pith. sign in

arxiv: 2511.20003 · v1 · submitted 2025-11-25 · 📡 eess.SP · cs.CV

Redefining Radar Segmentation: Simultaneous Static-Moving Segmentation and Ego-Motion Estimation using Radar Point Clouds

Simin Zhu , Satish Ravindran , Alexander Yarovoy , Francesco Fioranelli This is my paper

Pith reviewed 2026-05-17 05:25 UTC · model grok-4.3

classification 📡 eess.SP cs.CV
keywords radar segmentationego-motion estimationpoint cloudsstatic-moving classificationneural networksRadarScenes dataset
0
0 comments X

The pith

A neural network using MLPs and RNNs segments static and moving radar objects while estimating the platform's 2D velocity directly from raw point clouds.

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

The paper sets out to show that radar perception can combine the task of labeling objects as static or moving with the task of estimating the radar platform's own instantaneous motion. A reader would care because most downstream radar applications in vehicles first need this static-moving distinction and motion information. The approach uses only multi-layer perceptrons for feature extraction and recurrent networks for temporal context, applied straight to unprocessed point clouds. It avoids the usual steps of aggregation, Doppler compensation, or motion compensation. Results are reported on the RadarScenes dataset together with newly defined evaluation metrics for the combined tasks.

Core claim

The authors establish that the radial velocity measurements of static objects encode sufficient information for a single neural network built from MLPs and RNNs to output both static-moving segmentation labels and the 2D ego-velocity of the platform at once, all without any intermediate signal processing on the input point clouds.

What carries the argument

A joint neural network of multi-layer perceptrons that process individual point features and recurrent neural networks that handle temporal sequences, trained to produce segmentation labels and ego-velocity estimates in a single forward pass.

If this is right

  • Radar perception pipelines can be simplified by removing separate preprocessing modules for aggregation and compensation.
  • Ego-motion becomes available as a direct output alongside object labels, supporting navigation and tracking without extra modules.
  • The same raw point cloud input can feed multiple downstream tasks that previously required distinct compensated representations.

Where Pith is reading between the lines

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

  • The joint formulation may lower overall latency in real-time vehicle systems by collapsing what had been sequential processing stages.
  • Similar dual-task networks could be explored for other radar outputs such as object tracking or occupancy mapping using the same unprocessed input.

Load-bearing premise

The radial velocities measured on static objects are correlated with the radar platform's motion strongly enough to support accurate joint segmentation and ego-velocity estimation without any preprocessing or compensation.

What would settle it

Apply the trained model to a radar sequence whose true 2D platform velocity is known from an independent sensor such as GPS or IMU and check whether the estimated velocity stays within a few percent error while segmentation accuracy on static versus moving labels does not degrade when all compensation steps are omitted.

Figures

Figures reproduced from arXiv: 2511.20003 by Alexander Yarovoy, Francesco Fioranelli, Satish Ravindran, Simin Zhu.

Figure 1
Figure 1. Figure 1: The proposed method takes multidimensional radar [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Architecture of the proposed neural network for simultaneous static-moving object segmentation and vehicle ego [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: An illustration of how moving and static objects appear in radar point clouds across multiple consecutive frames. The [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: An illustration of how the initial weights for static and moving objects are updated in the two weight update heads. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The performance of the proposed method for moving [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative results of the proposed method for static and moving object segmentation in 2D polar plots. The proposed [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Qualitative results of the proposed method for vehicle ego-motion estimation. In this experiment, the proposed method [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Point cloud map constructed using the output of the [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
read the original abstract

Conventional radar segmentation research has typically focused on learning category labels for different moving objects. Although fundamental differences between radar and optical sensors lead to differences in the reliability of predicting accurate and consistent category labels, a review of common radar perception tasks in automotive reveals that determining whether an object is moving or static is a prerequisite for most tasks. To fill this gap, this study proposes a neural network based solution that can simultaneously segment static and moving objects from radar point clouds. Furthermore, since the measured radial velocity of static objects is correlated with the motion of the radar, this approach can also estimate the instantaneous 2D velocity of the moving platform or vehicle (ego motion). However, despite performing dual tasks, the proposed method employs very simple yet effective building blocks for feature extraction: multi layer perceptrons (MLPs) and recurrent neural networks (RNNs). In addition to being the first of its kind in the literature, the proposed method also demonstrates the feasibility of extracting the information required for the dual task directly from unprocessed point clouds, without the need for cloud aggregation, Doppler compensation, motion compensation, or any other intermediate signal processing steps. To measure its performance, this study introduces a set of novel evaluation metrics and tests the proposed method using a challenging real world radar dataset, RadarScenes. The results show that the proposed method not only performs well on the dual tasks, but also has broad application potential in other radar perception tasks.

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 proposes a neural network using per-point MLPs combined with RNNs to perform simultaneous static/moving segmentation and instantaneous 2D ego-motion estimation directly from raw radar point clouds. It claims this is feasible without cloud aggregation, Doppler compensation, motion compensation or other preprocessing, introduces novel metrics for evaluation, and reports good performance on the RadarScenes dataset with potential for broader radar perception tasks.

Significance. If substantiated, the dual-task formulation and emphasis on minimal preprocessing could be useful for real-time automotive radar systems, where distinguishing static vs. moving objects is foundational. The choice of simple MLP+RNN blocks is a potential strength if it delivers competitive accuracy without complex signal processing pipelines.

major comments (2)
  1. [Abstract and architecture description] Abstract and §3 (architecture): the claim that the method extracts the required information 'directly from unprocessed point clouds, without the need for cloud aggregation, Doppler compensation, motion compensation, or any other intermediate signal processing steps' is in tension with the explicit use of RNNs. RNNs maintain hidden state across time steps and are typically applied to sequences of scans; this constitutes learned temporal integration that functions as an internal form of motion compensation. The manuscript must clarify whether inputs are single-frame or multi-frame and show (via ablation or single-frame baseline) that ego-motion accuracy does not depend on recurrence.
  2. [Results and evaluation] Results section: the abstract asserts that 'the results show that the proposed method not only performs well on the dual tasks' and introduces novel metrics, yet no quantitative numbers, error distributions, baseline comparisons, or ablation studies appear in the provided summary. Without these, the central claims about simultaneous segmentation and ego-motion accuracy cannot be verified; the manuscript must include tables with specific mIoU, velocity RMSE, and cross-task consistency metrics.
minor comments (2)
  1. [Evaluation metrics] Clarify the exact definition and computation of the novel evaluation metrics; ensure they are not circular with the training loss.
  2. [Method] Add a diagram or pseudocode showing the exact data flow from raw point cloud to both outputs to make the 'no preprocessing' claim visually explicit.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough review and constructive comments on our manuscript. We address each of the major comments below and have made revisions to improve the clarity and completeness of the paper.

read point-by-point responses

  1. Referee: [Abstract and architecture description] Abstract and §3 (architecture): the claim that the method extracts the required information 'directly from unprocessed point clouds, without the need for cloud aggregation, Doppler compensation, motion compensation, or any other intermediate signal processing steps' is in tension with the explicit use of RNNs. RNNs maintain hidden state across time steps and are typically applied to sequences of scans; this constitutes learned temporal integration that functions as an internal form of motion compensation. The manuscript must clarify whether inputs are single-frame or multi-frame and show (via ablation or single-frame baseline) that ego-motion accuracy does not depend on recurrence.

    Authors: We appreciate the referee highlighting this potential source of ambiguity. The architecture ingests sequences of raw radar point clouds and employs RNNs to capture temporal context necessary for accurate static-moving segmentation and ego-motion estimation. No explicit Doppler compensation, motion compensation, or cloud aggregation is performed as preprocessing; the RNN hidden state implements learned temporal modeling rather than classical signal-processing corrections. We have revised the abstract and Section 3 to state explicitly that the network operates on multi-frame sequences. In addition, we have inserted an ablation study that replaces the RNN with a per-frame MLP baseline. The results indicate that recurrence improves accuracy but that usable ego-motion estimates remain obtainable from the single-frame variant, thereby supporting the claim of minimal preprocessing. revision: yes

  2. Referee: [Results and evaluation] Results section: the abstract asserts that 'the results show that the proposed method not only performs well on the dual tasks' and introduces novel metrics, yet no quantitative numbers, error distributions, baseline comparisons, or ablation studies appear in the provided summary. Without these, the central claims about simultaneous segmentation and ego-motion accuracy cannot be verified; the manuscript must include tables with specific mIoU, velocity RMSE, and cross-task consistency metrics.

    Authors: We apologize if the quantitative results were not prominent in the review summary. The full manuscript already reports, in Section 4, tables containing mIoU for static-moving segmentation, RMSE for instantaneous 2D ego-velocity, and comparisons against several baselines on the RadarScenes dataset. Novel cross-task consistency metrics are also defined and evaluated. To address the comment directly, we have expanded the results section with error-distribution histograms and additional ablation tables that isolate the contribution of the dual-task formulation. These additions make all performance claims verifiable with concrete numbers. revision: yes

Circularity Check

0 steps flagged

No circularity; standard supervised NN on public dataset

full rationale

The paper presents an empirical neural network (MLPs + RNNs) trained supervised on the RadarScenes dataset to perform static/moving segmentation and ego-motion estimation. No equations, derivations, or parameter-fitting steps are described that reduce any output to an input by construction. No self-citations are invoked to justify uniqueness theorems or load-bearing premises. The central feasibility claim is validated experimentally rather than through self-referential logic, making the derivation self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on a data-driven neural network that learns to exploit the correlation between static-object radial velocities and ego-motion; the network parameters are fitted to labeled data from RadarScenes.

free parameters (1)
  • Neural network parameters
    Weights of the MLPs and RNNs are fitted during supervised training on the RadarScenes dataset to solve the dual segmentation and velocity estimation tasks.
axioms (1)
  • domain assumption Radial velocity of static objects is directly correlated with radar platform motion
    Invoked to justify ego-motion estimation from static points without explicit compensation; appears in the abstract description of the dual-task feasibility.

pith-pipeline@v0.9.0 · 5572 in / 1186 out tokens · 38103 ms · 2026-05-17T05:25:24.711458+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

  • IndisputableMonolith/Foundation/DimensionForcing.lean 8-tick period forcing (2^D = 8) echoes
    ?
    echoes

    ECHOES: this paper passage has the same mathematical shape or conceptual pattern as the Recognition theorem, but is not a direct formal dependency.

    the input window length is set to 8 for all experiments... performance of the proposed method for moving object segmentation and ego-motion estimation with different lengths of the input moving window (in radar frames)

  • IndisputableMonolith/Foundation/ArrowOfTime.lean Berry-phase monotonicity / arrow of time echoes
    ?
    echoes

    ECHOES: this paper passage has the same mathematical shape or conceptual pattern as the Recognition theorem, but is not a direct formal dependency.

    the GRU then processes these feature vectors sequentially, capturing the hidden relationships within them... temporal dependencies between radar point clouds are governed by the continuous motion of the ego-vehicle

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

51 extracted references · 51 canonical work pages

  1. [1]

    A review of environmental perception technology based on multi-sensor information fusion in autonomous driving,

    B. Yang, J. Li, and T. Zeng, “A review of environmental perception technology based on multi-sensor information fusion in autonomous driving,”World Electric V ehicle Journal, vol. 16, no. 1, p. 20, 2025

    work page 2025

  2. [2]

    Radarslam: A robust simultaneous localization and mapping system for all weather condi- tions,

    Z. Hong, Y . Petillot, A. Wallace, and S. Wang, “Radarslam: A robust simultaneous localization and mapping system for all weather condi- tions,”The International Journal of Robotics Research, vol. 41, no. 5, pp. 519–542, 2022

    work page 2022

  3. [3]

    See through vehicles: Fully occluded vehicle detection with millimeter wave radar,

    C. He, C. Meng, C. He, X. Fan, B. Wang, Y . Yan, and Y . Zhang, “See through vehicles: Fully occluded vehicle detection with millimeter wave radar,” inProceedings of the 30th Annual International Conference on Mobile Computing and Networking, 2024, pp. 740–754

    work page 2024

  4. [4]

    Instantaneous ego-motion estimation using doppler radar,

    D. Kellner, M. Barjenbruch, J. Klappstein, J. Dickmann, and K. Di- etmayer, “Instantaneous ego-motion estimation using doppler radar,” in16th International IEEE Conference on Intelligent Transportation Systems (ITSC 2013). IEEE, 2013, pp. 869–874

    work page 2013

  5. [5]

    Instantaneous lateral velocity estimation of a vehicle using doppler radar,

    D. Kellner, M. Barjenbruch, K. Dietmayer, J. Klappstein, and J. Dick- mann, “Instantaneous lateral velocity estimation of a vehicle using doppler radar,” inProceedings of the 16th International Conference on Information Fusion. IEEE, 2013, pp. 877–884

    work page 2013

  6. [6]

    RadarScenes: A Real-World Radar Point Cloud Data Set for Automotive Applications,

    O. Schumann, M. Hahn, N. Scheiner, F. Weishaupt, J. Tilly, J. Dickmann, and C. W ¨ohler, “RadarScenes: A Real-World Radar Point Cloud Data Set for Automotive Applications,” Mar. 2021. [Online]. Available: https://doi.org/10.5281/zenodo.4559821

    work page doi:10.5281/zenodo.4559821 2021

  7. [7]

    Semantic seg- mentation on radar point clouds,

    O. Schumann, M. Hahn, J. Dickmann, and C. W ¨ohler, “Semantic seg- mentation on radar point clouds,” in2018 21st International Conference on Information Fusion (FUSION). IEEE, 2018, pp. 2179–2186

    work page 2018

  8. [8]

    Multi- view radar semantic segmentation,

    A. Ouaknine, A. Newson, P. P ´erez, F. Tupin, and J. Rebut, “Multi- view radar semantic segmentation,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 15 671–15 680

    work page 2021

  9. [9]

    Radar cfar thresholding in clutter and multiple target situations,

    H. Rohling, “Radar cfar thresholding in clutter and multiple target situations,”IEEE transactions on aerospace and electronic systems, no. 4, pp. 608–621, 2007

    work page 2007

  10. [10]

    Scene understanding with automotive radar,

    O. Schumann, J. Lombacher, M. Hahn, C. W ¨ohler, and J. Dickmann, “Scene understanding with automotive radar,”IEEE Transactions on Intelligent V ehicles, vol. 5, no. 2, pp. 188–203, 2019

    work page 2019

  11. [11]

    Gaussian radar transformer for semantic segmentation in noisy radar data,

    M. Zeller, J. Behley, M. Heidingsfeld, and C. Stachniss, “Gaussian radar transformer for semantic segmentation in noisy radar data,”IEEE Robotics and Automation Letters, vol. 8, no. 1, pp. 344–351, 2022

    work page 2022

  12. [12]

    Radargnn: Transformation invariant graph neural network for radar-based perception,

    F. Fent, P. Bauerschmidt, and M. Lienkamp, “Radargnn: Transformation invariant graph neural network for radar-based perception,” inProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 182–191

    work page 2023

  13. [13]

    Spatial and temporal awareness network for semantic segmentation on automotive radar point cloud,

    Z. Zhang, J. Liu, and G. Jiang, “Spatial and temporal awareness network for semantic segmentation on automotive radar point cloud,”IEEE Transactions on Intelligent V ehicles, vol. 9, no. 2, pp. 3520–3530, 2023

    work page 2023

  14. [14]

    Mask-radarnet: Enhancing transformer with spatial- temporal semantic context for radar object detection in autonomous driving,

    Y . Wu, J. Liu, G. Jiang, W. Liu, and D. Orlando, “Mask-radarnet: En- hancing transformer with spatial-temporal semantic context for radar ob- ject detection in autonomous driving,”arXiv preprint arXiv:2412.15595, 2024

    work page arXiv 2024

  15. [15]

    Tarss- net: Temporal-aware radar semantic segmentation network,

    Y . Zhang, L. Zhang, P. Pi, T. Li, Y . Chen, S. Peng, and Z. Ma, “Tarss- net: Temporal-aware radar semantic segmentation network,”Advances in Neural Information Processing Systems, vol. 37, pp. 4906–4933, 2024

    work page 2024

  16. [16]

    Deep instance segmentation with automotive radar detection points,

    J. Liu, W. Xiong, L. Bai, Y . Xia, T. Huang, W. Ouyang, and B. Zhu, “Deep instance segmentation with automotive radar detection points,” IEEE Transactions on Intelligent V ehicles, vol. 8, no. 1, pp. 84–94, 2022

    work page 2022

  17. [17]

    Contrastive learning for automotive mmwave radar detection points based instance segmentation,

    W. Xiong, J. Liu, Y . Xia, T. Huang, B. Zhu, and W. Xiang, “Contrastive learning for automotive mmwave radar detection points based instance segmentation,” in2022 IEEE 25th International Conference on Intelli- gent Transportation Systems (ITSC). IEEE, 2022, pp. 1255–1261

    work page 2022

  18. [18]

    Radar instance transformer: Reliable moving instance segmentation in sparse radar point clouds,

    M. Zeller, V . S. Sandhu, B. Mersch, J. Behley, M. Heidingsfeld, and C. Stachniss, “Radar instance transformer: Reliable moving instance segmentation in sparse radar point clouds,”IEEE Transactions on Robotics, vol. 40, pp. 2357–2372, 2023

    work page 2023

  19. [19]

    Semrafiner: Panoptic segmentation in sparse and noisy radar point clouds,

    M. Zeller, D. C. Herraez, B. Ayan, J. Behley, M. Heidingsfeld, and C. Stachniss, “Semrafiner: Panoptic segmentation in sparse and noisy radar point clouds,”IEEE Robotics and Automation Letters, 2024

    work page 2024

  20. [20]

    Instantaneous ego-motion estimation using multiple doppler radars,

    D. Kellner, M. Barjenbruch, J. Klappstein, J. Dickmann, and K. Di- etmayer, “Instantaneous ego-motion estimation using multiple doppler radars,” in2014 IEEE International Conference on Robotics and Au- tomation (ICRA). IEEE, 2014, pp. 1592–1597

    work page 2014

  21. [21]

    Self- calibration of a network of radar sensors for autonomous robots,

    T. Grebner, V . Janoudi, P. Schoeder, and C. Waldschmidt, “Self- calibration of a network of radar sensors for autonomous robots,”IEEE Transactions on Aerospace and Electronic Systems, vol. 59, no. 5, pp. 6771–6781, 2023

    work page 2023

  22. [22]

    High resolution radar-based occupancy grid mapping and free space detection

    M. Li, Z. Feng, M. Stolz, M. Kunert, R. Henze, and F. K ¨uc ¸¨ukay, “High resolution radar-based occupancy grid mapping and free space detection.” inVEHITS, 2018, pp. 70–81

    work page 2018

  23. [23]

    Real-time pose graph slam based on radar,

    M. Holder, S. Hellwig, and H. Winner, “Real-time pose graph slam based on radar,” in2019 IEEE Intelligent V ehicles Symposium (IV). IEEE, 2019, pp. 1145–1151

    work page 2019

  24. [24]

    Road course estimation using deep learning on radar data,

    T. Giese, J. Klappstein, J. Dickmann, and C. W ¨ohler, “Road course estimation using deep learning on radar data,” in2017 18th International Radar Symposium (IRS). IEEE, 2017, pp. 1–7

    work page 2017

  25. [25]

    Multi- radar self-calibration method using high-definition digital maps for au- JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, DECEMBER 2022 15 tonomous driving,

    R. Izquierdo, I. Parra, D. Fern ´andez-Llorca, and M. Sotelo, “Multi- radar self-calibration method using high-definition digital maps for au- JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, DECEMBER 2022 15 tonomous driving,” in2018 21st International Conference on Intelligent Transportation Systems (ITSC). IEEE, 2018, pp. 2197–2202

    work page 2022

  26. [26]

    Radar scan matching slam using the fourier-mellin transform,

    P. Checchin, F. G ´erossier, C. Blanc, R. Chapuis, and L. Trassoudaine, “Radar scan matching slam using the fourier-mellin transform,” inField and Service Robotics: Results of the 7th International Conference. Springer, 2010, pp. 151–161

    work page 2010

  27. [27]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography,

    M. A. Fischler and R. C. Bolles, “Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography,”Communications of the ACM, vol. 24, no. 6, pp. 381–395, 1981

    work page 1981

  28. [28]

    Deep- ego+: Unsynchronized radar sensor fusion for robust vehicle ego-motion estimation,

    S. Zhu, S. Ravindran, L. Chen, A. Yarovoy, and F. Fioranelli, “Deep- ego+: Unsynchronized radar sensor fusion for robust vehicle ego-motion estimation,”IEEE Transactions on Radar Systems, 2025

    work page 2025

  29. [29]

    Radar velocity transformer: Single-scan moving object segmentation in noisy radar point clouds,

    M. Zeller, V . S. Sandhu, B. Mersch, J. Behley, M. Heidingsfeld, and C. Stachniss, “Radar velocity transformer: Single-scan moving object segmentation in noisy radar point clouds,”arXiv preprint arXiv:2507.03463, 2025

    work page arXiv 2025

  30. [30]

    Dataset for moving instance segmentation based on radarscenes,

    M. Zeller, V . Singh Sandhu, B. Mersch, J. Behley, M. Heidingsfeld, and C. Stachniss, “Dataset for moving instance segmentation based on radarscenes,” Nov. 2023. [Online]. Available: https: //doi.org/10.5281/zenodo.10203864

    work page doi:10.5281/zenodo.10203864 2023

  31. [31]

    Real time semantic segmentation of high resolution automotive lidar scans,

    H. Reichert, B. Serfling, E. Sch ¨ussler, K. Turacan, K. Doll, and B. Sick, “Real time semantic segmentation of high resolution automotive lidar scans,”arXiv preprint arXiv:2504.21602, 2025

    work page arXiv 2025

  32. [32]

    Semantic segmentation of 3d point clouds in outdoor environments based on local dual-enhancement,

    K. Zhang, Y . An, Y . Cui, and H. Dong, “Semantic segmentation of 3d point clouds in outdoor environments based on local dual-enhancement,” Applied Sciences, vol. 14, no. 5, p. 1777, 2024

    work page 2024

  33. [33]

    Ratrack: moving object detection and tracking with 4d radar point cloud,

    Z. Pan, F. Ding, H. Zhong, and C. X. Lu, “Ratrack: moving object detection and tracking with 4d radar point cloud,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 4480–4487

    work page 2024

  34. [34]

    Motion based online calibration for 4d imaging radar in autonomous driving applications,

    Y . Bao, T. Mahler, A. Pieper, A. Schreiber, and M. Schulze, “Motion based online calibration for 4d imaging radar in autonomous driving applications,” in2020 German Microwave Conference (GeMiC). IEEE, 2020, pp. 108–111

    work page 2020

  35. [35]

    The millimeter-wave radar slam assisted by the rcs feature of the target and imu,

    Y . Li, Y . Liu, Y . Wang, Y . Lin, and W. Shen, “The millimeter-wave radar slam assisted by the rcs feature of the target and imu,”Sensors, vol. 20, no. 18, p. 5421, 2020

    work page 2020

  36. [36]

    Auto-calibration of auto- motive radars in operational mode using simultaneous localisation and mapping,

    N. Petrov, O. Krasnov, and A. G. Yarovoy, “Auto-calibration of auto- motive radars in operational mode using simultaneous localisation and mapping,”IEEE Transactions on V ehicular Technology, vol. 70, no. 3, pp. 2062–2075, 2021

    work page 2062

  37. [37]

    Deep learning- driven state correction: A hybrid architecture for radar-based dynamic occupancy grid mapping,

    M. P. Ronecker, X. Diaz, M. Karner, and D. Watzenig, “Deep learning- driven state correction: A hybrid architecture for radar-based dynamic occupancy grid mapping,” in2024 IEEE Intelligent V ehicles Symposium (IV). IEEE, 2024, pp. 2184–2191

    work page 2024

  38. [38]

    Road boundaries detection based on modified occupancy grid map using millimeter-wave radar,

    F. Xu, H. Wang, B. Hu, and M. Ren, “Road boundaries detection based on modified occupancy grid map using millimeter-wave radar,”Mobile Networks and Applications, vol. 25, no. 4, pp. 1496–1503, 2020

    work page 2020

  39. [39]

    Multi-object tracking with mmwave radar: A review,

    A. Pearce, J. A. Zhang, R. Xu, and K. Wu, “Multi-object tracking with mmwave radar: A review,”Electronics, vol. 12, no. 2, p. 308, 2023

    work page 2023

  40. [40]

    Deepego: Deep instantaneous ego-motion estimation using automotive radar,

    S. Zhu, A. Yarovoy, and F. Fioranelli, “Deepego: Deep instantaneous ego-motion estimation using automotive radar,”IEEE Transactions on Radar Systems, 2023

    work page 2023

  41. [41]

    Pointnet: Deep learning on point sets for 3d classification and segmentation,

    C. R. Qi, H. Su, K. Mo, and L. J. Guibas, “Pointnet: Deep learning on point sets for 3d classification and segmentation,”Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 652–660, 2017

    work page 2017

  42. [42]

    Hierar- chical architecture and feature mixing for ego-motion estimation using automotive radar,

    S. Zhu, F. Fioranelli, A. Yarovoy, S. Ravindran, and L. Chen, “Hierar- chical architecture and feature mixing for ego-motion estimation using automotive radar,” inICMIM 2024; 7th IEEE MTT Conference. VDE, 2024, pp. 99–102

    work page 2024

  43. [43]

    DelftBlue Supercomputer (Phase 2),

    Delft High Performance Computing Centre (DHPC), “DelftBlue Supercomputer (Phase 2),” https://www.tudelft.nl/dhpc/ark: /44463/DelftBluePhase2, 2024

    work page 2024

  44. [44]

    Tracking of multiple static and dynamic targets for 4d automotive millimeter- wave radar point cloud in urban environments,

    B. Tan, Z. Ma, X. Zhu, S. Li, L. Zheng, L. Huang, and J. Bai, “Tracking of multiple static and dynamic targets for 4d automotive millimeter- wave radar point cloud in urban environments,”Remote Sensing, vol. 15, no. 11, p. 2923, 2023

    work page 2023

  45. [45]

    A density-based algorithm for discovering clusters in large spatial databases with noise,

    M. Ester, H.-P. Kriegel, J. Sander, X. Xuet al., “A density-based algorithm for discovering clusters in large spatial databases with noise,” inkdd, vol. 96, no. 34, 1996, pp. 226–231

    work page 1996

  46. [46]

    Improving the hungarian assignment algorithm,

    R. Jonker and T. V olgenant, “Improving the hungarian assignment algorithm,”Operations research letters, vol. 5, no. 4, pp. 171–175, 1986

    work page 1986

  47. [47]

    Landmark-based radar slam for autonomous driving,

    A. N. Ramesh, C. M. Le ´on, J. C. Zafra, S. Br ¨uggenwirth, and M. A. Gonz´alez-Huici, “Landmark-based radar slam for autonomous driving,” in2021 21st International Radar Symposium (IRS). IEEE, 2021, pp. 1–10

    work page 2021

  48. [48]

    Phd filter based traffic target tracking framework with fmcw radar,

    X. Cao, C. Zhu, and W. Yi, “Phd filter based traffic target tracking framework with fmcw radar,” in2022 11th International Conference on Control, Automation and Information Sciences (ICCAIS). IEEE, 2022, pp. 468–475

    work page 2022

  49. [49]

    Point transformer,

    H. Zhao, L. Jiang, J. Jia, P. H. Torr, and V . Koltun, “Point transformer,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 16 259–16 268

    work page 2021

  50. [50]

    A new height- estimation method using fmcw radar doppler beam sharpening,

    A. Laribi, M. Hahn, J. Dickmann, and C. Waldschmidt, “A new height- estimation method using fmcw radar doppler beam sharpening,” in2017 25th European Signal Processing Conference (EUSIPCO). IEEE, 2017, pp. 1932–1396. BIOGRAPHYSECTION Simin Zhureceived his BSc degree in Electrical Engineering and Automation from the Central South University in 2016. Afterw...

    work page 2017

  51. [51]

    Micro-Doppler Radar and Its Applications

    From September 1994 through 1996 he was with Technical University of Ilmenau, Germany as a Visiting Researcher. Since 1999 he is with the Delft University of Technology, the Netherlands. Since 2009 he leads there a chair of Microwave Sensing, Systems and Signals. His main research interests are in high-resolution radar, microwave imaging and applied elect...

    work page 1994