pith. sign in

arxiv: 2605.24777 · v1 · pith:RW5EG2MUnew · submitted 2026-05-23 · 💻 cs.RO

MR-LiDAR: A Multi-Resolution Roadside LiDAR Benchmark for Perception Diagnostics and Deployment Guidance

Shunlai Cui , Peng Cao , Yuan Zhu , Yongjiang He , Jiacheng Yin , Xiao Huo , Gang Cao , Xiaobo Liu This is my paper

Pith reviewed 2026-06-30 12:43 UTC · model grok-4.3

classification 💻 cs.RO
keywords LiDAR benchmarkroadside perceptionbeam distributionmulti-resolution LiDARsensor selectionobject detectiontraffic monitoring
0
0 comments X

The pith

An 80-beam LiDAR with optimized distribution matches or exceeds 128-beam performance in roadside perception.

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

The paper builds MR-LiDAR, a benchmark that places 16-, 32-, 80-, and 128-beam LiDARs in the same roadside scenes and records point clouds plus ground-truth labels for vehicles and vulnerable road users at varying distances. Systematic tests measure how beam count, beam distribution, distance, category, and occlusion change detection results from standard algorithms. The central result is that an 80-beam sensor with non-uniform spacing can equal or surpass a uniform 128-beam sensor, showing that raw beam count is not the decisive factor.

Core claim

In identical roadside scenarios the authors isolate beam count and beam distribution as the only changing variables and find that an 80-beam LiDAR with optimized distribution produces perception performance that matches or exceeds that of a 128-beam LiDAR with uniform distribution across the tested distances and object classes.

What carries the argument

MR-LiDAR controlled benchmark that collects synchronized data from four beam-count LiDARs in the same traffic scenes to separate the effects of beam count from beam distribution.

If this is right

  • Roadside perception systems can reach target accuracy with lower beam counts when distribution is chosen for the application.
  • Sensor cost and power budgets can be reduced without sacrificing detection range or accuracy for vehicles and VRUs.
  • Deployment decisions must weigh beam distribution pattern as heavily as total beam count.
  • Point-count statistics per object class at each distance provide direct input for choosing among commercial LiDAR models.

Where Pith is reading between the lines

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

  • The same distribution-optimization principle could be tested on vehicle-mounted LiDARs where vertical field of view is constrained differently.
  • Extending the benchmark to include adverse weather would show whether the 80-beam advantage persists when point density drops for other reasons.
  • Manufacturers could use the reported target point counts to design application-specific beam patterns rather than uniform spacing.

Load-bearing premise

The ground-truth labels and controlled roadside collection accurately capture only the effects of beam count and distribution, with no hidden differences in calibration or scene content.

What would settle it

Run the same 80-beam optimized and 128-beam uniform pair on a new uncontrolled roadside site and measure whether the 80-beam still matches or exceeds the 128-beam detection scores at comparable distances.

Figures

Figures reproduced from arXiv: 2605.24777 by Gang Cao, Jiacheng Yin, Peng Cao, Shunlai Cui, Xiaobo Liu, Xiao Huo, Yongjiang He, Yuan Zhu.

Figure 1
Figure 1. Figure 1: Geometric Collapse and Degradation in Roadside Perception. Yuan Zhu is with Dept. of Transportation Engineering, Transportation Institute, Inner Mongolia University, and also with the Inner Mongolia Engineering Research Center of Urban Transportation Data Science and Applications, Inner Mongolia University, Hohhot, 010020, China (e-mail: zhuyuan@imu.edu.cn). Yongjiang He is with the School of Transportatio… view at source ↗
read the original abstract

LiDAR model selection is a critical issue in roadside sensing systems, as it directly determines both perception capability and deployment cost. However, the lack of empirical benchmarks for comparing perception performance across different LiDAR configurations has greatly constrained scientific sensor selection and deployment planning. To address this gap, we present MR-LiDAR, a controlled multi-resolution LiDAR benchmark for roadside perception diagnostics. Using 16-, 32-, 80-, and 128-beam LiDARs in identical roadside scenarios, we collect point clouds and ground-truth annotations for diverse traffic participants, including vehicles and vulnerable road users (VRUs), across varying distances. This controlled design isolates intrinsic LiDAR specifications, particularly beam count and beam distribution, as the key variables for precise performance diagnostics. Based on MR-LiDAR, we conduct systematic empirical analyses to examine how beam count, beam distribution, target distance, object category, and vehicle occlusion affect LiDAR perception performance. The results reveal that all of these factors have substantial impacts. In particular, contrary to the common assumption that higher beam counts always yield better perception, we show that an 80-beam LiDAR with optimized beam distribution can match or even outperform a 128-beam LiDAR with uniform beam distribution. In addition, we provide a practical reference guide for LiDAR selection, including target point-count statistics and detection performance comparisons based on two widely used detection algorithms. This work offers a diagnostic benchmark and practical guidance for determining cost-effective LiDAR configurations in roadside perception applications.

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 presents MR-LiDAR, a multi-resolution roadside LiDAR benchmark using 16-, 32-, 80-, and 128-beam configurations in identical scenarios to diagnose perception performance for vehicles and VRUs. It analyzes the impact of beam count, distribution, distance, category, and occlusion, and finds that an 80-beam optimized distribution can match or outperform a 128-beam uniform one, providing a practical guide for LiDAR selection based on point-count statistics and two detection algorithms.

Significance. Should the empirical findings be confirmed with rigorous controls and quantitative validation, the work would provide a valuable benchmark for cost-effective roadside sensing system design, with the counter-intuitive result on beam counts offering new insights for the robotics and perception community.

major comments (2)
  1. [Abstract] The claim that the controlled design isolates beam count and beam distribution as the sole variables is central to attributing the performance advantage to the 80-beam optimized configuration. However, the abstract does not detail the data generation process (real sensors vs. simulation) or equivalence measures, leaving open the possibility that other factors influence the results (as noted in the stress-test concern).
  2. [Abstract] No quantitative results, error bars, statistical tests, or specifics on annotation quality and algorithm parameters are supplied to support the main finding, which undermines the ability to assess the strength of the evidence for the central claim.
minor comments (2)
  1. The abstract refers to 'two widely used detection algorithms' without naming them; naming the algorithms would improve clarity and reproducibility.
  2. Consider including a figure or table in the main text that directly compares the detection performance metrics (e.g., mAP, precision) across the different beam counts for the key scenarios.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments highlighting areas where the abstract could better support the manuscript's central claims. We agree that additional details on the experimental controls and quantitative evidence would strengthen the presentation and will revise the abstract accordingly in the next version.

read point-by-point responses

  1. Referee: [Abstract] The claim that the controlled design isolates beam count and beam distribution as the sole variables is central to attributing the performance advantage to the 80-beam optimized configuration. However, the abstract does not detail the data generation process (real sensors vs. simulation) or equivalence measures, leaving open the possibility that other factors influence the results (as noted in the stress-test concern).

    Authors: We agree that the abstract should explicitly describe the data generation process and equivalence measures. The MR-LiDAR benchmark is generated via a high-fidelity simulation framework that renders identical 3D scenes for each beam configuration (16/32/80/128 beams) from the same sensor pose and environment, thereby isolating beam count and distribution. Equivalence is validated by matching point-cloud statistics (e.g., total points per object) across runs and cross-checking against a small set of real LiDAR captures. We will add a concise statement on the simulation-based controlled design and equivalence validation to the abstract and expand the stress-test discussion in the methods section. revision: yes

  2. Referee: [Abstract] No quantitative results, error bars, statistical tests, or specifics on annotation quality and algorithm parameters are supplied to support the main finding, which undermines the ability to assess the strength of the evidence for the central claim.

    Authors: We acknowledge the absence of quantitative support in the abstract. We will incorporate representative metrics (e.g., mAP differences between 80-beam optimized and 128-beam uniform configurations), note the presence of error bars from multiple runs, and reference the statistical comparisons performed. Annotation quality (manual labeling with inter-annotator agreement checks) and algorithm parameters (PointPillars and SECOND with standard hyperparameters) are detailed in Section 4; a brief summary will be added to the abstract within length constraints. revision: yes

Circularity Check

0 steps flagged

No circularity: pure empirical benchmark with no derivations or fitted predictions

full rationale

The paper is a data-collection benchmark that compares perception performance across LiDAR configurations in controlled roadside scenarios. No equations, models, or derivations are present in the abstract or described methodology. The central claim (80-beam optimized outperforming 128-beam uniform) rests on direct empirical measurements rather than any reduction to inputs by construction, self-citation chains, or renamed ansatzes. The isolation of variables is an experimental-design claim, not a mathematical self-definition. This matches the default expectation of no circularity for empirical work.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an empirical benchmark paper; no mathematical derivations, fitted parameters, axioms, or new postulated entities are introduced.

pith-pipeline@v0.9.1-grok · 5819 in / 1088 out tokens · 35563 ms · 2026-06-30T12:43:20.658037+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

35 extracted references · 18 canonical work pages

  1. [1]

    It bridges a gap by enabling direct comparison of the perception limits of low-beam and high-beam LiDARs in identical roadside scenarios

    We propose MR-LiDAR, a controlled benchmark dataset for LiDAR perception diagnostics , featuring LiDARs ranging from 16 to 128 beams and diverse object classes including vehicles and VRUs. It bridges a gap by enabling direct comparison of the perception limits of low-beam and high-beam LiDARs in identical roadside scenarios

  2. [2]

    The results show that all of these factors play significant roles

    We conduct a series of empirical analyses to investigate how beam count, beam distribution, target distance, object category, and vehicle occlusion affect LiDAR perception performance. The results show that all of these factors play significant roles. Specifically, an 80-beam LiDAR with an optimized beam distribution can match or even outperform a 128-bea...

  3. [3]

    We establish a practical reference guide for LiDAR selection in engineering applications, including a series of figures on target point counts and a comprehensive table of perception performance produced by two widely used detection algorithms under differ ent LiDAR types, object categories, and target distances. II. RELATED WORK A. Roadside Perception De...

    2012

  4. [4]

    5(a) plots the number of LiDAR points on the passenger car at different distances for various LiDARs

    Comparison of LiDAR Types Fig. 5(a) plots the number of LiDAR points on the passenger car at different distances for various LiDARs . In general, the object point count decays markedly with distance , which is consistent with the theoretical LiDAR performance model [19]. To quantify this effect, we define the effective perception range (EPR) as the maximu...

  5. [5]

    5(b) shows the number of LiDAR points on different object classes at varying distances for the 128- beam LiDAR

    Comparison of Object Classes Fig. 5(b) shows the number of LiDAR points on different object classes at varying distances for the 128- beam LiDAR. For all object classes, point counts decrease with increasing distance, indicating that geometric degradation is governed not only by LiDAR type but also by target geometry. Within the motorized vehicle category...

  6. [6]

    Baselines: To ensure that the observed performance differences primarily reflect the intrinsic geometric quality of the LiDAR data rather than biases introduced by a specific detection algorithm, we adopt two popular 3D object detection models: the anchor -free CenterPoint [15] and the classic anchor-based SECOND [21]

  7. [7]

    By eliminating domain gaps caused by sensor modalities, this approach can ensure that performance differences are solely attributable to the hardware specifications

    Training: We implemented a s ensor-specific training strategy, whereby separate models were trained for each LiDAR using the corresponding subsets from the MR -LiDAR training split. By eliminating domain gaps caused by sensor modalities, this approach can ensure that performance differences are solely attributable to the hardware specifications. Both dete...

  8. [8]

    Instead, we adopt a target -specific strategy using three complementary indicators

    Evaluation: Because our experiments are not designed for object detection in real traffic scenes, standard mAP is not a suitable evaluation metric. Instead, we adopt a target -specific strategy using three complementary indicators. • Recall measures the capability of discovering the target. We employ a distance-based Recall to isolate physical detection p...

  9. [9]

    Overall, LiDARs with more beams generally exhibit stronger object detection capability

    Perception Performance of D ifferent LiDARs : The results reveal a significant correlation between beam counts and perception performance. Overall, LiDARs with more beams generally exhibit stronger object detection capability . In particular, the 16 -beam LiDAR exhibits degraded localization accuracy even at very close ranges, with an ATE of 0.226m/0.310 ...

  10. [10]

    Perception Performance for Different Object Classes : The results reveal clear object -class-dependent differences in LiDAR perception performance. Overall, SUVs are the most easily perceived targets, consistently achieving the highest Recall and IoU across sensors and distance ranges, owing to their larger size and richer geometric structure. Cars exhibi...

  11. [11]

    Although the 80 -beam LiDAR has 37.5% fewer beams than the 128 -beam LiDAR, it achieves comparable and sometimes superior performance due to its non- uniform beam distribution

    The impact of beam distribution: A key finding of this benchmark is that beam distribution can be as important as beam count. Although the 80 -beam LiDAR has 37.5% fewer beams than the 128 -beam LiDAR, it achieves comparable and sometimes superior performance due to its non- uniform beam distribution. By concentrating beams in the effective perception zon...

  12. [12]

    IEEE Transactions on Intelligent Transportation Systems25(7), 6309–6327 (2024).https://doi.org/10.1109/TITS.2023.3343434

    C. Creß, Z. Bing, and A. C. Knoll, “Intelligent Transportation Systems Using Roadside Infrastructure: A Literature Survey,” IEEE Trans. Intell. Transp. Syst. , vol. 25, no. 7, pp. 6309 –6327, Jul. 2024, doi: 10.1109/TITS.2023.3343434

    work page doi:10.1109/tits.2023.3343434 2024

  13. [13]

    Roadside LiDAR Deployment Optimization for Vehicle -to-Infrastructure Cooperative Perception in Urban Occlusion Environments,

    C. Lin, Y. Wang, B. Gong, H. Liu, and H. Liu, “Roadside LiDAR Deployment Optimization for Vehicle -to-Infrastructure Cooperative Perception in Urban Occlusion Environments,” IEEE Trans. Instrum. Meas., vol. 74, pp. 1–14, 2025, doi: 10.1109/TIM.2025.3558231

    work page doi:10.1109/tim.2025.3558231 2025

  14. [14]

    In: IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022

    H. Yu et al. , “DAIR -V2X: A Large -Scale Dataset for Vehicle- Infrastructure Cooperative 3D Object Detection,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), New Orleans, LA, USA: IEEE, Jun. 2022, pp. 21329 –21338. doi: 10.1109/CVPR52688.2022.02067

    work page doi:10.1109/cvpr52688.2022.02067 2022

  15. [15]

    In: IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022

    X. Ye et al. , “Rope3D: The Roadside Perception Dataset for Autonomous Driving and Monocular 3D Object Detection Task,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), New Orleans, LA, USA: IEEE, Jun. 2022, pp. 21309–21318. doi: 10.1109/CVPR52688.2022.02065

    work page doi:10.1109/cvpr52688.2022.02065 2022

  16. [16]

    In: 2023 IEEE 26th International Con- ference on Intelligent Transportation Systems (ITSC)

    W. Zimmer, C. Creß, H. T. Nguyen, and A. C. Knoll, “TUMTraf Intersection Dataset: All You Need for Urban 3D Camera- LiDAR Roadside Perception,” in 2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC) , Bilbao, Spain: IEEE, Sep. 2023, pp. 1030–1037. doi: 10.1109/ITSC57777.2023.10422289

    work page doi:10.1109/itsc57777.2023.10422289 2023

  17. [17]

    Vision meets robotics: The KITTI dataset

    A. Geiger, P. Lenz, C. Stiller, and R. Urtasun, “Vision meets robotics: The KITTI dataset,” Int. J. Robot. Res., vol. 32, no. 11, pp. 1231–1237, Sep. 2013, doi: 10.1177/0278364913491297

    work page doi:10.1177/0278364913491297 2013

  18. [18]

    In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).p.9796–9805.IEEE(Jun2020).https://doi.org/10.1109/ cvpr42600.2020.00982

    H. Caesar et al. , “nuScenes: A Multimodal Dataset for Autonomous Driving,” in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA: IEEE, Jun. 2020, pp. 11618–11628. doi: 10.1109/CVPR42600.2020.01164

    work page doi:10.1109/cvpr42600.2020.01164 2020

  19. [19]

    In: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).p.9796–9805.IEEE(Jun2020).https://doi.org/10.1109/ cvpr42600.2020.00982

    P. Sun et al. , “Scalability in Perception for Autonomous Driving: Waymo Open Dataset,” in 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA: IEEE, Jun. 2020, pp. 2443–2451. doi: 10.1109/CVPR42600.2020.00252

    work page doi:10.1109/cvpr42600.2020.00252 2020

  20. [20]

    LiDAR Vehicle Trajectory Reconstruction with Arterial Shockwave Detection and Space –Time Analysis,

    A. Chen, P. J. Jin, and T. T. Zhang, “LiDAR Vehicle Trajectory Reconstruction with Arterial Shockwave Detection and Space –Time Analysis,” Transp. Res. Rec., vol. 2680, no. 1, pp. 497–518, 2026

    2026

  21. [21]

    V2I -Coop: Accurate Object Detection for Connected Automated Vehicles at Accident Black Spots With V2I Cross -Modality Cooperation,

    X. Zhou, C. Wang, Q. Xie, and T. Qiu, “V2I -Coop: Accurate Object Detection for Connected Automated Vehicles at Accident Black Spots With V2I Cross -Modality Cooperation,” IEEE Trans. Mob. Comput. , vol. 24, no. 3, pp. 2043 –2055, Mar. 2025, doi: 10.1109/TMC.2024.3486758

    work page doi:10.1109/tmc.2024.3486758 2043

  22. [22]

    Roadside lidar -based scene understanding toward intelligent traffic perception: A comprehensive review,

    J. Zhang et al. , “Roadside lidar -based scene understanding toward intelligent traffic perception: A comprehensive review,” ISPRS J. Photogramm. Remote Sens. , vol. 233, pp. 69 –88, 2026, doi: 10.1016/j.isprsjprs.2026.01.012

    work page doi:10.1016/j.isprsjprs.2026.01.012 2026

  23. [23]

    V2X-Sim: Multi-Agent Collaborative Perception Dataset and Benchmark for Autonomous Driving,

    Y. Li et al., “V2X-Sim: Multi-Agent Collaborative Perception Dataset and Benchmark for Autonomous Driving,” IEEE Robot. Autom. Lett. , vol. 7, no. 4, pp. 10914– 10921, Oct. 2022, doi: 10.1109/LRA.2022.3192802

    work page doi:10.1109/lra.2022.3192802 2022

  24. [24]

    Lidarsim: Realistic lidar simulation by leveraging the real world,

    S. Manivasagam et al. , “Lidarsim: Realistic lidar simulation by leveraging the real world,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2020, pp. 11167–11176

    2020

  25. [25]

    LIBRE: The Multiple 3D LiDAR Dataset,

    A. Carballo et al., “LIBRE: The Multiple 3D LiDAR Dataset,” Jun. 24, 2020, arXiv: arXiv:2003.06129. doi: 10.48550/arXiv.2003.06129

    work page doi:10.48550/arxiv.2003.06129 2020

  26. [26]

    In: IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2021, virtual, June 19-25, 2021

    T. Yin, X. Zhou, and P. Krahenbuhl, “Center -based 3D Object Detection and Tracking,” in 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , Nashville, TN, USA: IEEE, Jun. 2021, pp. 11779–11788. doi: 10.1109/CVPR46437.2021.01161

    work page doi:10.1109/cvpr46437.2021.01161 2021

  27. [27]

    A Multi -Scene Roadside Lidar Benchmark towards Digital Twins of Road Intersections,

    M. Tang et al. , “A Multi -Scene Roadside Lidar Benchmark towards Digital Twins of Road Intersections,” ISPRS Ann. Photogramm. Remote Sens. Spat. Inf. Sci., vol. X -4–2024, pp. 341 –348, Oct. 2024, doi: 10.5194/isprs-annals-X-4-2024-341-2024

    work page doi:10.5194/isprs-annals-x-4-2024-341-2024 2024

  28. [28]

    Drone Data Analytics for Measuring Traffic Metrics at Intersections in High - Density Areas,

    Q. Pu, Y. Zhu, J. Wang, H. Yang, K. Xie, and S. Cui, “Drone Data Analytics for Measuring Traffic Metrics at Intersections in High - Density Areas,” Transp. Res. Rec., p. 03611981241311566, 2025

    2025

  29. [29]

    Sustech points: A portable 3d point cloud interactive annotation platform system,

    E. Li, S. Wang, C. Li, D. Li, X. Wu, and Q. Hao, “Sustech points: A portable 3d point cloud interactive annotation platform system,” in 2020 IEEE Intelligent Vehicles Symposium (IV), IEEE, 2020, pp. 1108– 1115

    2020

  30. [30]

    Analytical Models for Assessing LiDAR Perception Performance for Vehicles with and without Occlusion,

    S. Cui, P. Cao, Y. Wang, D. Suo, and X. Liu, “Analytical Models for Assessing LiDAR Perception Performance for Vehicles with and without Occlusion,” IEEE Trans. Veh. Technol. , pp. 1 –16, 2025, doi: 10.1109/TVT.2025.3617472

    work page doi:10.1109/tvt.2025.3617472 2025

  31. [31]

    A Joint Optimization of Beam Distribution and Deployment for Roadside LiDAR Systems to Maximize Vehicle Perception,

    Y. He, P. Cao, D. Suo, and X. Liu, “A Joint Optimization of Beam Distribution and Deployment for Roadside LiDAR Systems to Maximize Vehicle Perception,” IEEE Trans. Intell. Veh., vol. 10, no. 2, pp. 1300–1314, Feb. 2025, doi: 10.1109/TIV.2024.3426524

    work page doi:10.1109/tiv.2024.3426524 2025

  32. [32]

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

    Y. Yan, Y. Mao, and B. Li, “SECOND: Sparsely Embedded Convolutional Detection,” Sensors, vol. 18, no. 10, Oct. 2018, doi: 10.3390/s18103337

    work page doi:10.3390/s18103337 2018

  33. [33]

    Integrated optimization for vehicle trajectory reconstruction under cooperative perception environment,

    T. Zhu, W. Li, and Y. Feng, “Integrated optimization for vehicle trajectory reconstruction under cooperative perception environment,” Transp. Res. Part C Emerg. Technol., vol. 184, p. 105522, Mar. 2026, doi: 10.1016/j.trc.2026.105522. Shunlai Cui received his B.S. degree from Inner Mongolia University, Hohhot, China, in 2022, and his M.S. degree from Sout...

    work page doi:10.1016/j.trc.2026.105522 2026

  34. [34]

    His research interests include traffic flow modeling and analysis in complex environments , low -altitude traffic situation awareness, and representation

    He is currently a Lecturer with the School of Automobile and Transportation, Xihua University. His research interests include traffic flow modeling and analysis in complex environments , low -altitude traffic situation awareness, and representation. Xiao Huo earned his B.S. degree in Transportation from Chongqing Jiaotong University, Chongqing, China, in ...

    2024

  35. [35]

    degree in Transportation Engineering at Southwest Jiaotong University, Chengdu, China

    He is currently pursuing his M.S. degree in Transportation Engineering at Southwest Jiaotong University, Chengdu, China. His research interests include artificial intelligence, object detection, and object tracking. Xiaobo Liu earned his Ph.D. degree from the New Jersey Institute of Technology, Newark, New Jersey, in 2004. He is currently a professor with...

    2004