Ph.D. Dissertation Defense: Anousheh Gholami

Friday, August 5, 2022
2:00 p.m.
AVW 1146
Emily Irwin
301 405 0680


NAME:  Anousheh Gholami

Professor John S. Baras (Chair)
Professor Richard J. La
Professor Behtash Babadi
Professor Andre Tits
Professor Bruce L. Golden (Dean’s Representative)

Friday, Aug 5, 2022 at 2:00 pm
AVW 1146

Title:  Resource Allocation in Next-Generation Mobile Networks


The increasing heterogeneity of the mobile network infrastructure together with the explosively growing demand for bandwidth-hungry services with diverse quality of service (QoS) requirements leads to a degradation in the performance of traditional networks. To address this issue in next-generation mobile networks (NGMN), various technologies such as software-defined networking (SDN), network function virtualization (NFV), mobile edge/cloud computing (MEC/MCC), non-terrestrial networks (NTN), and edge ML are essential. Towards this direction, an optimal allocation and management of heterogeneous network resources to achieve the required low latency, energy efficiency, high reliability, enhanced coverage and connectivity, etc. is a key challenge to be solved urgently. In this dissertation, we address four critical and challenging resource allocation problems in NGMN and propose efficient solutions to tackle them.

In the first part, we address the network slice resource provisioning problem in NGMN for delivering a wide range of services promised by 5G systems and beyond, including enhanced mobile broadband (eMBB), ultra-reliable and low latency (URLLC), and massive machine-type communication (mMTC). Network slicing is one of the major solutions needed to meet the differentiated service requirements of NGMN, under one common network infrastructure. Towards robust mobile network slicing, we propose a novel approach for the end-to-end (E2E) resource allocation in a realistic scenario with uncertainty in slices' demands using stochastic programming. The effectiveness of our proposed methodology is validated through simulations.

Despite the significant benefits that network slicing has demonstrated to bring to the management and performance of NGMN, the real-time response required by many emerging delay-sensitive applications, such as autonomous driving, remote health, and smart manufacturing, necessitates the integration of multi-access edge computing (MEC) into network sliding for 5G networks and beyond. To this end, we discuss a novel collaborative cloud-edge-local computation offloading scheme in the next two parts of this dissertation. The first part studies the problem from the perspective of the infrastructure provider and shows the effectiveness of the proposed approach in addressing the rising number of latency-sensitive services and improving energy efficiency which has become a primary concern in NGMN. Moreover, taking into account the perspective of application (higher layer), we propose a novel framework for the optimal reservation of resources by applications, resulting in significant resource savings and reduced cost. The proposed method utilizes application-specific resource coupling relationships modeled using linear regression analysis.  We further improve this approach by using Reinforcement Learning to automatically derive resource coupling functions in dynamic environments.

Enhanced connectivity and coverage are other key objectives of NGMN. In this regard, unmanned aerial vehicles (UAVs) have been extensively utilized to provide wireless connectivity in rural and under-developed areas, enhance network capacity, and provide support for peaks or unexpected surges in user demand. The popularity of UAVs in such scenarios is mainly owing to their fast deployment, cost-efficiency, and superior communication performance resulting from line-of-sight (LoS)-dominated wireless channels. In the fifth part of this dissertation, we formulate the problem of aerial platform resource allocation and traffic routing in multi-UAV relaying systems wherein UAVs are deployed as flying base stations. Our proposed solution is shown to improve the supported traffic with minimum deployment cost.

Moreover, the new breed of intelligent devices and applications such as UAVs, AR/VR, remote health, autonomous vehicles, etc. requires a novel paradigm shift from traditional cloud-based learning to a distributed, low-latency, and reliable ML at the network edge. To this end, Federated Learning (FL) has been proposed as a new learning scheme that enables devices to collaboratively learn a shared model while keeping the training data locally. However, the performance of FL is significantly affected by various security threats such as data and model poisoning attacks. Towards reliable edge learning, in the last part of this dissertation, we propose trust as a metric to measure the trustworthiness of the FL agents and thereby enhance the reliability of FL.

Audience: Graduate  Faculty 

remind we with google calendar


August 2022

31 1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30 31 1 2 3
Submit an Event