2.5D object modeling using Gaussian processes for robotic mapping and navigation

Date

2024-9-04

Author

Toraman, Erdem

Metadata

Show full item record

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

Item Usage Stats

158
views

0
downloads

Cite This

In this thesis, we address the challenge of mapping and planning for robotic navigation using LiDAR-based measurements, which provide precise distance and geometric information about the surroundings. We explore a probabilistic mapping framework based on Gaussian processes (GPs) to leverage the available information. We develop a two-and-a-half-dimensional (2.5D) object modeling approach for map construction that effectively uses the information provided by three-dimensional (3D) point cloud measurements. The resulting method provides a compact volumetric representation using contours and heights, ensuring robust shape learning even with sparse data. Additionally, we propose a reactive path planning approach that exploits the spatial correlation information provided by the simultaneously constructed GP-based map. This approach incorporates the dynamic window approach for informed decision-making in uncertain environments, making it well-suited for navigation and exploration. The effectiveness of the proposed methods is demonstrated through simulations, illustrating the benefits of the 2.5D GP-based object modeling and the path planning approaches.

Subject Keywords

Object modeling, Robotic mapping, Gaussian processes, Path planning

URI

https://hdl.handle.net/11511/111403

Collections

Graduate School of Natural and Applied Sciences, Thesis