Semester

Spring

Date of Graduation

2024

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Lane Department of Computer Science and Electrical Engineering

Committee Chair

Matthew C. Valenti, Ph.D.

Committee Member

Nasser M. Nasrabadi, Ph.D.

Committee Member

Brian Woerner, Ph.D.

Abstract

Providing broadband access to rural communities continues to be an important societal problem whose solution would help to break down the digital divide. While 5G wireless networks may be used for rural broadband, a key challenge is the placement of base stations, which is exacerbated by the use of high frequencies in the millimeter-wave band. Such technology requires an unobstructed line of sight, demanding meticulous planning of the number, height, and location of base stations for optimal coverage. Conventional methods, such as ray-tracing to simulate signal propagation across varied terrain, are computational costly and not feasible for vast coverage areas. These constraints pose significant hurdles for extensive network deployment, especially in rural U.S. areas with complex topographies like mountainous regions.

In this thesis, we investigate the fusion of machine learning algorithms with Digital Terrain Elevation Data (DTED) and leverage MATLAB's analytical capabilities to identify optimal locations for base stations in order to enhance line-of-sight (LoS) coverage in rural areas. Preston County, West Virginia, serves as a compelling case study for this research. The choice of this location is based on two main factors: its proximity to West Virginia University (WVU) provides logistical advantages for field studies, and its characteristics of sparse population and rugged mountainous terrain present typical obstacles encountered in rural telecommunications.

The methodology begins by spatially sampling the region, considering 100 potential sites for base stations. An exhaustive search technique is then deployed to identify an optimal subset of locations that could significantly enhance coverage. Although this approach is resource-intensive, it establishes a reliable benchmark for assessing the efficacy of machine learning methods applied later in the study. Subsequently, the optimization of base station placement is obtained through the application of the Differential Evolution algorithm. This method demonstrates its effectiveness for the specific challenge by aligning with results from the exhaustive search, thereby validating its utility. The research culminates with an exploration into the use of Graph Neural Networks (GNN) for further optimizing the problem. This innovative approach seeks to leverage the spatial relationships and network topology inherent in base station placement, offering a novel perspective on optimizing telecommunications infrastructure within complex terrains.

Share

COinS