pith. sign in

arxiv: 2607.02074 · v1 · pith:DAKBI3ZZnew · submitted 2026-07-02 · 💻 cs.CV

Comprehensive Robustness Analysis of LiDAR-based 3D Object Detection in Autonomous Driving

Adwait Chandorkar , Kai Krink , Yerdana Maulenbay , Hasan Tercan , Tobias Meisen This is my paper

Pith reviewed 2026-07-03 15:52 UTC · model grok-4.3

classification 💻 cs.CV
keywords LiDAR3D object detectionadversarial robustnessautonomous drivingvoxel-based detectorspillar-based detectorspoint cloud
0
0 comments X

The pith

High-capacity voxel-based detectors prove more susceptible to structured coordinate perturbations than pillar-based detectors in LiDAR 3D object detection.

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

The paper establishes a new evaluation framework for adversarial robustness in LiDAR-based 3D object detectors that goes beyond simple mAP scores. It incorporates two structural factors—point cloud density and localization—and three predictive factors: misclassification, localization error, and distance from ego. Using this, the authors test recent and legacy models against LiDAR-specific adversarial attacks. They find that voxel-based high-capacity models are more vulnerable to coordinate perturbations, while non-anchor-based detectors show poor robustness overall. This suggests that current accuracy-focused designs do not automatically improve security against attacks, calling for benchmarks that reward robustness as well.

Core claim

High-capacity voxel-based detectors are more susceptible to structured coordinate perturbations than pillar-based detectors, and non-anchor-based detectors demonstrate poor adversarial robustness, indicating that recent models remain as vulnerable as their predecessors.

What carries the argument

A holistic robustness evaluation framework using structural factors (point cloud density and localization) and predictive factors (misclassification, localization error, distance from ego).

Load-bearing premise

The adversarial attacks used are representative of realistic threats to deployed autonomous driving systems.

What would settle it

Observing whether the same vulnerability patterns hold when testing the models against physical-world LiDAR spoofing attacks or natural perturbations in real driving data.

Figures

Figures reproduced from arXiv: 2607.02074 by Adwait Chandorkar, Hasan Tercan, Kai Krink, Tobias Meisen, Yerdana Maulenbay.

Figure 1
Figure 1. Figure 1: Methodology [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Visualization of inner and outer PCs for a Car in nuScenes. Impact of Point Cloud localization. In order to assess the maximum impact of the modifications to the PC, we split the PCs within the ground truth box into two categories before generating the adversarial PC. As shown in [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Visualization of average change (△) in false positives (FP) and false negatives (FN) for each model on nuScenes and Waymo datasets. is disturbed, resulting in higher misclassifications. On the bright side, we ob￾serve that although the PN model uses CenterHead as detection head, which demonstrated poor resilience against NE attack, PN exhibits strong robustness against NE attack. This underlines that pilla… view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of geometric errors for each model|attack com￾bination on nuScenes. determine the directional bias of the induced errors. Our results indicate that these attacks are predominantly erasure-dominant, resulting in a state of systemic blind￾ness. The transformer-based FF designed to mitigate FN in high capacity detectors, in contrast, generates high FNs in PB attack. Although FF leverages long-range… view at source ↗
Figure 5
Figure 5. Figure 5: Impact on ASR for near-far objects for each model-attack pair. The proximity of the objects is represented as : Near, : Mid and : Far. sequently, a failure of 3D-OD propagates through the stack, potentially leading to risky planning decisions. 4.3 Impact on Distance Since the PC density has inverse correlation to the proximity of object from ego, it is hypothesized that objects in the immediate vicinity, c… view at source ↗
Figure 6
Figure 6. Figure 6: Visualization of ASR vs point cloud density of models under different attacks. 0.0 0.2 0.4 0.6 0.8 1.0 ASR PP | PA CP | PD FF | PD CP | PB FF | PB ASR vs Inner/Outer Change Inner Outer nuScenes Waymo [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Impact of PC lo￾calization on ASR. context, density is defined as the cardinal point count within an object’s (e.g. Car ) volumetric bounds fol￾lowing adversarial manipulation. We observe that few model-attack pairs (FF-PB, CP-PB and PP-PA) report density-invariant susceptibility, analogous to observa￾tions in Tab. 2. We observe a paradoxical inverse corre￾lation between PC density and adversarial vulnerab… view at source ↗
Figure 8
Figure 8. Figure 8: Mean of ratio between inner and outer point clouds across all objects for nuScenes and Waymo dataset. varying levels of LiDAR sparsity, we calibrate the threshold to equi-partition the points between the internal core and the peripheral volume. Empirical results in [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
read the original abstract

Recent advancements in LiDAR-only 3D object detection have demonstrated improved detection accuracy over benchmark datasets. However, the adversarial robustness of these models remains untested. Very few adversarial robustness studies exist for LiDAR-only 3D object detection and unfortunately, even they are limited to legacy models. Moreover, there is a systemic gap in the existing evaluation frameworks that rely simply on mAP ignoring other structural and predictive factors. To fill this gap, we propose a holistic framework that evaluates adversarial robustness using two structural factors (point cloud density and point cloud localization) and three predictive factors (misclassification, localization error, distance from ego). Using this framework, we perform an empirical study and critical analysis on recent and legacy state-of-the-art models using adversarial attacks specifically designed for LiDAR-based models. Our key finding is that high-capacity, voxel-based detectors are more susceptible to structured coordinate perturbations than pillar-based detectors. Additionally, non-anchor-based detectors demonstrate poor adversarial robustness, which necessitates rethinking model training techniques. Overall, our results demonstrate that recent models are as vulnerable to adversarial attacks as their predecessors. Therefore, we argue that there is a need to improve the evaluation benchmarks for 3D object detection that not only reward architectural modifications for improving detection accuracy, but also evaluate whether the design choices improve adversarial robustness.

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 / 1 minor

Summary. The manuscript proposes a holistic evaluation framework for adversarial robustness of LiDAR-based 3D object detectors that incorporates two structural factors (point cloud density and localization) and three predictive factors (misclassification, localization error, distance from ego). It conducts an empirical study on recent and legacy SOTA models using attacks specifically designed for LiDAR, reporting that high-capacity voxel-based detectors are more susceptible to structured coordinate perturbations than pillar-based detectors, that non-anchor-based detectors exhibit poor robustness, and that recent models remain as vulnerable as predecessors, thereby arguing for robustness-aware benchmarks beyond mAP.

Significance. If the attack implementations and controls are shown to be representative, the architecture-level distinctions (voxel vs. pillar, anchor vs. non-anchor) would provide actionable guidance for designing more robust detectors in safety-critical autonomous driving. The multi-factor framework is a constructive step beyond single-metric evaluations. The current lack of implementation details, however, leaves the mapping from reported perturbations to real-world risk unanchored.

major comments (2)
  1. [Abstract] Abstract and evaluation sections: the central claims rest on adversarial attacks 'specifically designed for LiDAR-based models,' yet no description of attack generation, physical realizability constraints (sensor spoofing, beam geometry), or comparison to natural sensor noise is supplied. This directly undermines the assertion that the observed vulnerabilities justify rethinking training techniques and that recent models are 'as vulnerable as their predecessors.'
  2. [Evaluation Framework] Evaluation framework and experimental setup: the manuscript provides no information on dataset splits, statistical testing procedures, or controls for confounding factors (e.g., model capacity, training data volume). Without these, the comparative findings on voxel-based vs. pillar-based susceptibility and non-anchor-based robustness cannot be verified as load-bearing.
minor comments (1)
  1. [Abstract] The abstract states 'systemic gap in the existing evaluation frameworks' without citing specific prior LiDAR robustness papers that were limited to legacy models; adding these references would clarify the novelty of the proposed factors.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the major comments below and will revise the manuscript to incorporate additional details where the current version is lacking.

read point-by-point responses

  1. Referee: [Abstract] Abstract and evaluation sections: the central claims rest on adversarial attacks 'specifically designed for LiDAR-based models,' yet no description of attack generation, physical realizability constraints (sensor spoofing, beam geometry), or comparison to natural sensor noise is supplied. This directly undermines the assertion that the observed vulnerabilities justify rethinking training techniques and that recent models are 'as vulnerable as their predecessors.'

    Authors: We agree that the manuscript does not currently supply a detailed description of attack generation, physical realizability constraints, or comparisons to natural sensor noise. In the revision we will add a dedicated subsection describing the LiDAR-specific attack implementations (including coordinate perturbation methods drawn from prior literature), discuss sensor spoofing and beam geometry considerations, and include a comparison of perturbation magnitudes to typical sensor noise levels. These additions will better support the claims about rethinking training techniques and the vulnerability of recent models relative to predecessors. revision: yes

  2. Referee: [Evaluation Framework] Evaluation framework and experimental setup: the manuscript provides no information on dataset splits, statistical testing procedures, or controls for confounding factors (e.g., model capacity, training data volume). Without these, the comparative findings on voxel-based vs. pillar-based susceptibility and non-anchor-based robustness cannot be verified as load-bearing.

    Authors: We acknowledge that the current text omits explicit details on dataset splits, statistical testing, and controls for confounding factors. The revised manuscript will specify the dataset splits used, describe any statistical procedures applied to the multi-factor robustness metrics, and discuss controls or accounting for model capacity and training data volume when interpreting the voxel/pillar and anchor/non-anchor distinctions. These additions will improve verifiability of the reported architecture-level findings. revision: yes

Circularity Check

0 steps flagged

No circularity; empirical robustness claims rest on direct evaluations, not derivations or self-referential fits.

full rationale

The paper is an empirical study that proposes a multi-factor evaluation framework and applies existing LiDAR-specific adversarial attacks to compare detector architectures. No equations, fitted parameters renamed as predictions, or self-citation chains appear in the provided text. The central claims (voxel vs. pillar susceptibility; non-anchor robustness) are presented as outcomes of those direct comparisons rather than reductions to inputs by construction. This matches the default case of a self-contained empirical paper with no load-bearing circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an empirical robustness study; the central claims rest on the validity of the five evaluation factors and the representativeness of the chosen LiDAR-specific attacks, which are treated as given rather than derived.

pith-pipeline@v0.9.1-grok · 5778 in / 1042 out tokens · 28322 ms · 2026-07-03T15:52:59.909869+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

52 extracted references · 49 canonical work pages · 2 internal anchors

  1. [1]

    Sensors25(19) (2025).https://doi.org/10.3390/s25196026,https: //www.mdpi.com/1424-8220/25/19/6026

    Abraha, T.M., Graham-Knight, J.B., Lasserre, P., Najjaran, H., Lucet, Y.: Convnet-generated adversarial perturbations for evaluating 3d object detection robustness. Sensors25(19) (2025).https://doi.org/10.3390/s25196026,https: //www.mdpi.com/1424-8220/25/19/6026

    work page doi:10.3390/s25196026 2025

  2. [2]

    In: 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2 (ASPLOS ’24)

    Ansel, J., Yang, E., He, H., Gimelshein, N., Jain, A., Voznesensky, M., Bao, B., Bell, P., Berard, D., Burovski, E., Chauhan, G., Chourdia, A., et al.: PyTorch 2: Faster Machine Learning Through Dynamic Python Bytecode Transformation and Graph Compilation. In: 29th ACM International Conference on Architectural Support for Programming Languages and Operati...

    work page doi:10.1145/3620665.3640366 2024

  3. [3]

    Aung, N.H.H., Sangwongngam, P., Jintamethasawat, R., Shah, S., Wuttisittikulkij, L.: A review of lidar-based 3d object detection via deep learning approaches to- wardsrobustconnectedandautonomousvehicles.IEEETransactionsonIntelligent Vehicles10(1), 526–547 (2025).https://doi.org/10.1109/TIV.2024.3415771

    work page doi:10.1109/tiv.2024.3415771 2025

  4. [4]

    Caesar, H., Bankiti, V., Lang, A.H., Vora, S., Liong, V.E., Xu, Q., Krishnan, A., Pan, Y., Baldan, G., Beijbom, O.: nuscenes: A multimodal dataset for autonomous driving (2020),https://doi.org/10.1109/CVPR42600.2020.01164

    work page doi:10.1109/cvpr42600.2020.01164 2020

  5. [5]

    In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision (ICCV) Workshops

    Chandorkar,A.,Tercan,H.,Meisen,T.:Rethinkingbackbonedesignforlightweight 3d object detection in lidar. In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision (ICCV) Workshops. pp. 1698–1706 (October 2025). https://doi.org/https://doi.org/10.48550/arXiv.2508.00744

    work page doi:10.48550/arxiv.2508.00744 2025

  6. [6]

    IEEE Transactions on Intelligent Transportation Systems25(11), 16118–16132 (2024)

    Chen, H., Yan, H., Yang, X., Su, H., Zhao, S., Qian, F.: Efficient adversarial at- tack strategy against 3d object detection in autonomous driving systems. IEEE Transactions on Intelligent Transportation Systems25(11), 16118–16132 (2024). https://doi.org/10.1109/TITS.2024.3410038

    work page doi:10.1109/tits.2024.3410038 2024

  7. [7]

    IEEE Transactions on Intelligent Trans- portation Systems26(9), 13563–13572 (2025).https://doi.org/10.1109/TITS

    Chen, J., Liao, D., Yan, Y., Xiang, S., Zheng, H.: Lidattack: Robust black-box attack on lidar-based object detection. IEEE Transactions on Intelligent Trans- portation Systems26(9), 13563–13572 (2025).https://doi.org/10.1109/TITS. 2025.3573055

    work page doi:10.1109/tits 2025

  8. [8]

    Peebles and S

    Chen, Y., Yu, Z., Chen, Y., Lan, S., Anandkumar, A., Jia, J., Alvarez, J.M.: FocalFormer3D : Focusing on Hard Instance for 3D Object Detection . In: 2023 IEEE/CVF International Conference on Computer Vision (ICCV). pp. 8360–8371. IEEE Computer Society, Los Alamitos, CA, USA (Oct 2023).https://doi.org/ 10.1109/ICCV51070.2023.00771,https://doi.ieeecomputers...

    work page doi:10.1109/iccv51070.2023.00771 2023

  9. [9]

    2023.02076

    Chen, Y., Liu, J., Zhang, X., Qi, X., Jia, J.: Voxelnext: Fully sparse voxelnet for 3d object detection and tracking (2023).https://doi.org/10.1109/CVPR52729. 2023.02076

    work page doi:10.1109/cvpr52729 2023

  10. [10]

    Contributors, M.: MMDetection3D: OpenMMLab next-generation platform for general 3D object detection.https://github.com/open- mmlab/mmdetection3d (2020)

    2020

  11. [12]

    Accident Analysis & Prevention207, 107742 (2024).https://doi.org/ https://doi.org/10.1016/j.aap.2024.107742,https://www.sciencedirect

    Du, B., Zhang, C., Sarkar, A., Shen, J., Telikani, A., Hu, H.: Identifying factors related to pedestrian and cyclist crashes in act, australia with an extended crash dataset. Accident Analysis & Prevention207, 107742 (2024).https://doi.org/ https://doi.org/10.1016/j.aap.2024.107742,https://www.sciencedirect. com/science/article/pii/S0001457524002872

    work page doi:10.1016/j.aap.2024.107742 2024

  12. [13]

    Erabati, G.K., Araujo, H.: Li3detr: A lidar based 3d detection transformer (2022), https://arxiv.org/abs/2210.15365

    work page arXiv 2022

  13. [14]

    In: IEEE Conf

    Eskandar, G.: An empirical study of the generalization ability of lidar 3d object detectors to unseen domains. In: 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 23815–23825 (2024).https://doi.org/10. 1109/CVPR52733.2024.02248

    work page arXiv 2024

  14. [15]

    Gamerdinger, J., Teufel, S., Amann, S., Bringmann, O.: Criticality metrics for relevanceclassificationinsafetyevaluationofobjectdetectioninautomateddriving (2025),https://arxiv.org/abs/2512.15181

    work page arXiv 2025

  15. [16]

    International Journal of Robotics Research (IJRR)32(11), 1231–1237 (2013).https://doi.org/https://dl.acm.org/doi/10.1177/02783649134912

    Geiger, A., Lenz, P., Stiller, C., Urtasun, R.: Vision meets robotics: The KITTI dataset. International Journal of Robotics Research (IJRR)32(11), 1231–1237 (2013).https://doi.org/https://dl.acm.org/doi/10.1177/02783649134912

    work page doi:10.1177/02783649134912 2013

  16. [17]

    Explaining and Harnessing Adversarial Examples

    Goodfellow, I.J., Shlens, J., Szegedy, C.: Explaining and harnessing adversarial examples. arXiv preprint arXiv:1412.6572 (2014).https://doi.org/10.48550/ arXiv.1412.6572

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  17. [18]

    International Journal of Intelligent Transportation Systems Research (12 2025).https://doi.org/https://doi.org/10.1007/s13177-025- 00598-2

    Hariya, K., Inoshita, H., Fukuda, Y., Yoneda, K., Suganuma, N.: Real-time 3d object detection with distance-aware hybrid point cloud representation toward long range detection. International Journal of Intelligent Transportation Systems Research (12 2025).https://doi.org/https://doi.org/10.1007/s13177-025- 00598-2

    work page doi:10.1007/s13177-025- 2025

  18. [19]

    Peebles and S

    Kong, L., Liu, Y., Li, X., Chen, R., Zhang, W., Ren, J., Pan, L., Chen, K., Liu, Z.: Robo3d: Towards robust and reliable 3d perception against corruptions. In: Pro- ceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 19994–20006 (October 2023),https://doi.ieeecomputersociety.org/10. 1109/ICCV51070.2023.01830

    work page arXiv 2023

  19. [20]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Lang, A.H., Vora, S., Caesar, H., Zhou, B., Yang, J., Beijbom, O.: Pointpillars: Fast encoders for object detection from point clouds. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 12697–12705 (2019). https://doi.org/10.1109/CVPR.2019.01298

    work page doi:10.1109/cvpr.2019.01298 2019

  20. [21]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Li, J., Luo, C., Yang, X.: Pillarnext: Rethinking network designs for 3d ob- ject detection in lidar point clouds. In: 2023 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR). pp. 17567–17576 (2023).https: //doi.org/10.1109/CVPR52729.2023.01685

    work page doi:10.1109/cvpr52729.2023.01685 2023

  21. [22]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Liu, Z., Mao, H., Wu, C.Y., Feichtenhofer, C., Darrell, T., Xie, S.: A convnet for the 2020s. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 11976–11986 (June 2022).https://doi.org/ 10.1109/CVPR52688.2022.01186

    work page doi:10.1109/cvpr52688.2022.01186 2022

  22. [23]

    IET Computer Vision19(1), e70011 (2025)

    Long, H., Chen, H., Xu, M., Zhang, C., Qian, F.: Crafting transferable adversarial examples against 3d object detection. IET Computer Vision19(1), e70011 (2025). https://doi.org/https://doi.org/10.1049/cvi2.70011

    work page doi:10.1049/cvi2.70011 2025

  23. [24]

    IEEE Transactions on Vehicular Technology74(1), 292–305 (2025)

    Lu, Y., Hao, X., Li, Y., Chai, W., Sun, S., Velipasalar, S.: Range-aware atten- tion network for lidar-based 3d object detection with auxiliary point density level estimation. IEEE Transactions on Vehicular Technology74(1), 292–305 (2025). https://doi.org/10.1109/TVT.2024.3454607 18 A. Chandorkar et al

    work page doi:10.1109/tvt.2024.3454607 2025

  24. [25]

    Madry, A., Makelov, A., Schmidt, L., Tsipras, D., Vladu, A.: Towards deep learning models resistant to adversarial attacks (2019),https://arxiv.org/abs/1706. 06083

    2019

  25. [26]

    Towards Deep Learning Models Resistant to Adversarial Attacks

    Madry, A., Makelov, A., Schmidt, L., Tsipras, D., Vladu, A.: Towards deep learning models resistant to adversarial attacks. arXiv preprint arXiv:1706.06083 (2017). https://doi.org/10.48550/arXiv.1706.06083

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1706.06083 2017

  26. [27]

    IEEE Transactions on Intelligent Vehicles10(4), 2905–2914 (2025).https://doi.org/10.1109/TIV

    Mao,W.,Wang,T.,Zhang,D.,Yan,J.,Yoshie,O.:Pillarnest:Embracingbackbone scaling and pretraining for pillar-based 3d object detection. IEEE Transactions on Intelligent Vehicles10(4), 2905–2914 (2025).https://doi.org/10.1109/TIV. 2024.3386576

    work page doi:10.1109/tiv 2025

  27. [28]

    Accident Analysis & Pre- vention184, 106996 (2023).https://doi.org/https://doi.org/10.1016/ j.aap.2023.106996,https://www.sciencedirect.com/science/article/pii/ S000145752300043X

    Scarano, A., Aria, M., Mauriello, F., Riccardi, M.R., Montella, A.: Systematic literature review of 10 years of cyclist safety research. Accident Analysis & Pre- vention184, 106996 (2023).https://doi.org/https://doi.org/10.1016/ j.aap.2023.106996,https://www.sciencedirect.com/science/article/pii/ S000145752300043X

    work page arXiv 2023

  28. [29]

    In: European Conference on Computer Vision

    Shi, G., Li, R., Ma, C.: Pillarnet: Real-time and high-performance pillar-based 3d object detection. In: European Conference on Computer Vision. pp. 35–52 (2022). https://doi.org/https://doi.org/10.1007/978-3-031-20080-9_3

    work page doi:10.1007/978-3-031-20080-9_3 2022

  29. [30]

    In: IEEE Conf

    Shi, S., Guo, C., Jiang, L., Wang, Z., Shi, J., Wang, X., Li, H.: Pv-rcnn: Point-voxel feature set abstraction for 3d object detection. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 10526–10535 (2020). https://doi.org/10.1109/CVPR42600.2020.01054

    work page doi:10.1109/cvpr42600.2020.01054 2020

  30. [31]

    Shi, S., Jiang, L., Deng, J., Wang, Z., Guo, C., Shi, J., Wang, X., Li, H.: Pv- rcnn++: Point-voxel feature set abstraction with local vector representation for 3d object detection (2022),https://doi.org/10.48550/arXiv.2102.00463

    work page doi:10.48550/arxiv.2102.00463 2022

  31. [32]

    In: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Shi, S., Wang, X., Li, H.: Pointrcnn: 3d object proposal generation and detection from point cloud. In: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 770–779 (2019).https://doi.org/10.1109/CVPR.2019. 00086

    work page doi:10.1109/cvpr.2019 2019

  32. [33]

    Shi,S.,Wang,Z.,Shi,J.,Wang,X.,Li,H.:Frompointstoparts:3dobjectdetection from point cloud with part-aware and part-aggregation network (2020),https: //doi.org/10.48550/arXiv.1907.03670

    work page doi:10.48550/arxiv.1907.03670 2020

  33. [34]

    IEEE Transactions on Intelligent Transportation Systems25(11), 15407– 15436 (2024).https://doi.org/10.1109/TITS.2024.3439557

    Song, Z., Liu, L., Jia, F., Luo, Y., Jia, C., Zhang, G., Yang, L., Wang, L.: Robustness-aware 3d object detection in autonomous driving: A review and out- look. IEEE Transactions on Intelligent Transportation Systems25(11), 15407– 15436 (2024).https://doi.org/10.1109/TITS.2024.3439557

    work page doi:10.1109/tits.2024.3439557 2024

  34. [35]

    In: Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR) (June 2020).https://doi.org/http: //ieeexplore.ieee.org/document/9156973/

    Sun, P., Kretzschmar, H., Dotiwalla, X., Chouard, A., Patnaik, V., Tsui, P., Guo, J., Zhou, Y., Chai, Y., Caine, B., Vasudevan, V., Han, W., Ngiam, J., Zhao, H., Timofeev, A., Ettinger, S., Krivokon, M., Gao, A., Joshi, A., Zhang, Y., Shlens, J., Chen, Z., Anguelov, D.: Scalability in perception for autonomous driving: Waymo open dataset. In: Proceedings ...

    work page arXiv 2020

  35. [36]

    In: IEEE Conf

    Tu, J., Ren, M., Manivasagam, S., Liang, M., Yang, B., Du, R., Cheng, F., Urtasun, R.: Physically realizable adversarial examples for lidar object detection. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 13713–13722 (2020).https://doi.org/10.1109/CVPR42600.2020.01373

    work page doi:10.1109/cvpr42600.2020.01373 2020

  36. [37]

    In: European conference on computer vision

    Wang, J., Meng, Q., Liu, G., Yan, L., Wang, K., Cheng, M.M., Hou, Q.: Towards stable 3d object detection. In: European conference on computer vision. Springer (2024).https://doi.org/https://doi.org/10.1007/978-3-031-72973-7_12 Robustness Analysis of LiDAR-based 3D-OD in Autonomous Driving 19

    work page doi:10.1007/978-3-031-72973-7_12 2024

  37. [38]

    IEEE Internet of Things Journal12(12), 21404–21414 (2025).https://doi.org/10.1109/JIOT

    Wang, Y., Wu, L., Jin, J., Wang, E., Zhang, Z., Zhao, Y.: An imperceptible adver- sarial attack against 3-d object detectors in autonomous driving. IEEE Internet of Things Journal12(12), 21404–21414 (2025).https://doi.org/10.1109/JIOT. 2025.3547966

    work page doi:10.1109/jiot 2025

  38. [39]

    In: Proceedings of the 37th International Conference on Neural Infor- mation Processing Systems

    Wang, Z., Li, Y., Chen, X., Zhao, H., Wang, S.: Uni3detr: unified 3d detection transformer. In: Proceedings of the 37th International Conference on Neural Infor- mation Processing Systems. NIPS ’23, Curran Associates Inc., Red Hook, NY, USA (2023).https://doi.org/https://dl.acm.org/doi/10.5555/3666122.3667855

    work page doi:10.5555/3666122.3667855 2023

  39. [40]

    In: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Xiang, C., Qi, C.R., Li, B.: Generating 3d adversarial point clouds. In: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 9128–9136 (2019).https://doi.org/10.1109/CVPR.2019.00935

    work page doi:10.1109/cvpr.2019.00935 2019

  40. [41]

    Xiang, X., Zhang, J.: Fusionvit: Hierarchical 3d object detection via lidar-camera vision transformer fusion (2023),https://arxiv.org/abs/2311.03620

    work page arXiv 2023

  41. [42]

    arXiv preprint arXiv:2404.07495 (2024),https://doi.org/10.48550/arXiv.2404.07495

    Xu, W., Zhou, S., Yuan, Z.: Pillartrack: Redesigning pillar-based transformer net- work for single object tracking on point clouds. arXiv preprint arXiv:2404.07495 (2024),https://doi.org/10.48550/arXiv.2404.07495

    work page doi:10.48550/arxiv.2404.07495 2024

  42. [43]

    Sen- sors18(10), 3337 (2018).https://doi.org/10.3390/s18103337

    Yan, Y., Mao, Y., Li, B.: Second: Sparsely embedded convolutional detection. Sen- sors18(10), 3337 (2018).https://doi.org/10.3390/s18103337

    work page doi:10.3390/s18103337 2018

  43. [44]

    In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR)

    Yan, Z., Lin, Y., Wang, K., Zheng, Y., Wang, Y., Zhang, Z., Li, J., Yang, J.: Tri-perspective view decomposition for geometry-aware depth completion. In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR). pp. 4874–4884 (June 2024).https://doi.org/https://doi.org/ 10.48550/arXiv.2403.15008

    work page doi:10.48550/arxiv.2403.15008 2024

  44. [45]

    In: IEEE Conf

    Yang,Z.,Sun,Y.,Liu,S.,Jia,J.:3dssd:Point-based3dsinglestageobjectdetector. In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 11037–11045 (2020).https://doi.org/10.1109/CVPR42600.2020. 01105

    work page doi:10.1109/cvpr42600.2020 2020

  45. [46]

    Zee- shan Zia, and Quoc-Huy Tran

    Yin, T., Zhou, X., Krähenbühl, P.: Center-based 3d object detection and tracking. In: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 11779–11788 (2021).https://doi.org/10.1109/CVPR46437.2021. 01161

    work page doi:10.1109/cvpr46437.2021 2021

  46. [47]

    Zhang, P., Li, X., Lin, X., He, L.: A new literature review of 3d object detection on autonomous driving. J. Artif. Int. Res.82(Jun 2025),https://doi.org/10. 1613/jair.1.15961

    2025

  47. [48]

    Zhang, Y., Hou, J., Yuan, Y.: A comprehensive study of the robustness for lidar- based 3d object detectors against adversarial attacks (2023),https://doi.org/ 10.1007/s11263-023-01934-3

    work page doi:10.1007/s11263-023-01934-3 2023

  48. [49]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Zhou, C., Zhang, Y., Chen, J., Huang, D.: Octr: Octree-based transformer for 3d object detection. In: 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 5166–5175 (2023).https://doi.org/10.1109/ CVPR52729.2023.00500

    work page arXiv 2023

  49. [50]

    Zhou, S., Tian, Z., Chu, X., Zhang, X., Zhang, B., Lu, X., Feng, C., Jie, Z., Chiang, P.Y., Ma, L.: Fastpillars: A deployment-friendly pillar-based 3d detector (2023), https://arxiv.org/abs/2302.02367

    work page arXiv 2023

  50. [51]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Zhou, Y., Tuzel, O.: Voxelnet: End-to-end learning for point cloud based 3d object detection. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 4490–4499 (2018).https://doi.org/10.1109/CVPR.2018.00472

    work page doi:10.1109/cvpr.2018.00472 2018

  51. [52]

    IEEE Transactions on Intelligent Transportation Systems (2023)

    Zhu, J., et al.: Vulnerability analysis of lidar-based 3d object detection to adversar- ial lidar points. IEEE Transactions on Intelligent Transportation Systems (2023). https://doi.org/10.1109/TITS.2023.3274635 20 A. Chandorkar et al

    work page doi:10.1109/tits.2023.3274635 2023

  52. [53]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Zhu, Z., Zhang, Y., Chen, H., Dong, Y., Zhao, S., Ding, W., Zhong, J., Zheng, S.: Understanding the robustness of 3d object detection with bird’view rep- resentations in autonomous driving. In: 2023 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR). pp. 21600–21610 (2023).https: //doi.org/10.1109/CVPR52729.2023.02069

    work page doi:10.1109/cvpr52729.2023.02069 2023