Abstract

Most computer networks require continuous end-to-end connectivity between sending and destination nodes for data transfer. However, in certain scenarios—such as disaster zones, geographically remote areas with intermittent or no connectivity, or regions facing peak network loads—such continuous communication paths may be unavailable. A notable example of this challenge occurred during Hurricane Milton in 2024 and the Los Angeles wildfires in 2025, where Starlink provided free satellite internet to support disaster relief efforts and enable communication in affected areas. However, while offering satellite services at no cost can be effective in emergencies, it may not always be feasible on a global scale. In large-scale disasters, the associated costs and logistical challenges could present significant obstacles. These critical situations necessitate a shift toward infrastructure-less communication, making opportunistic networks (OppNets) a viable solution. OppNets have already been deployed across various domains, including smart cities, underwater communication, wildlife monitoring, and interplanetary missions. Notable applications include tourist tracking through LASSO in 2020, oceanic data exchange via the SWIM model in 2016, and NASA's use of the Saratoga routing protocol for interplanetary communication since 2004. Their adaptability demonstrates the potential for infrastructure-free communication in diverse environments. Researchers have developed numerous OppNets forwarding algorithms to help network nodes select the next node for data forwarding based on specific criteria. In this work, we specifically focus on social-aware, interest- aware, and power-aware forwarding algorithms, as they involve nodes interested in the content while conserving their valuable power resources. The aim of such algorithms is to maximize successful message delivery to the intended users while minimizing the number of replicas generated along the way and preserving power resources. This study particularly investigates the performance of key algorithms, namely Interest-Aware PeopleRank (IPeR) and Power-Aware IPeR (PIPeR), using Epidemic and PeopleRank(PeR) as benchmark algorithms. While these algorithms have demonstrated strong performance in confined environments such as conferences, shopping malls, and university campuses, they have predominantly been evaluated under pedestrian mobility models. Their application in subway environments—characterized by hybrid mobility involving both pedestrian movement and subway train rides—remains unexplored. Subway mobility scenarios are particularly relevant for contexts such as emergency response during crises, where fixed network infrastructure is limited or unavailable. To address this gap, we evaluate the IPeR, and PIPeR algorithms in subway stations, where the mobility model integrates subway train rides with pedestrian movement, using the AnyLogic simulator to accurately model passenger flows. Our results reveal that PIPeR demonstrates the best performance, achieving a 64% increase in F-measure and a 63% reduction in delay compared to the pedestrian environment, albeit with increased power consumption and cost. Additionally, key zones of vigorous content dissemination unique to this hybrid environment are identified. Building on these findings, we propose five enhanced variants of the PIPeR algorithm, designed to address the unique challenges of subway environments and any other environments of similar mobility patterns aiming to improve the

ii

algorithm's overall efficiency. The proposed variants effectively overcome the limitations of the original PIPeR algorithm in the hybrid mobility environment, achieving an impressive 83% reduction in power consumption and a 38% decrease in cost, with a trade-off of a 20% reduction in F-measure. The best-performing variant is then identified and rigorously tested under varying conditions, including various interest distributions, battery distributions, user densities, and message volumes per user. Notably, this variant excels even in challenging scenarios such as those involving a majority of uninterested nodes and a lack of intermediate forwarders. Finally, we quantified the environmental impact of the proposed PIPeR variants by measuring their carbon footprint. The results revealed an extraordinary reduction, with the variants achieving a 98% decrease compared to the original PIPeR algorithm in AnyLogic and an exceptional 99% reduction relative to the PeopleRank algorithm. These findings highlight the pivotal role of power-aware enhancements in fostering green, sustainable, and energy-efficient opportunistic networks.

School

School of Sciences and Engineering

Department

Computer Science & Engineering Department

Degree Name

MS in Computer Science

Graduation Date

Spring 6-18-2025

Submission Date

5-27-2025

First Advisor

Dr. Sherif Aly

Second Advisor

Dr. Soumaia Al Ayyat

Committee Member 1

Dr. Amr El Mougy

Committee Member 2

Dr. Amal El Nahhas

Extent

145 p.

Document Type

Master's Thesis

Institutional Review Board (IRB) Approval

Not necessary for this item

Download

Share

COinS