RadarLoc: Learning to Relocalize in FMCW Radar
| Autor: | Pedro P. B. de Gusmao, Niki Trigoni, Wei Wang, Andrew Markham, Bo Yang |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
FOS: Computer and information sciences
0209 industrial biotechnology Computer Science - Artificial Intelligence Computer science ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION 02 engineering and technology computer.software_genre Field (computer science) law.invention Computer Science - Robotics 020901 industrial engineering & automation law Margin (machine learning) Computer vision Radar Artificial neural network business.industry Deep learning Location awareness Robotics Continuous-wave radar Artificial Intelligence (cs.AI) Artificial intelligence business computer Robotics (cs.RO) |
| Zdroj: | ICRA |
| DOI: | 10.48550/arxiv.2103.11562 |
| Popis: | Relocalization is a fundamental task in the field of robotics and computer vision. There is considerable work in the field of deep camera relocalization, which directly estimates poses from raw images. However, learning-based methods have not yet been applied to the radar sensory data. In this work, we investigate how to exploit deep learning to predict global poses from Emerging Frequency-Modulated Continuous Wave (FMCW) radar scans. Specifically, we propose a novel end-to-end neural network with self-attention, termed RadarLoc, which is able to estimate 6-DoF global poses directly. We also propose to improve the localization performance by utilizing geometric constraints between radar scans. We validate our approach on the recently released challenging outdoor dataset Oxford Radar RobotCar. Comprehensive experiments demonstrate that the proposed method outperforms radar-based localization and deep camera relocalization methods by a significant margin. Comment: To appear in ICRA 2021 |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|