Artificial Intelligence, Machine Learning, and Optimization Tools for Smart Cities

Scientific and technological inventiveness scarcely appears isolated. It is rather accompanied by a general cultural atmosphere, which may transfer influences from one cultural domain to another, often in a non-rational unconscious way. We may unveil analogous epistemological correlations between different scientific approaches, usually imposing the predominance of leading paradigms to others, of minor importance. We may also signalize the association of such leading “smart” scientific or technological paradigms to the totality of social expression, and thus to the representational art practices, to generalized social behavior and even to the political context of a given historic period. In this sense, we may also detect preceding scientific or cultural influences, which participate in future scientific and cultural formations and may incubate a future “smart” reality and give birth to it. Such correlations could probably be detected, between the contemporary “smart” approaches of the topology oriented design references or present-day digital technology and the scientific and cultural interest of the nineteenth century, the latter focusing on change, movement, and evolution. We could even use the terms “digital vitalism,” in order to refer to contemporary “animate” digital design in correlation to the notion of “vitalism,” correlated to the eighteenth and nineteenth-century mystical belief that a non-physical element, an inner “spark,” gives motion to beings. According to romantic thought, such vital energy could even be extended to inorganic beings; why not, to the inorganic, morphogenetic presentations of our computer screens.

In this chapter, we present a few architectural ideas and concepts relating with avant-garde technologies that have marked architectural thinking over the last few decades. Over this process, we question why architectural thinking in relation with new technological advancements has not yet combined forces to face pressing global challenges in public health and sustainability.

In 2008 bitcoin was born, based on blockchain technology. Soon after various blockchain frameworks came out, offering unparallel security and supporting rule-based logic. At their core two concepts are prominent, identity and rule-based transactions. These two inherent characteristics are cornerstone in any process. Cities can benefit from programmatically facilitating their own internal processes to achieve efficiencies at scale, securing their citizens’ personal data, and eventually creating their own individual identities. It is a long road ahead, where calculated steps must be taken in integrating the promise of blockchain technology in the quest of building the smart cities of tomorrow.

With the rapid growth of emerging technologies and services, smart city has been extensively studied with the development of modern societies. There are various applications and strategic views in different smart city domains such as smart urban planning or smart mobility, for designing smart services. However, it is still very complex to understand the interconnections and mutual influences among city services across the application domains. Therefore, based on a layered model of smart city, this paper investigates the emerging technologies and domain-specific services in a holistic way and plots them in the defined layers, so that the mutual interconnections and similarities can be identified. Our results show which technologies and services are developed in certain domains and also demonstrates how to organize technologies and services in terms of a smart city layered landscape model. The model allows us to compare the similarities and differences in each layer and identify possible interactions of smart services for each smart city layer across different smart city domains.

In the past decades, the evolution of big cities was followed by the emergence of smart city technologies. These new technologies enable in-depth analysis, optimization of public city services, and new modes of governance. However, as the cities’ infrastructure develops, the emerging data sources generate massive datasets. Currently, the efficient and accurate processing of smart cities’ enormous time series datasets poses a particular challenge to data scientists.

To overcome this problem, many high-level representations of time series have been proposed, including Fourier transform, wavelet transform, or symbolic representation. Applying fundamental symbolization techniques for time series with multiple variables, such as Symbolic Aggregate Approximation (SAX), results in distinct sequences of symbols for each variable.

A novel multivariate extension of SAX will be presented, which allows to express a multivariate time series with one sequence of symbols. Integrating individual sequences in one symbolic sequence provides better expressive power, while our modified SAX distance measure can be applied for clustering and classification tasks in smart cities, decreasing the enormous dataset storage and speeding up the big data processing. Performance evaluation shows that our multivariate symbolic representation results in better accuracy and dimension reduction than the classical SAX method.

Too much sunlight exposure would cause heat stress for people during the hot summer, although a minimum amount of sunlight is required for humans. Unprotected exposure to ultraviolet (UV) radiation in the sunlight is one of the major risk factors for skin cancer. Mitigating the heat stress and UV exposure caused by too much sunlight exposure becomes a pressing issue in the context of increasing temperature in urban areas. In this study, we propose an individualized and short-term effective strategy to reduce sunlight exposure for urban residents. We developed a routing algorithm minimizing pedestrian’s outdoor sunlight exposure based on the spatiotemporal distribution of sunlight in street canyons, which was generated by the simulation of sunlight reaching the ground using Google Street View (GSV) panoramas. The deep convolutional neural network-based image segmentation algorithm PSPNet was used to segment the GSV panoramas into categories of sky, trees, buildings, road, etc. Based on the GSV image segmentation results, we further estimated the spatiotemporal distribution of sunlight in street canyons by projecting the sun path over time on the segmented GSV panoramas. The simulation results in Shibuya, Tokyo, show that the routing algorithm can help to reduce human sunlight exposure significantly compared with the shortest path. The proposed method is highly scalable and can be easily extended to other cities with GSV data available. This study would provide a pedestrian-level strategy to reduce the negative effects of sunlight exposure on urban residents.

Electric vehicle (EV) sharing has experienced rapid development and has served as a flexible and environmental friendly means for urban transportation. However, charging an EV sharing fleet is still a challenge for business operators because of limited or costly access to charging facilities. In this chapter, we focus on how to charge a fleet to make EV sharing viable and profitable. Adopting the real data of car2go, we propose a queueing network model to characterize how customers endogenously pick EVs according to energy levels, as well as the implementation of a charging-up-to policy. In order to solve the proposed nonlinear optimization program, we develop mixed-integer second order cone programs as tractable lower- and upper-bound formulations. These models lead to practical insights related to charger resource availability and locations, EV charging policy, battery technological advancements, and urban spatial structure.

The fog/edge computing paradigm has been proposed to tackle the challenges inherent to the Internet of Things realm. Timely response, bandwidth efficiency, context awareness, data privacy and safety, and mobility support are some of the requirements that are only partially covered by cloud computing. A collaboration of both paradigms when developing deep learning solutions for the Internet of Things can be seen as a win–win approach. Time-consuming and hardware demanding deep learning models are built in the cloud with data provided by the fog/edge, and then these models are returned to the fog/edge for use. This work proposes a new architecture, based on the principles of reactive systems, for building responsive, resilient and elastic systems, where all components interact with one another through asynchronous message passing. As a proof of concept, two particular applications of this architecture in the realms of e-health and precision agriculture are presented.

We forecast that smart city technologies will reinforce the primacy of the most successful global metropolises at least for a decade or more. A few select metropolises in emerging countries may also leverage these technologies to leapfrog on the provision of local public services. In the long run, all cities throughout the urban system will end up adopting successful and cost-effective smart city initiatives. Nevertheless, smaller scale interventions are likely to crop up everywhere, even in the short run. Such targeted programs are more likely to improve conditions in blighted or relatively deprived neighborhoods, which could generate gentrification and higher valuations there.

It is unclear whether urban information systems will have a centralizing or suburbanizing impact. They are likely to make denser urban centers more attractive, but they are also bound to make suburban or exurban locations more accessible.

This chapter highlights the key challenges of our ongoing project in developing an information technology solution for emergency response and management in smart cities. We aim to develop a cloud-based big data framework that will enable us to utilize heterogeneous data sources and sophisticated machine learning techniques to gather, process, and integrate information intelligently to support emergency response to any disaster or crisis rapidly. After identifying the right data sources, we turn our attentions into investigating suitable techniques that can be utilized in disaster-event detection as well as extraction and representation of useful features related to the disaster. We also outline our approach in analysis and integration of disaster-related knowledge with the help of a disaster ontology. Our ultimate goal is to display and disseminate actionable information to the decision-makers in the format most appropriate for carrying out emergency response and coordination efficiently. We developed a dashboard-like interface to facilitate such goal.

For any disaster or emergency, the heterogeneous nature (texts, image, audio, and videos) and sheer volume of data instantly available on the social media platforms necessitate fast and automated processing (including integration and fusion of information originating from disparate sources). This chapter highlights our ongoing research in addressing such challenges in an automated fashion using state-of-the-art artificial intelligence and machine learning techniques suitable for processing multimodal social-media data. Our research contributions will eventually facilitate building a comprehensive disaster management framework and system that may streamline emergency response operations in the smart cities.