Long-term indoor localization in floor plans using semantic cues : for robot autonomy in human-oriented environments
- Zimmerman, Nicky
- Hormann , Kai (Degree supervisor)
- Giusti, Alessandro (Degree committee member)
- Guzzi, Jérôme (Degree committee member)
- Benini, Luca (Degree committee member)
- 2024
PhD: Università della Svizzera italiana
English
Localization in a given map is an essential capability of most autonomous robots, and robust long-term localization is crucial in the case of service robots. This is a challenging task, especially in a dynamic, human-occupied environment, and it motivates the use of sparse map representations containing structural elements that remain constant over time. Floor plans, in particular, are a sparse map representation that is often readily-available without the additional cost and effort of sensor-based mapping. We present different strategies for achieving robust long-term localization in floor plans, by taking inspiration from the way humans navigate in indoor environments. We start with a classical range sensor-based particle filter framework and augment it by integrating textual in- formation. We then improve localization by considering a variety of semantic cues and propose a 3D metric-semantic map representation that enriches floor plans with semantic information. We address the challenge of localization on resource-constrained platforms and verify that our semantic localization approach is suitable for a variety of robotic platforms. Finally, we explore the benefits of collaborative localization, where robots in a team assist each other in improving the pose estimation.
- Collections
- Language
-
- English
- Classification
- Computer science and technology
- License
- License undefined
- Open access status
- green
- Identifiers
-
- NDP-USI 2024INF013
- ARK ark:/12658/srd1329703
- URN urn:nbn:ch:rero-006-122451
- Persistent URL
- https://n2t.net/ark:/12658/srd1329703
Statistics
Document views: 37
File downloads:
- 2024INF013.pdf: 41