Electronic Theses and Dissertations Archive

Date

2025

Document Type

Thesis

Degree Name

Master of Science

Department

Computer Science

Committee Chair

Lan Wang

Committee Member

Christos Papadopoulos

Committee Member

Xiaolei Huang

Abstract

Named Data Networking (NDN) is a promising data-centric Internet architecture that focuses on data rather than the hosts. By design, NDN supports multicast communication, enabling data distribution to multiple recipients. This behavior is beneficial for modern applications such as video conferencing, online gaming, and vehicular networks. However, when multiple NDN nodes access the same data on a multicast link, they may send duplicate Interest and Data packets. It causes bandwidth wastage and latency in the network. Our research employs Actor Critic network which is a type of Reinforcement Learning, to optimize the wait time each node should observe before sending Interest or Data packets in a single-hop network. The Actor Critic network enables nodes to learn and adjust wait times based on real-time network conditions. This wait time minimizes unnecessary duplications. In our experiments, we have compared our method named Reinforcement Learning-based Adaptive Duplicate Suppression (RL-ADS) with Adaptive Duplicate Suppression (ADS) and current NDN Forwarding Daemon (NFD) with no suppression. Our experiments demonstrate a substantial reduction of more than 60% in duplicate packet transmissions compared to the current NFD with no suppression, and about a 17% decrease compared to ADS.

Library Comment

Dissertation or thesis originally submitted to ProQuest.

Notes

Embargoed until 10-30-2026

Download

Available for download on Friday, October 30, 2026

Share

COinS