Abstract
In recent years we have seen both the proliferation of smart devices in people's daily habits and activities and the advent of the Internet of Things as the basis on which numerous smart applications and services are being developed to solve various problems in cities, from traffic monitoring, pollution and noise monitoring to healthcare, transport infrastructure and financial data processing. The plethora of available data sources has provided a range of important options for monitoring and assessing the state of a smart city in real time. New distributed big data processing and analysis systems, such as Apache Hadoop, Apache Storm and serverless systems, have also been proposed to manage this large volume of data and provide low-latency, real-time data processing. Beyond data processing, however, there are significant research challenges that need to be studied in order to achieve effective use of these systems. These challenges arise from the fact that each urban data source is gove ...
In recent years we have seen both the proliferation of smart devices in people's daily habits and activities and the advent of the Internet of Things as the basis on which numerous smart applications and services are being developed to solve various problems in cities, from traffic monitoring, pollution and noise monitoring to healthcare, transport infrastructure and financial data processing. The plethora of available data sources has provided a range of important options for monitoring and assessing the state of a smart city in real time. New distributed big data processing and analysis systems, such as Apache Hadoop, Apache Storm and serverless systems, have also been proposed to manage this large volume of data and provide low-latency, real-time data processing. Beyond data processing, however, there are significant research challenges that need to be studied in order to achieve effective use of these systems. These challenges arise from the fact that each urban data source is governed by its own characteristics, such as sampling rate, data volume, data validity, etc. These challenges include, among others, analysing and modelling the behaviour of urban data in different contexts (during periods of normal city operation or during periods when different events occur in a city) and drawing meaningful conclusions from the data in order to enhance the sustainability of urban data. Finally, the challenge remains to bring out the maximum potential value of the data by using these models and inferences in applications and systems for smart cities. The aim of this thesis is to propose and practical methods to address the aforementioned issues. The first part of the thesis deals with the analysis and modelling of urban data by studying different sources of such data, either by examining the data unilaterally or by combining multiple data sources simultaneously. In this part of the thesis, in order to understand what smart city data actually represent, metrics and techniques are proposed that serve to model their behaviour based on their characteristics, and subsequently allow further combination of multiple urban data sources by creating synergies between them. On the other hand, this requires addressing characteristics such as, for example, data volume, temporal resolution and also data sampling rate. Considering these challenges, in the context of this thesis, in the first part a set of methodologies is proposed that allow smart city authorities and citizens to understand and capture the pulse of the urban environment in real time as well as to draw important conclusions from the data collected in it. The second part of the thesis focuses on the problem of exploiting the inferences derived from urban data as well as their modelling and aims to highlight their value in applications and systems using urban data. The problems studied in this part of the thesis, concerning the appropriate exploitation of inferences from data, are particularly challenging due to the heterogeneity of urban data sources, their different characteristics and the real-time response requirements imposed by the applications, making the development of applications and systems a challenge. In this part of the thesis, a set of new algorithms and related applications that exploit urban data analysis and modelling and the conclusions drawn through them are proposed to address real-world problems within a smart city, such as traffic congestion, enhancing alternative forms of mobility and reducing the environmental footprint of citizens, and ultimately improving the quality of life of its inhabitants and enhancing its sustainability. This part includes graph algorithms for resource allocation problems within transportation networks (with application to any resource allocation and reallocation problem that utilizes a transportation network), routing algorithms within graphs with dynamic weights (allowing the creation of paths adaptive to real-time data), and techniques involving data privacy from users of data analytics services.
show more
