Feasibility Study of Smart City Applications

智慧城市應用的可行性研究

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date19 Jun 2023

Abstract

Smart city construction is a great vision for global urban development. As the backbone of smart city, Internet of Things (IoT) technologies enable the interconnection between people and nature, things and environment. In particular, the emerging Low Power Wide Area (LPWA) technologies (e.g., Long Range (LoRa), Narrowband Internet of Things (NB-IoT), Sigfox) facilitate the realization of large-scale smart city blueprint, given the advantages of low power consumption, low cost, and wide coverage. Based on these IoT technologies, smart city applications integrating advanced Artificial Intelligent (AI) algorithms have been explored. However, the feasibility of diverse smart solutions in the long-term development still needs to be studied, such as practicality, scalability, standardization, coexistence of multiple services, security, etc. Without prospective feasibility analysis, the problems of performance degradation (e.g., high Packet Loss Rate (PLR), high latency, etc.) or even service downtime would be exposed as network scale expands and applications grow explosively. Therefore, this paper focuses on feasibility study of application development and management in selected smart city domains to promote the coordinated and sustainable development of smart cities, including smart building, smart transportation, smart evaluation, smart harmonization, and smart security. The contributions are summarized as follows:

1. A multi-leak detection strategy, namely Multi-Classification Multi-Leak Detection (MC-MLD) is developed for Water Supply Network (WSN), which shows its feasibility in smart building. In the MC-MLD, a novel Adaptive Kernel (AK) scheme and Multi-Classification (MC) scheme are designed to facilitate efficient adaptation to different scenarios and modifications of WSN, which is applicable for scalable smart WSN. The experimental results showed the developed MC-MLD achieved 96% accuracy, 96% sensitivity, 95% specificity, and about 5 s detection time.

2. The feasibility of a new driver stress recognition scheme using Ensembled Multiscale Classifier (EMC) is analyzed, aiming to improve the efficiency and safety of traffic management in smart transportation. In EMC, the driver stress level is classified into three categories, including Stress Level 1 (SL1), Stress Level 2 (SL2), and Stress Level 3 (SL3). Fiducial features extracted from Electrocardiogram (ECG) signals and non-fiducial features extracted from the transformed ECG signals are analyzed using Neural Network with Backpropagation (NNBP) and the 1-D Convolutional Neural Network (1-D CNN) respectively. An ensemble decision-making layer is added to coordinate the outputs of NNBP and 1-D CNN and provides probabilistic prediction of drivers’ stress level. The experimental results reveal the proposed scheme achieves 95.9% detection accuracy and high practicability.

3. A new performance evaluation indicator based on IEEE 2668 standard, namely Satisfaction IoT Index (SDex) is proposed in smart evaluation field, which guides developers in choosing the best IoT configuration solution to maximize its performance. Given diverse Quality of Service (QoS) requirements of different types of applications, SDex is established from the user-centric aspect to evaluate the user’s satisfaction degree for the specific application by measuring difference between practical and required performances. In addition, SDex-based Fuzzy Comprehensive Evaluation (SDex-FCE) scheme is developed to grade the SDex score taking attribution uncertainty at adjacent levels into consideration. Finally, the best configuration solution can be identified as the one with the highest SDex score. Experimental results in case studies (i.e., smart metering and smart healthcare) illustrated the effectiveness of proposed SDex and SDex-FCE scheme.

4. A Priority-Aware Resource Allocation (PA-RA) strategy is developed to effectively coordinate multiple applications to facilitate smart IoT harmonization. LoRaWAN is selected as case study. There are three primary types of application services, namely safety services, control services, and monitoring services. Given the varying levels of criticality among these services, the PA-RA scheme allocates Spreading Factors (SFs) to end devices based on the priority parameter with the highest precedence. Additionally, to provide a comprehensive and quantitative assessment of coordination capabilities, the Harmonization Index (HDex) is introduced. Afterwards, Genetic Algorithm (GA)-based optimization approach is employed to identify the optimal service criticality parameters. Experimental results demonstrate that, compared to the conventional Adaptive Data Rate (ADR) scheme, the proposed PA-RA strategy achieves a 50% capacity improvement while maintaining the demands of each application.

5. A Directed Acyclic Graph (DAG)-based IoT framework is proposed to further enhance security of smart city applications. To avoid single point of failure in IoT system, distributed gateways/base stations and networks servers are deployed to record data transmission in tangle network. A secure Data Ledger (DL) using digital signature algorithm and symmetric-key encryption is designed to defend against malicious gateway attacks and eavesdropping. The results demonstrated that the proposed DAG-based IoT system is an effective solution to ensure the security and efficiency of smart city applications.