pith. machine review for the scientific record.
sign in

arxiv: 2605.07041 · v1 · submitted 2026-05-07 · 💻 cs.RO · cs.CV

Dr-BA: Separable Optimization for Direct Radar Bundle Adjustment & Localization

Daniil Lisus , Cedric Le Gentil , Timothy D. Barfoot This is my paper

Pith reviewed 2026-05-11 01:12 UTC · model grok-4.3

classification 💻 cs.RO cs.CV
keywords radar bundle adjustmentseparable optimizationdirect localizationspinning radarstate estimationautonomous navigationall-weather sensing
0
0 comments X

The pith

Dr-BA performs radar bundle adjustment directly on full intensity images by separating pose estimation from map construction.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper introduces a bundle adjustment method that keeps the complete radar intensity returns from spinning sensors instead of first turning them into point clouds. It splits the optimization into separate steps for vehicle poses and scene maps, which makes the computation efficient and extends naturally to localizing the vehicle inside a previously built map. The approach targets all-weather autonomy because radar keeps working when cameras and lidars are blinded by rain or snow. Results are shown on more than 200 km of real driving data collected on five different routes, where the method outperforms earlier radar-only techniques.

Core claim

Dr-BA formulates the radar bundle adjustment problem as a separable optimization that decouples the estimation of sensor poses from the recovery of dense maps built from full 2D spinning-radar intensity images. The same decoupled structure directly supports localization inside a pre-built map without any additional machinery. The method therefore avoids the information loss that occurs when radar data are first reduced to sparse point clouds.

What carries the argument

The separable optimization that decouples pose estimation from mapping, letting the full intensity images from multiple scans be used without point-cloud extraction.

If this is right

  • Radar state estimation can retain dense information instead of discarding most returns during point-cloud conversion.
  • The same code solves both mapping and cross-session localization without reformulation.
  • Vehicle motion and scene structure can be recovered at higher density than sparse-feature approaches allow.
  • All-weather navigation benefits because the method runs on a sensor that functions in precipitation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The separability may allow incremental updates when new radar scans arrive, supporting online map maintenance.
  • Dense maps produced by the method could be reused for radar-based object detection or terrain classification.
  • The same decoupling principle might apply to other modalities that produce dense returns, such as sonar or certain thermal cameras.

Load-bearing premise

The separable formulation produces pose and map estimates whose accuracy matches what a single joint optimization would achieve while still using every intensity return.

What would settle it

On the same 200 km dataset, a non-separable joint optimizer produces measurably lower localization error than Dr-BA, or Dr-BA's errors exceed those of existing point-cloud radar methods.

Figures

Figures reproduced from arXiv: 2605.07041 by Cedric Le Gentil, Daniil Lisus, Timothy D. Barfoot.

Figure 1
Figure 1. Figure 1: Dr-BA is a method for direct bundle adjustment of [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: An overview of the implementation of Dr-BA, including initialization steps and a downstream localization task. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Different processing steps for leveraging raw radar im [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison of trajectory estimates from [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Average ATE and convergence for the Industrial route with increasingly degraded Dr-PoGO initialization. E. Ablation studies 1) Processing impact: Table IV presents an ablation study on the impact of different processing parameters on BA and ultimately localization performance. Ablations are done on only a Suburbs and Skyway sequence as they contain a large number of overlapping regions in their traversals.… view at source ↗
Figure 7
Figure 7. Figure 7: Challenging segments of the Skyway, Farm, and Forest routes, characterized by sparse features, diffuse nat￾ural structure, or poor scan co-visibility. The lengthscale is shared by all plots. localization using Dr-BA-generated maps. Tests are run on five diverse trajectories, ranging from structured suburban environments to highly unstructured farm roads. REFERENCES [1] Nader J. Abu-Alrub and Nathir A. Rawa… view at source ↗
read the original abstract

This paper introduces Dr-BA, a first-of-its-kind radar bundle adjustment (BA) framework that operates directly on 2D spinning radar intensity images. Unlike camera or lidar sensors, radar is largely unaffected by precipitation, making it a critical modality for autonomous systems that require all-weather robustness. Existing state estimation approaches using spinning radar typically extract sparse point clouds from range-azimuth-intensity measurements and apply point cloud alignment techniques to estimate vehicle motion, scene structure, or to localize within an existing map. In contrast, Dr-BA uses the full radar returns from multiple scans to jointly estimate dense maps and sensor poses. By formulating the problem as a separable optimization, we derive an efficient and general solution that decouples pose estimation from mapping. In addition to solving the BA problem, this formulation naturally extends to direct radar-only localization (DRL) within a previously built map. Dr-BA achieves state-of-the-art radar-based BA and cross-session localization performance, demonstrated on more than 200 km of on-road data across five distinct routes. Our implementation is publicly available at https://github.com/utiasASRL/dr_ba.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces Dr-BA, a separable optimization framework for direct radar bundle adjustment (BA) that operates on full 2D spinning radar intensity images rather than extracted point clouds. It jointly estimates dense maps and sensor poses, extends naturally to direct radar-only localization (DRL) in a pre-built map, and reports state-of-the-art radar-based BA and cross-session localization performance on more than 200 km of on-road data across five routes. The implementation is released publicly.

Significance. If the separable formulation maintains accuracy comparable to joint optimization while delivering efficiency gains and avoiding point-cloud extraction, the work would represent a meaningful advance for all-weather robust localization in robotics. The scale of the evaluation (200 km across multiple routes) and public code release are notable strengths that support reproducibility and practical impact.

major comments (2)
  1. [Separable optimization formulation (method section)] The central claim that the separable optimization preserves accuracy equivalent to a joint formulation over full intensity images (without point-cloud extraction) is load-bearing for the SOTA assertion but lacks direct validation. No comparison of residual norms, pose errors, or convergence behavior between the separable solver and a non-separable joint baseline on the same intensity data is provided, despite radar-specific correlations (multipath, speckle) that could violate the independence assumptions required for exact equivalence.
  2. [Experimental results (§5)] §5 (experimental results) and the abstract report SOTA performance on >200 km of data but provide no error bars, statistical significance tests, or ablation details on key design choices (e.g., separability impact, map density). This weakens the ability to assess whether the efficiency gains come at an accuracy cost that would affect the cross-session localization claims.
minor comments (2)
  1. [Abstract] The abstract states the method is 'first-of-its-kind' for direct radar BA on intensity images; a brief comparison table or sentence distinguishing it from prior radar point-cloud BA methods would improve clarity.
  2. [Method] Notation for the separable cost function and the mapping/pose variables could be introduced earlier with an explicit equation reference to aid readers following the decoupling argument.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful review and constructive suggestions. Below, we provide point-by-point responses to the major comments. We are committed to making the suggested revisions to strengthen the paper's claims regarding the separable optimization and experimental validation.

read point-by-point responses

  1. Referee: [Separable optimization formulation (method section)] The central claim that the separable optimization preserves accuracy equivalent to a joint formulation over full intensity images (without point-cloud extraction) is load-bearing for the SOTA assertion but lacks direct validation. No comparison of residual norms, pose errors, or convergence behavior between the separable solver and a non-separable joint baseline on the same intensity data is provided, despite radar-specific correlations (multipath, speckle) that could violate the independence assumptions required for exact equivalence.

    Authors: We appreciate the referee pointing out the need for direct empirical validation of the separable formulation. The derivation in Section 3 shows that the separable approach is mathematically equivalent to the joint optimization when solving for the map parameters analytically, assuming measurement independence. However, to rigorously address concerns about radar-specific effects such as multipath and speckle potentially affecting this equivalence, we will include a new subsection or appendix in the revised manuscript comparing the separable solver to a joint baseline on a representative subset of the data. This will report residual norms, pose errors, and convergence behavior to confirm no significant accuracy loss. revision: yes

  2. Referee: [Experimental results (§5)] §5 (experimental results) and the abstract report SOTA performance on >200 km of data but provide no error bars, statistical significance tests, or ablation details on key design choices (e.g., separability impact, map density). This weakens the ability to assess whether the efficiency gains come at an accuracy cost that would affect the cross-session localization claims.

    Authors: We agree that providing error bars, statistical tests, and ablations would improve the robustness of our experimental claims. In the revised version, we will augment Section 5 with error bars on all reported metrics (e.g., mean and standard deviation across routes), include p-values or significance tests for comparisons against baselines, and add ablations specifically on the effect of the separable formulation versus joint, as well as varying map density. These additions will help demonstrate that the efficiency gains do not compromise the accuracy needed for the cross-session localization results. revision: yes

Circularity Check

0 steps flagged

No significant circularity; separable formulation derived independently of performance claims

full rationale

The paper's core contribution is a mathematical formulation of separable optimization for direct radar BA that decouples pose and mapping variables, presented as a derivation from the problem structure rather than a fit to data or self-citation. Performance claims rest on empirical evaluation over 200 km of real-world radar data, not on any prediction that reduces to the inputs by construction. No equations, uniqueness theorems, or ansatzes are shown to loop back to the target result; the separability is introduced as a modeling choice whose accuracy equivalence is asserted but not derived tautologically. This is a standard non-circular engineering derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No free parameters, axioms, or invented entities are described in the abstract.

pith-pipeline@v0.9.0 · 5503 in / 944 out tokens · 29412 ms · 2026-05-11T01:12:53.886021+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 40 canonical work pages

  1. [2]

    TBV Radar SLAM – Trust but Verify Loop Candidates.IEEE Robotics and Automation Letters (RA-L), 8(6):3613– 3620, 2023

    Daniel Adolfsson, Mattias Karlsson, Vladim ´ir Kubelka, Martin Magnusson, and Henrik Andreasson. TBV Radar SLAM – Trust but Verify Loop Candidates.IEEE Robotics and Automation Letters (RA-L), 8(6):3613– 3620, 2023. doi: 10.1109/LRA.2023.3268040

    work page doi:10.1109/lra.2023.3268040 2023

  2. [3]

    Lilienthal, and Henrik Andreasson

    Daniel Adolfsson, Martin Magnusson, Anas Alhashimi, Achim J. Lilienthal, and Henrik Andreasson. Lidar-Level Localization With Radar? The CFEAR Approach to Accurate, Fast, and Robust Large-Scale Radar Odometry in Diverse Environments.IEEE Trans. on Robotics (TRO), 39(2):1476–1495, 2023. doi: 10.1109/TRO.2022. 3221302

    work page doi:10.1109/tro.2022 2023

  3. [4]

    Cambridge University Press, 2024

    Timothy D Barfoot.State Estimation for Robotics. Cambridge University Press, 2024

    work page 2024

  4. [5]

    Masking by Moving: Learning Distraction-Free Radar Odometry from Pose Information

    Dan Barnes, Rob Weston, and Ingmar Posner. Masking by Moving: Learning Distraction-Free Radar Odometry from Pose Information. InConf. Robot Learn., 2019. URL https://arxiv.org/pdf/1909.03752

    work page arXiv 2019

  5. [6]

    Are We Ready for Radar to Replace Lidar in All-weather Mapping and Localization?IEEE Robotics and Automation Letters (RA-L), 7(4):10328–10335, 2022

    Keenan Burnett, Yuchen Wu, David J Yoon, Angela P Schoellig, and Timothy D Barfoot. Are We Ready for Radar to Replace Lidar in All-weather Mapping and Localization?IEEE Robotics and Automation Letters (RA-L), 7(4):10328–10335, 2022

    work page 2022

  6. [7]

    Boreas: A Multi-Season Autonomous Driving Dataset , url =

    Keenan Burnett, David J Yoon, Yuchen Wu, Andrew Z Li, Haowei Zhang, Shichen Lu, Jingxing Qian, Wei-Kang Tseng, Andrew Lambert, Keith YK Leung, Angela P Schoellig, and Timothy D Barfoot. Boreas: A Multi- season Autonomous Driving Dataset.Int. J. Robot. Res., 42(1-2):33–42, 2023. doi: 10.1177/02783649231160195

    work page doi:10.1177/02783649231160195 2023

  7. [8]

    Schoellig, and Timothy D

    Keenan Burnett, Angela P. Schoellig, and Timothy D. Barfoot. Continuous-Time Radar-Inertial and Lidar- Inertial Odometry Using a Gaussian Process Motion Prior.IEEE Trans. on Robotics (TRO), 41:1059–1076, January 2025. ISSN 1552-3098. doi: 10.1109/TRO.2024. 3521856

    work page doi:10.1109/tro.2024 2025

  8. [9]

    Radar SLAM Using Visual Features.EURASIP Journal on Advances in Signal Processing, 2011, 12 2011

    Jonas Callmer, David T ¨ornqvist, Fredrik Gustafsson, Henrik Svensson, and Pelle Carlbom. Radar SLAM Using Visual Features.EURASIP Journal on Advances in Signal Processing, 2011, 12 2011. doi: 10.1186/ 1687-6180-2011-71

    work page 2011

  9. [10]

    Precise Ego-Motion Estimation with Millimeter-Wave Radar Under Diverse and Challenging Conditions

    Sarah Cen and Paul Newman. Precise Ego-Motion Estimation with Millimeter-Wave Radar Under Diverse and Challenging Conditions. InProc. of the IEEE Intl. Conf. on Robotics & Automation (ICRA), pages 1–8, 05 2018. doi: 10.1109/ICRA.2018.8460687

    work page doi:10.1109/icra.2018.8460687 2018

  10. [11]

    Radar Scan Matching SLAM Using the Fourier-Mellin Transform

    Paul Checchin, Franck G ´erossier, Christophe Blanc, Roland Chapuis, and Laurent Trassoudaine. Radar Scan Matching SLAM Using the Fourier-Mellin Transform. In Field and Service Robotics: Results of the 7th Interna- tional Conference, pages 151–161. Springer, 2010

    work page 2010

  11. [12]

    Dissanayake, P

    M.W.M.G. Dissanayake, P. Newman, S. Clark, H.F. Durrant-Whyte, and M. Csorba. A Solution to the Simultaneous Localization and Map Building (SLAM) Problem.IEEE Trans. on Robotics and Automation, 17 (3):229–241, 2001. doi: 10.1109/70.938381

    work page doi:10.1109/70.938381 2001

  12. [13]

    Visual Teach and Repeat for Long-range Rover Autonomy.Journal of Field Robotics (JFR), 27(5):534–560, 2010

    Paul Furgale and Timothy D Barfoot. Visual Teach and Repeat for Long-range Rover Autonomy.Journal of Field Robotics (JFR), 27(5):534–560, 2010

    work page 2010

  13. [14]

    Separable Nonlinear Least Squares: the Variable Projection Method and its Applications.Inverse Problems, 19:R1–R26(1), 01 2003

    Gene Golub and Victor Pereyra. Separable Nonlinear Least Squares: the Variable Projection Method and its Applications.Inverse Problems, 19:R1–R26(1), 01 2003

    work page 2003

  14. [15]

    Golub and Victor Pereyra

    G.H. Golub and Victor Pereyra. The Differentiation of Pseudo-Inverses and Nonlinear Least Squares Problems Whose Variables Separate.SIAM Journal on Numerical Analysis, 10:413–432, 04 1973. doi: 10.1137/0710036

    work page doi:10.1137/0710036 1973

  15. [16]

    A New Wave in Robotics: Survey on Recent mmWave Radar Applications in Robotics , url =

    Kyle Harlow, Hyesu Jang, Timothy D. Barfoot, Ay- oung Kim, and Christoffer Heckman. A New Wave in Robotics: Survey on Recent MmWave Radar Applica- tions in Robotics.IEEE Trans. on Robotics (TRO), 40: 4544–4560, 2024. doi: 10.1109/TRO.2024.3463504

    work page doi:10.1109/tro.2024.3463504 2024

  16. [17]

    M2DP: A Novel 3D Point Cloud Descriptor and its Application in Loop Closure Detection

    Li He, Xiaolong Wang, and Hong Zhang. M2DP: A Novel 3D Point Cloud Descriptor and its Application in Loop Closure Detection. InProc. of the IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems (IROS), pages 231–237, 2016. doi: 10.1109/IROS.2016.7759060

    work page doi:10.1109/iros.2016.7759060 2016

  17. [18]

    Large scale place recognition in 2D LIDAR scans using Geometrical Land- mark Relations

    Marian Himstedt, Jan Frost, Sven Hellbach, Hans- Joachim B ¨ohme, and Erik Maehle. Large scale place recognition in 2D LIDAR scans using Geometrical Land- mark Relations. InProc. of the IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems (IROS), pages 5030–5035,

    work page

  18. [19]

    doi: 10.1109/IROS.2014.6943277

    work page doi:10.1109/iros.2014.6943277 2014

  19. [20]

    Real-Time Pose Graph SLAM based on Radar

    Martin Holder, Sven Hellwig, and Hermann Winner. Real-Time Pose Graph SLAM based on Radar. In2019 IEEE Intelligent V ehicles Symposium (IV), pages 1145– 1151, 2019. doi: 10.1109/IVS.2019.8813841

    work page doi:10.1109/ivs.2019.8813841 2019

  20. [21]

    Connor Holmes and Timothy D. Barfoot. An Effi- cient Global Optimality Certificate for Landmark-Based SLAM.IEEE Robotics and Automation Letters, 8(3): 1539–1546, 2023. doi: 10.1109/LRA.2023.3238173

    work page doi:10.1109/lra.2023.3238173 2023

  21. [22]

    RadarSLAM: A Robust Simultaneous Localiza- tion and Mapping System for All Weather Conditions

    Ziyang Hong, Yvan Petillot, Andrew Wallace, and Sen Wang. RadarSLAM: A Robust Simultaneous Localiza- tion and Mapping System for All Weather Conditions. Intl. Journal of Robotics Research (IJRR), 41(5):519– 542, 2022. doi: 10.1177/02783649221080483

    work page doi:10.1177/02783649221080483 2022

  22. [23]

    RaPlace: Place Recognition for Imaging Radar using Radon Trans- form and Mutable Threshold

    Hyesu Jang, Minwoo Jung, and Ayoung Kim. RaPlace: Place Recognition for Imaging Radar using Radon Trans- form and Mutable Threshold. InProc. of the IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems (IROS), pages 11194–11201, 2023. doi: 10.1109/IROS55552. 2023.10341883

    work page doi:10.1109/iros55552 2023

  23. [24]

    XPRESS: X-Band Radar Place Recognition via Elliptical Scan Shaping.IEEE Robotics and Automation Letters, 10(12):13121–13128, 2025

    Hyesu Jang, Wooseong Yang, Ayoung Kim, Dongje Lee, and Hanguen Kim. XPRESS: X-Band Radar Place Recognition via Elliptical Scan Shaping.IEEE Robotics and Automation Letters, 10(12):13121–13128, 2025

    work page 2025

  24. [25]

    Scan Context++: Structural Place Recognition Robust to Rota- tion and Lateral Variations in Urban Environments.IEEE Trans

    Giseop Kim, Sunwook Choi, and Ayoung Kim. Scan Context++: Structural Place Recognition Robust to Rota- tion and Lateral Variations in Urban Environments.IEEE Trans. on Robotics (TRO), 38(3):1856–1874, 2022. doi: 10.1109/TRO.2021.3116424

    work page doi:10.1109/tro.2021.3116424 2022

  25. [26]

    Parallel Tracking and Mapping for Small AR Workspaces

    Georg Klein and David Murray. Parallel Tracking and Mapping for Small AR Workspaces. InProc. Sixth IEEE and ACM International Symposium on Mixed and Augmented Reality (ISMAR’07), Nara, Japan, November 2007

    work page 2007

  26. [27]

    Wang, and Wen-Chieh Lin

    Pou-Chun Kung, C. Wang, and Wen-Chieh Lin. A Normal Distribution Transform-Based Radar Odometry Designed For Scanning and Automotive Radars.Proc. of the IEEE Intl. Conf. on Robotics & Automation (ICRA), pages 14417–14423, 2021

    work page 2021

  27. [28]

    Cedric Le Gentil, Leonardo Brizi, Daniil Lisus, Xinyuan Qiao, Giorgio Grisetti, and Timothy D. Barfoot. DRO: Doppler-Aware Direct Radar Odometry. InProc. of Robotics: Science and Systems (RSS), 2025

    work page 2025

  28. [29]

    Cedric Le Gentil, Raphael Falque, Daniil Lisus, and Timothy D. Barfoot. 2FAST-2LAMAA: Large-Scale Lidar-Inertial Localization and Mapping with Continuous Distance Fields, 2025. URL https://arxiv.org/abs/2410. 05433

    work page 2025

  29. [30]

    Cedric Le Gentil, Weican Li, Leonardo Brizi, and Tim- othy D. Barfoot. Dr-PoGO: Direct Radar Pose-Graph Optimization. InProc. of the IEEE Intl. Conf. on Robotics & Automation (ICRA), 2026

    work page 2026

  30. [31]

    ORORA: Outlier-robust Radar Odometry

    Hyungtae Lim, Kawon Han, Gunhee Shin, Giseop Kim, Songcheol Hong, and Hyun Myung. ORORA: Outlier-robust Radar Odometry. InProc. of the IEEE Intl. Conf. on Robotics & Automation (ICRA), pages 2046–2053, 2023

    work page 2046

  31. [32]

    Yoon, Richard Poulton, John Marshall, and Timothy D

    Daniil Lisus, Keenan Burnett, David J. Yoon, Richard Poulton, John Marshall, and Timothy D. Barfoot. Are Doppler Velocity Measurements Useful for Spinning Radar Odometry?IEEE Robotics and Automation Letters (RA-L), 10(1):224–231, 2025. doi: 10.1109/LRA.2024. 3505821

    work page doi:10.1109/lra.2024 2025

  32. [33]

    Papais, Cedric Le Gentil, Elliot Preston-Krebs, Andrew Lambert, Keith Y

    Daniil Lisus, Katya M. Papais, Cedric Le Gentil, Elliot Preston-Krebs, Andrew Lambert, Keith Y . K. Leung, and Timothy D. Barfoot. Boreas Road Trip: A Multi-Sensor Autonomous Driving Dataset on Challenging Roads,

    work page

  33. [34]

    URL https://arxiv.org/abs/2602.16870

    work page arXiv

  34. [35]

    David G. Lowe. Distinctive Image Features from Scale- Invariant Keypoints.Intl. Journal of Computer Vision (IJCV), 60(2):91–110, November 2004

    work page 2004

  35. [36]

    Zhehua Mao, Liang Zhao, Shoudong Huang, Yiting Fan, and Alex P. W. Lee. DSR: Direct Simultaneous Registration for Multiple 3D Images. InMedical Image Computing and Computer Assisted Intervention – MIC- CAI 2022, pages 98–107, 2022

    work page 2022

  36. [37]

    Sparse Variable Projection in Robotic Perception: Exploiting Separable Structure for Efficient Nonlinear Optimization,

    Alan Papalia, Nikolas Sanderson, Haoyu Han, Heng Yang, Hanumant Singh, and Michael Everett. Sparse Variable Projection in Robotic Perception: Exploiting Separable Structure for Efficient Nonlinear Optimization,

    work page

  37. [38]

    URL https://arxiv.org/abs/2512.07969

    work page arXiv

  38. [39]

    In: 2020 IEEE Interna- tional Conference on Robotics and Automation, ICRA 2020, Paris, France, May 31 - August 31, 2020

    Yeong Sang Park, Young-Sik Shin, and Ayoung Kim. PhaRaO: Direct Radar Odometry using Phase Corre- lation. InProc. of the IEEE Intl. Conf. on Robotics & Automation (ICRA), pages 2617–2623, 2020. doi: 10.1109/ICRA40945.2020.9197231

    work page doi:10.1109/icra40945.2020.9197231 2020

  39. [40]

    Bar- foot

    Elliot Preston-Krebs, Daniil Lisus, and Timothy D. Bar- foot. The Finer Points: A Systematic Comparison of Point-Cloud Extractors for Radar Odometry.Proceedings of the Conference on Robots and Vision, may 27 2025. https://crv.pubpub.org/pub/urc6a1fo

    work page 2025

  40. [41]

    Mrdjan Jankovic

    Arthur Venon, Yohan Dupuis, Pascal Vasseur, and Pierre Merriaux. Millimeter Wave FMCW RADARs for Percep- tion, Recognition and Localization in Automotive Appli- cations: A Survey.IEEE Trans. on Intelligent V ehicles (T-IV), 7:1–1, 09 2022. doi: 10.1109/TIV .2022.3167733. SUPPLEMENTARYMATERIAL A. DETAILEDSEQUENCEVISUALS A. Suburbs 500 m Fig. A.1: Sparsity ...

    work page doi:10.1109/tiv 2022