In this research project, I conducted a comparative analysis of two distinct approaches for Simultaneous Localization and Mapping (SLAM) - ORB-SLAM (Visual-based) and Cartographer (LiDAR-based). These SLAM algorithms are crucial for a wide range of applications, from robotics and autonomous vehicles to augmented and virtual reality.
Our analysis primarily focused on evaluating the performance of these algorithms in real-world environments, considering factors like accuracy, speed, and robustness. We applied these algorithms to a unique dataset collected from the NuANCE semi-autonomous vehicle, which included 3D LiDAR data, stereo-camera data, raw IMU data, and GPS data.
The results of our study revealed some significant insights. Cartographer demonstrated high accuracy and robustness in mapping and localization, but it comes with high computational and memory costs. In contrast, ORB-SLAM showed faster processing times and lower resource demands, but it may be less accurate and robust in challenging scenarios.
Our work highlights the importance of choosing the right SLAM algorithm for specific applications, depending on the requirements and constraints. This research contributes to the growing field of SLAM, providing valuable insights into the trade-offs between LiDAR-based and visual-based approaches.