Multisensory integration for topological indoor localization of mobile robots in complex symmetrical environments

Indoor localization is essential for robotic navigation by using different sensors on board. Specifically, visual localization with a single camera is a great challenge in highly symmetric environments (e.g. offices, hospitals or residences), where appearance patterns are repetitive and captures fro...

Descripción completa

Detalles Bibliográficos
Autores: Lafuente Arroyo, Sergio|||0000-0002-4841-2501, Maldonado Bascón, Saturnino|||0000-0001-6472-5359, Delgado Mena, Diego|||0000-0002-5672-1865, Gutiérrez Álvarez, Carlos|||0000-0002-5624-0076, Acevedo Rodríguez, Francisco Javier|||0000-0002-4727-1575
Tipo de recurso: artículo
Fecha de publicación:2024
País:España
Institución:Universidad de Alcalá (UAH)
Repositorio:e_Buah Biblioteca Digital Universidad de Alcalá
Idioma:inglés
OAI Identifier:oai:ebuah.uah.es:10017/66985
Acceso en línea:http://hdl.handle.net/10017/66985
https://dx.doi.org/10.1016/j.eswa.2023.122561
Access Level:acceso abierto
Palabra clave:Indoor global localization
Autonomous mobile robot
Topological map
Complex symmetrical environments
Telecomunicaciones
Telecommunication
Descripción
Sumario:Indoor localization is essential for robotic navigation by using different sensors on board. Specifically, visual localization with a single camera is a great challenge in highly symmetric environments (e.g. offices, hospitals or residences), where appearance patterns are repetitive and captures from different locations provide very similar images. To overcome this issue, in this paper, we present a method that integrates multisensory information from an RGB-D camera, a LiDAR and motor encoders. Our approach simultaneously utilizes spatial consistency from a reference topological map and temporal consistency from time-series observations. Inspired by human cognitive perception, we define a two layered topological architecture that encompasses both coarse information of object distributions and structural information with some metric references. Categories of common objects in the environments, such as fire extinguishers or doors, are used as natural beacons. We evaluated our approach in two real-world buildings based on a multi-aisle structure with corridors of very similar appearance. Results demonstrated accurate localization despite the high level of symmetry of the scenario, and how ambiguity was significantly reduced as the agent progressed along its trajectory.