SLAM Drift and Localization Error Reduction with RTAB-Map tools
Ingren, Siina (2023)
Ingren, Siina
2023
Tieto- ja sähkötekniikan kandidaattiohjelma - Bachelor's Programme in Computing and Electrical Engineering
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2023-05-03
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202305035172
https://urn.fi/URN:NBN:fi:tuni-202305035172
Tiivistelmä
This bachelor's thesis investigates techniques for reducing drift and localization errors in a 3D reconstruction made with the SLAM (simultaneous localization and mapping) technique. The errors are mainly present in corridors, which will be called the upper and canteen hallway. The error reduction methods will be performed with RTAB-Map tools Database and Odometry viewers. In addition, relevant aspects of simultaneous localization and mapping are briefly introduced.
The RTAB-Map database modified in this research is of a night-time query image sequence of Tampere University Hervanta campus’ main foyer. This sequence is part of the TAU-Indoors dataset, which was introduced in the study “TAU-Indoors Dataset for Visual and LiDAR Place Recognition” conducted by Tampere University Vision Group. The dataset contains RGB-D images and LiDAR scans. The RTAB-Map 3D reconstructions introduced errors which prevented utilizing the models for place recognition studies.
To address the errors, this thesis employs error reduction methods such as manual correction of map constraints, adding loop closures to specific map points and parameter tuning of relevant algorithms. Additionally, different graph optimization and odometry methods are experimented with. The utilized methods are divided into graph optimization and odometry parts. For a full SLAM approach, ROS (Robot Operating System) needs to be employed because the TAU-indoors dataset’s data is saved as rosbag files.
The Database viewer tool allows for manually modifying map constraints, experimenting with different graph optimization methods, and tuning certain parameters. The upper hallway’s drift was fixed by adding loop closures to specific map points and restricting the map into three degrees of freedom. The localization errors in the canteen hallway were reduced by refining map constraints. Changing graph optimization methods had subsidiary effect.
The Odometry viewer tool allows for testing different odometry methods and the related parameters. In this work the parameter tuning was experimental. As a results the canteen hallway’s localization errors were reduced but the upper hallway was left unimproved.
The RTAB-Map database modified in this research is of a night-time query image sequence of Tampere University Hervanta campus’ main foyer. This sequence is part of the TAU-Indoors dataset, which was introduced in the study “TAU-Indoors Dataset for Visual and LiDAR Place Recognition” conducted by Tampere University Vision Group. The dataset contains RGB-D images and LiDAR scans. The RTAB-Map 3D reconstructions introduced errors which prevented utilizing the models for place recognition studies.
To address the errors, this thesis employs error reduction methods such as manual correction of map constraints, adding loop closures to specific map points and parameter tuning of relevant algorithms. Additionally, different graph optimization and odometry methods are experimented with. The utilized methods are divided into graph optimization and odometry parts. For a full SLAM approach, ROS (Robot Operating System) needs to be employed because the TAU-indoors dataset’s data is saved as rosbag files.
The Database viewer tool allows for manually modifying map constraints, experimenting with different graph optimization methods, and tuning certain parameters. The upper hallway’s drift was fixed by adding loop closures to specific map points and restricting the map into three degrees of freedom. The localization errors in the canteen hallway were reduced by refining map constraints. Changing graph optimization methods had subsidiary effect.
The Odometry viewer tool allows for testing different odometry methods and the related parameters. In this work the parameter tuning was experimental. As a results the canteen hallway’s localization errors were reduced but the upper hallway was left unimproved.
Kokoelmat
- Kandidaatintutkielmat [7047]