Key Performance Indicators (KPIs), along with their collection methodology, to capture the performance of a city in multiple dimensions and to support transparent monitoring and the comparability among smart cities. In this context, the Smart Governance performed in smart cities is in charge of decision-making processes, by exploiting KPIs assessment in order to have a complete vision of the cities in terms of smartness and sustainability. Overall, a framework for KPIs measurements in smart cities could be defined as a Quality Evaluation System (QES) that performs the assessment of a subject, i.e., the candidate smart city, w.r.t. some quality metrics, i.e., the selected interesting KPIs, as usually done for software quality analysis. However, despite the growing interest in smart cities evaluation and the existing guidelines on KPIs, no standard tools, languages and models to support systematic KPIs assessment processes do exist. This implies the lack of efficiency in the process of smart city evaluation and comparison that, in turn, affects the growth and improvement of smart cities. Moreover, this limitation hinders knowledge sharing among the smart city stakeholders, thus negatively affecting the smart city decision making processes. These challenges are further exacerbated by the complex nature of smart cities themselves. In fact, when speaking about smart cities, multiple dimensions (e.g., mobility, economy, environment) come into play, together with their corresponding stakeholders (e.g., private companies, public administrations, service providers). These dimensions are very heterogeneous, making it difficult also the interconnection among them. Furthermore, different cities can show diverse features and peculiarities (e.g., size, economic growth), thus affecting the relevance that some KPIs might have in their specific context. This entails that the relevant sets of KPIs might differ among different cities, thus implying the need for KPIs customization. KPIs evolve over time, which means that new KPIs can be defined or existing ones can be implemented in slightly different manners. Unfortunately, the currently available frameworks (e.g., manual, spreadsheet based, Web-based platforms) for KPIs assessment are still far from being flexible enough. To the contrary, they are tailored to a specific domain, or closed to customization, or rely on manual and error-prone tasks. Model-Driven Engineering (MDE) techniques are widely used to represent complex systems through abstract models. In this dissertation, we propose MIKADO– a Smart City KPIs Assessment Modeling Framework. Our solution is an approach supporting (i) the uniform modeling of both smart cities and KPIs, (ii) the automatic calculation of KPIs, and (iii) graphical visualization of assessed KPIs by means of dynamic dashboards. MIKADO enables the continuous monitoring and evaluation of the KPIs’ input parameters, by weaving open services and runtime models. The resulting approach provides a standard, but at the same time, a customizable process for smart cities governance administrators. MIKADO is characterized by domain-independence, indeed we demonstrate that it can be generalized into a QES for multiple domains, by exploiting the Multilevel modeling paradigm. We evaluated the approach in terms of (i) understandability of the MIKADO Domain Specific Languages, (ii) performance measured as execution time given variable models; and (iii) the latency.