Details

Pose Graph SLAM from Scratch

Year: 2026

Term: Winter

Student Name: Mario Pardo

Supervisor: Mark Lanthier

Abstract: Humans are able to navigate and understand new environments quickly and intuitively, combining sensory inputs with spatial memory to create an understanding of their environment and their location within it. Replicating this in machines is a complicated task encompassing various areas of research and is a key challenge in developing autonomous robots and vehicles. Simultaneous Localization and Mapping (SLAM) is a class of algorithms dedicated to just this task, allowing systems to understand the environment in which they operate and where they are within it --- especially critical in scenarios where external positioning sources, such as GPS, are unavailable. Pose Graph SLAM is a popular implementation of SLAM used by most modern state-of-the-art systems, and provides a computationally efficient, yet effective, solution to the localization and mapping problem. This project presents a C++ implementation of a Pose Graph SLAM system developed from first principles. The resulting system demonstrates reliable convergence in pose-graph optimization, validating that a ground-up implementation can manage the underlying mathematical complexity and high-performance requirements of modern robotics.