Practical Insights on Automotive SLAM in Urban Environments
[ 1 ] Instytut Robotyki i Inteligencji Maszynowej, Wydział Automatyki, Robotyki i Elektrotechniki, Politechnika Poznańska | [ P ] pracownik
[2.2] Automatyka, elektronika, elektrotechnika i technologie kosmiczne
2023
rozdział w monografii naukowej
angielski
- SLAM
- LiDAR
- vision
- GNSS
- factor graph
- evaluation
EN This chapter tackles the issues of simultaneous localization and mapping (SLAM) using laser scanners or vision as a viable alternative to the accurate modes of satellitebased localization, which are popular and easy to implement with modern technology but might fail in many urban scenarios. This chapter considers two state-of-the-art localization algorithms, LOAM and ORB-SLAM3 that use the optimization-based formulation of SLAM and utilize laser and vision sensing, respectively. The focus is on the practical aspects of localization and the accuracy of the obtained trajectories. It contributes to a series of experiments conducted using an electric car equipped with a carefully calibrated multisensory setup with a 3D laser scanner, camera, and a smartphone for collecting the exteroceptive measurements. Results of applying the two different SLAM algorithms to the data sequences collected with the vehicle-based multisensory setup clearly demonstrate that not only the expensive laser sensors but also monocular vision, including the commodity smartphone camera, can be used to obtain off-line reasonably accurate vehicle trajectories in an urban environment.
127 - 142
CC BY (uznanie autorstwa)
witryna wydawcy
ostateczna wersja opublikowana
10.05.2023
w momencie opublikowania
publiczny
5