A Laser SLAM Framework with Gaussian Process Based Map Objects

Date

2024-9-6

Author

Balcı, Ali Emre

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

74
views

0
downloads

Cite This

In this thesis, a novel Gaussian process-based technique for map representation and robot localization is studied. As accurate detection and interpretation of objects in the environment are crucial for robot navigation and perception, there has been a growing need for improvements in the representation of mapped objects. Addressing this need, we propose a novel Simultaneous Localization and Mapping (SLAM) approach with Gaussian process-based object representations. Our proposed object representation framework enables the SLAM algorithm to function without performing point cloud registration or using grid-based map representations. Our approach also improves the robustness of the robot's data association capabilities by providing a probabilistic measurement-to-object association framework, thanks to the Gaussian process-based representations. The integration of Gaussian processes allows for a more flexible and adaptive modeling of the surroundings, enabling a SLAM method that is efficient in terms of memory and CPU usage, with probabilistic traversability maps as a natural by-product. We studied the capabilities of our approach in various scenarios where several objects of diverse shapes and sizes needed to be mapped. We demonstrate that our method is is able to estimate object shapes and sizes even with low-end range-bearing sensors that produce sparse point clouds.

Subject Keywords

Gaussian Processes, Localization, Mapping

URI

https://hdl.handle.net/11511/111300

Collections

Graduate School of Natural and Applied Sciences, Thesis