Kidnapped radar: topological radar localisation using rotationally-invariant metric learning
This paper presents a system for robust, large-scale topological localisation using Frequency-Modulated ContinuousWave (FMCW) scanning radar. We learn a metric space for embedding polar radar scans using CNN and NetVLAD architectures traditionally applied to the visual domain. However, we tailor the...
| Main Authors: | , , , , |
|---|---|
| 格式: | Conference item |
| 語言: | English |
| 出版: |
IEEE Xplore
2020
|
| _version_ | 1826281752404951040 |
|---|---|
| author | Gadd, M Sǎftescu, Ş De Martini, D Barnes, D Newman, P |
| author_facet | Gadd, M Sǎftescu, Ş De Martini, D Barnes, D Newman, P |
| author_sort | Gadd, M |
| collection | OXFORD |
| description | This paper presents a system for robust, large-scale topological localisation using Frequency-Modulated ContinuousWave (FMCW) scanning radar. We learn a metric space for embedding polar radar scans using CNN and NetVLAD architectures traditionally applied to the visual domain. However, we tailor the feature extraction for more suitability to the polar nature of radar scan formation using cylindrical convolutions, anti-aliasing blurring, and azimuth-wise max-pooling; all in order to bolster the rotational invariance. The enforced metric space is then used to encode a reference trajectory, serving as a map, which is queried for nearest neighbours (NNs) for recognition of places at run-time. We demonstrate the performance of our topological localisation system over the course of many repeat forays using the largest radar-focused mobile autonomy dataset released to date, totalling 280 km of urban driving, a small portion of which we also use to learn the weights of the modified architecture. As this work represents a novel application for FMCW radar, we analyse the utility of the proposed method via a comprehensive set of metrics which provide insight into the efficacy when used in a realistic system, showing improved performance over the root architecture even in the face of random rotational perturbation. |
| first_indexed | 2024-03-07T00:33:32Z |
| format | Conference item |
| id | oxford-uuid:80a2d643-623e-4edc-8be2-457f0394690f |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T00:33:32Z |
| publishDate | 2020 |
| publisher | IEEE Xplore |
| record_format | dspace |
| spelling | oxford-uuid:80a2d643-623e-4edc-8be2-457f0394690f2022-03-26T21:24:38ZKidnapped radar: topological radar localisation using rotationally-invariant metric learningConference itemhttp://purl.org/coar/resource_type/c_5794uuid:80a2d643-623e-4edc-8be2-457f0394690fEnglishSymplectic ElementsIEEE Xplore2020Gadd, MSǎftescu, ŞDe Martini, DBarnes, DNewman, PThis paper presents a system for robust, large-scale topological localisation using Frequency-Modulated ContinuousWave (FMCW) scanning radar. We learn a metric space for embedding polar radar scans using CNN and NetVLAD architectures traditionally applied to the visual domain. However, we tailor the feature extraction for more suitability to the polar nature of radar scan formation using cylindrical convolutions, anti-aliasing blurring, and azimuth-wise max-pooling; all in order to bolster the rotational invariance. The enforced metric space is then used to encode a reference trajectory, serving as a map, which is queried for nearest neighbours (NNs) for recognition of places at run-time. We demonstrate the performance of our topological localisation system over the course of many repeat forays using the largest radar-focused mobile autonomy dataset released to date, totalling 280 km of urban driving, a small portion of which we also use to learn the weights of the modified architecture. As this work represents a novel application for FMCW radar, we analyse the utility of the proposed method via a comprehensive set of metrics which provide insight into the efficacy when used in a realistic system, showing improved performance over the root architecture even in the face of random rotational perturbation. |
| spellingShingle | Gadd, M Sǎftescu, Ş De Martini, D Barnes, D Newman, P Kidnapped radar: topological radar localisation using rotationally-invariant metric learning |
| title | Kidnapped radar: topological radar localisation using rotationally-invariant metric learning |
| title_full | Kidnapped radar: topological radar localisation using rotationally-invariant metric learning |
| title_fullStr | Kidnapped radar: topological radar localisation using rotationally-invariant metric learning |
| title_full_unstemmed | Kidnapped radar: topological radar localisation using rotationally-invariant metric learning |
| title_short | Kidnapped radar: topological radar localisation using rotationally-invariant metric learning |
| title_sort | kidnapped radar topological radar localisation using rotationally invariant metric learning |
| work_keys_str_mv | AT gaddm kidnappedradartopologicalradarlocalisationusingrotationallyinvariantmetriclearning AT saftescus kidnappedradartopologicalradarlocalisationusingrotationallyinvariantmetriclearning AT demartinid kidnappedradartopologicalradarlocalisationusingrotationallyinvariantmetriclearning AT barnesd kidnappedradartopologicalradarlocalisationusingrotationallyinvariantmetriclearning AT newmanp kidnappedradartopologicalradarlocalisationusingrotationallyinvariantmetriclearning |