Author ORCID Identifier

https://orcid.org/0000-0001-7474-2689

Date Available

4-28-2022

Year of Publication

2022

Degree Name

Doctor of Philosophy (PhD)

Document Type

Doctoral Dissertation

College

Engineering

Department/School/Program

Computer Science

First Advisor

Dr. Hana Khamfroush

Abstract

Smart cities are an ambitious vision for future urban environments. The ultimate aim of smart cities is to use modern technology to optimize city resources and operations while improving overall quality-of-life of its citizens. Realizing this ambitious vision will require embracing advancements in information communication technology, data analysis, and other technologies. Because smart cities naturally produce vast amounts of data, recent artificial intelligence (AI) techniques are of interest due to their ability to transform raw data into insightful knowledge to inform decisions (e.g., using live road traffic data to control traffic lights based on current traffic conditions). However, training and providing these AI applications is non-trivial and will require sufficient computing resources. Traditionally, cloud computing infrastructure have been used to process computationally intensive tasks; however, due to the time-sensitivity of many of these smart city applications, novel computing hardware/technologies are required. The recent advent of edge computing provides a promising computing infrastructure to support the needs of the smart cities of tomorrow. Edge computing pushes compute resources close to end users to provide reduced latency and improved scalability — making it a viable candidate to support smart cities. However, it comes with hardware limitations that are necessary to consider.

This thesis explores the use of the edge computing paradigm for smart city applications and how to make efficient, smart decisions related to their available resources. This is done while considering the quality-of-service provided to end users. This work can be seen as four parts. First, this work touches on how to optimally place and serve AI-based applications on edge computing infrastructure to maximize quality-of-service to end users. This is cast as an optimization problem and solved with efficient algorithms that approximate the optimal solution. Second, this work investigates the applicability of compression techniques to reduce offloading costs for AI-based applications in edge computing systems. Finally, this thesis then demonstrate how edge computing can support AI-based solutions for smart city applications, namely, smart energy and smart traffic. These applications are approached using the recent paradigm of federated learning.

The contributions of this thesis include the design of novel algorithms and system design strategies for placement and scheduling of AI-based services on edge computing systems, formal formulation for trade-offs between delivered AI model performance and latency, compression for offloading decisions for communication reductions, and evaluation of federated learning-based approaches for smart city applications.

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2022.82

Funding Information

This research was supported by the National Science Foundation under grant no. CSR-1948387 and Cisco Systems Inc. under research grant number 1215519250.

Share

COinS