Date of Award
12-2025
Degree Type
Thesis
Degree Name
M.S.
Degree Program
Computer Science
Department
Computer Science
Major Professor
Dr. James Wagner
Second Advisor
Dr. Abdullah Al Redwan Newaz
Third Advisor
Dr. Abdullah Yasin Nur
Abstract
Conventional graph-based Simultaneous Localization and Mapping (SLAM) systems rely pri marily on in-memory data structures to manage LiDAR and odometry information. Although this design ensures rapid access during map construction, it poses significant challenges in large-scale applications, including excessive memory utilization and the risk of data loss dur ing system interruptions. These limitations motivate the exploration of persistent and scalable data management approaches for SLAM pipelines.
This thesis presents a Graph SLAM framework integrated with a relational database system, incorporating a persistent storage layer designed to enhance memory efficiency and system re liability through a carefully structured database design for managing pose and scan data. The central hypothesis is that the structured query capabilities, indexing mechanisms, configurable parameter management, and transactional guarantees of relational databases enable scalable and fault-tolerant data handling without compromising runtime performance significantly. The framework is implemented using three relational database systems—PostgreSQL, MonetDB, and SQLite3—and benchmarked against a traditional baseline in-memory through systematic simulation and ablation studies. Evaluation metrics include average and peak memory con sumption as well as total execution time for data ingestion and map generation.
The experimental evaluations are conducted using datasets of varying scale and complexity, including Intel datasets, marsyard dataset, and large-scale KITTI datasets. Simulation results demonstrate that database integration considerably reduces memory consumption by offloading intermediate pose and scan data from volatile memory to persistent storage, while maintaining comparable execution times.
Overall, the findings confirm that embeddingrelational databases within the SLAMdata pipeline enhances scalability, persistence, and memory efficiency. The proposed framework establishes a foundation for further research toward distributed and multi-robot SLAM systems driven by database-centric architectures.
Recommended Citation
Khatri, Abhishek, "Toward Resource-Efficient Graph-Based Simultaneous Localization and Mapping (SLAM) through Relational Database Systems" (2025). University of New Orleans Theses and Dissertations. 3313.
https://scholarworks.uno.edu/td/3313
Rights
The University of New Orleans and its agents retain the non-exclusive license to archive and make accessible this dissertation or thesis in whole or in part in all forms of media, now or hereafter known. The author retains all other ownership rights to the copyright of the thesis or dissertation.