Institutional-Repository, University of Moratuwa
Welcome to the University of Moratuwa Digital Repository, which houses postgraduate theses and dissertations, research articles presented at conferences by faculties and departments, university-published journal articles and research publications authored by academic staff. This online repository stores, preserves and distributes the University's scholarly work. This service allows University members to share their research with a larger audience.
 Research Publications |  Thesis & Dissertation |  E- Books |
 UoM Journal Publications |  UoM Conference Proceedings |  Articles published in Scimago's Q1 journals |  UoM Research Reports |  Other Articles authored by UoM staff |
Recent Submissions
item: Thesis-Full-text
A Multi-stakeholder roadmap for 5G deployment in Sri Lanka : regulatory, operator, and enterprise user perspectives
(2026) Madanayake, NM; Dias, D
The fifth generation of mobile technology (5G) is widely recognized as a transformative platform for digital economies, offering ultra-low latency, high device density, and enhanced mobile broadband. While early adopters such as South Korea, China, and the United States have demonstrated significant socio-economic benefits, developing nations continue to face barriers related to spectrum scarcity, infrastructure costs, and regulatory uncertainty. Sri Lanka represents a case where national digital ambitions, outlined in strategies for 2030, are not yet matched by an operational roadmap for 5G deployment. This study examines Sri Lanka’s readiness to adopt 5G through a multi-stakeholder lens, focusing on regulators, operators, infrastructure providers, and industry users. Using a qualitative-dominant mixed-methods design, the research combines expert interviews with global benchmarks and secondary data. The analysis is organized around five key themes: spectrum allocation, NSA to SA migration strategies, infrastructure and power cost challenges, policy and regulatory incentives, and enterprise adoption models. Findings indicate that while operators are technically prepared to launch NSA 5G, progress is constrained by fragmented spectrum, limited fiber backhaul, high energy tariffs, and unclear regulatory processes. Industry adoption is further slowed by device affordability issues and weak enterprise demand. At the same time, lessons from regional peers highlight the importance of phased strategies, transparent spectrum frameworks, and coordinated infrastructure sharing. Drawing on these insights, the thesis proposes a four-phase roadmap of Enablement, Optimization, Innovation, and Reinvention, designed to align technical, regulatory, and market priorities. The roadmap emphasizes spectrum harmonization, infrastructure cost-sharing, regulatory reform, and targeted enterprise pilots as prerequisites for sustainable rollout. The study contributes by offering a context-specific model for 5G deployment in a developing economy, balancing international best practices with local constraints. Its outcomes are intended to guide policymakers, operators, and industry stakeholders in ensuring that 5G supports national goals of economic resilience, digital inclusion, and long-term competitiveness..
item: Thesis-Full-text
Robust regression as a benchmark for regression based federated learning
(2025) Karunarathna, JHSP; Premaratne, U
This research evaluates the effectiveness of robust regression techniques, specifically Theil-Sen and Repeated Median Regression (RMR) as benchmarks for regression- based Federated Learning (FL) across a range of simulated sensor configurations. FL is a privacy-preserving paradigm that enables collaborative model training across distributed nodes without centralizing the raw data. Standard regression techniques such as linear regression, which is not robust against outliers and heterogeneity are commonly employed in real world FL environments. In contrast, RMR demonstrates the highest resilience and lowest error rates, with Theil-Sen also showing strong per- formance as a more computationally efficient alternative. RMR achieves the overall highest performance gains, with the best efficiency improvement with respect to the linear federated regression. The findings of the study support the use of robust regres- sion as a dependable alternative in distributed, real-world sensor networks.
item: Thesis-Full-text
Implementation of deep neural networks on embedded systems for real-time hand gesture recognition using a 24GHz Doppler radar
(2025) Ijaz, MAM; Edussooriya, C; Samarasinghe, K
The rapid proliferation of 5G networks has led to increased demand for efficient resource allocation at the edge cloud to support latency-sensitive and compute-intensive applications. This thesis presents an optimal resource allocation framework for 5G edge cloud environments, leveraging a closed-loop architecture that integrates demand prediction, intelligent scheduling, and load balancing. The proposed framework employs machine learning-based traffic prediction models, such as Prophet, to forecast real-time demand and optimize resource provisioning dynamically. The framework consists of key components, including a central cloud resource orchestrator, edge cloud nodes, a traffic prediction module, and a resource orchestrator for cluster-wide task scheduling and load balancing. A feedback mechanism ensures continuous updates on resource utilization, enabling adaptive scaling and efficient workload distribution. Additionally, an optimization engine is incorporated to balance energy efficiency and privacy constraints while maintaining service quality. This research is implemented and evaluated using CloudSim Plus, a simulation framework for cloud computing environments and NS3, a discrete-event network simulator for internet systems. The effectiveness of the proposed approach is assessed based on metrics such as resource utilization, energy consumption, and service latency. Experimental results demonstrate that the closed-loop framework enhances the adaptability of 5G edge clouds, reducing underutilization and overprovisioning while improving overall system efficiency. The findings of this study contribute to the advancement of intelligent resource management in 5G edge computing, paving the way for scalable and autonomous cloud-native network operations. Future work may explore the integration of reinforcement learning-based optimization techniques to further enhance decision-making in dynamic edge cloud environments.
item: Thesis-Full-text
Novel resource allocation technique for demand aware 5G edge cloud dimensioning
(2025) Sivanujan, Y; Sumanasena, A; Hemachandra, K
The rapid proliferation of 5G networks has led to increased demand for efficient resource allocation at the edge cloud to support latency-sensitive and compute-intensive applications. This thesis presents an optimal resource allocation framework for 5G edge cloud environments, leveraging a closed-loop architecture that integrates demand prediction, intelligent scheduling, and load balancing. The proposed framework employs machine learning-based traffic prediction models, such as Prophet, to forecast real-time demand and optimize resource provisioning dynamically. The framework consists of key components, including a central cloud resource orchestrator, edge cloud nodes, a traffic prediction module, and a resource orchestrator for cluster-wide task scheduling and load balancing. A feedback mechanism ensures continuous updates on resource utilization, enabling adaptive scaling and efficient workload distribution. Additionally, an optimization engine is incorporated to balance energy efficiency and privacy constraints while maintaining service quality. This research is implemented and evaluated using CloudSim Plus, a simulation framework for cloud computing environments and NS3, a discrete-event network simulator for internet systems. The effectiveness of the proposed approach is assessed based on metrics such as resource utilization, energy consumption, and service latency. Experimental results demonstrate that the closed-loop framework enhances the adaptability of 5G edge clouds, reducing underutilization and overprovisioning while improving overall system efficiency. The findings of this study contribute to the advancement of intelligent resource management in 5G edge computing, paving the way for scalable and autonomous cloud-native network operations. Future work may explore the integration of reinforcement learning-based optimization techniques to further enhance decision-making in dynamic edge cloud environments.
item: Thesis-Abstract
Service-based 5G network dimensioning framework
(2025) Keeragala, KAVCK; Sumanasena, A; Hemachandra, K
The fifth generation (5G) mobile networks are designed to deliver multiple services with varying requirements and are mostly delivered through a single New Radio (NR) carrier in an early 5G network. However, designing a network to serve multiple services will not be simple and will drastically vary depending on the user demand and the business requirements of the operator. This thesis introduces a service-based 5G network dimensioning framework designed to address the multifaceted requirements of such 5G networks that deliver multiple services through the same NR carrier. Unlike previous studies that rely solely on theoretical models and are based on population census data which do not provide an accurate distribution of users, this research uses real user traffic data and Minimization of Drive Test (MDT) based serving locations to create an accurate and very granular traffic map with user geolocation to service areas of dimension 100m x 100m, offering a practical approach to accurate network dimensioning. The thesis clearly differentiates between the Mobile and Fixed service throughput requirements and how network operators should consider both services in dimensioning a network considering the each service users separately, while previous studies mainly focused on coverage and capacity dimensioning for a common cell edge throughput. The thesis elaborates how the increase in user density due to demographic variation or capture rate impacts the utilization of 5G NR resources and impacts dimensioning parameters contributing to the increase of required Next Generation Node B (gNodeB) to cover certain areas such as hot-spots. Further, the thesis briefly goes through pre-dimensioning process before dimensioning of gNodeB cells, where crucial decision for spectrum selection is done giving a holistic view of the dimensioning process, which are not commonly addressed in existing literature. This work is particularly beneficial for operators in emerging markets, providing a practical guide to address challenges in early stage 5G deployments with mixed services on the same carrier.