Pith. sign in

REVIEW 4 major objections 5 minor 34 references

R-CARLA: High-Fidelity Sensor Simulations with Interchangeable Dynamics for Autonomous Racing

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read R-CARLA, a CARLA extension, claims to shrink the sim-to-real gap for autonomous racing by 42% for vehicle dynamics and by 82% for sensor simulation, unifying accurate dynamics with high-fidelity perception in one platform.

desk verdict Useful open-source racing simulator integration, but the headline 42%/82% Sim-to-Real reductions are not supported by the experiments as reported. read the letter →

arxiv 2506.09629 v1 pith:6YFEA265 submitted 2025-06-11 cs.RO

Maurice Brunner , Edoardo Ghignone , Nicolas Baumann , Michele Magno This is my paper
classification cs.RO
keywords autonomousracingsimulationsim-to-realgapdigitaltwinvehicledynamicsLiDARCARLAsensorfidelity
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

R-CARLA aims to give autonomous racing developers a single simulation environment that is accurate enough in both vehicle dynamics and sensors to test an entire autonomy stack end-to-end, from LiDAR input to control output. The paper claims that adding a custom dynamics interface, a dynamic single-track tire model, and a digital-twin pipeline to the CARLA simulator reduces the sim-to-real gap by 42% for car dynamics and by 82% for sensor simulation. If true, this would let teams develop race-ready perception, planning, and control software in one simulator rather than juggling separate dynamics and perception tools. The work is evaluated on two scaled racecars (F1TENTH and a gokart) across three real tracks.

What carries the argument

The load-bearing pieces are (1) a modular dynamics interface implemented in ROS that replaces CARLA's physics with any external state-update function; (2) a dynamic single-track vehicle model, a two-wheel bicycle model that includes tire slip through Pacejka tire-force curves, which lets the same code simulate a 1:10 F1TENTH car and a 1:2 gokart; and (3) a digital-twin creation pipeline that turns a LiDAR point cloud into an Unreal Engine map via outlier rejection, Poisson-disk sampling, and ball-pivoting meshing. Opponent simulation, which advances NPCs along waypoint trajectories, rounds out the closed-loop test capability.

What would settle it

Replay the gokart's real recorded LiDAR scans through the same SLAM localization used for the synthetic data in the digital-twin map; if the resulting RMSE is substantially larger than 0.164 m, the claimed closeness of synthetic and real sensors would not survive a direct real-data test. Alternatively, computing a point-cloud-to-point-cloud distance between synthetic and real scans at identical poses would settle it directly.

Watch

Extended reading notes

Core claim

The central claim is that R-CARLA, by disabling CARLA's default urban vehicle dynamics and injecting an external dynamics model through a ROS interface, makes CARLA a faithful full-stack simulator for autonomous racing. The proposed dynamic single-track model with Pacejka tire forces reproduces real lap times and lateral deviations more closely than CARLA's built-in dynamics, and the digital-twin maps built from real LiDAR point clouds support LiDAR localization with errors within 1.74 cm of those obtained with real data. This leads the authors to report a 42% average reduction in the sim-to-real gap for dynamics across all tested metrics, and an 82% reduction for the sensor-simulation or holistic-testing scenario (stated as 81% in the conclusion).

Load-bearing premise

The sensor-simulation improvement is measured with synthetic LiDAR scans localized against a map built from real LiDAR, which is a proxy for matching real sensor data, not a direct comparison of synthetic and real point clouds.

Editorial extensions

If this is right

  • Autonomous racing teams could develop and validate an entire perception-to-control stack in R-CARLA, since the simulator now claims to reproduce both vehicle handling and sensor behavior faithfully.
  • Scaled platforms like F1TENTH, which CARLA's stock dynamics cannot simulate at all, become testable in a high-fidelity visual environment for the first time.
  • A digital twin of a real track, built from one LiDAR pass, can serve as a deployment rehearsal environment for the same racestack software.
  • Holistic closed-loop testing (planner plus controller plus state estimation) shows smaller sim-to-real deviations than testing dynamics alone, suggesting that sensor fidelity and dynamics interact when validating full stacks.
  • The open-source release makes these capabilities reproducible and extensible by other racing and research groups.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The 82% sensor-fidelity claim is not yet a direct test of whether simulated LiDAR matches real LiDAR; it rests on the proxy that synthetic scans localize as well in a real-built map as real scans do, so a direct point-cloud-to-point-cloud distance comparison at matched poses would be a stronger falsifier.
  • If the digital-twin environments truly reproduce real tracks, reinforcement-learning and simulator-to-real transfer methods for racing could use them directly, potentially improving sample efficiency on real vehicles.
  • The modular dynamics interface makes R-CARLA a natural testbed for learned dynamics models, which could further shrink the remaining 58% dynamics gap by adapting to friction changes online.
  • The same pipeline could be applied beyond racing, e.g., to autonomous warehouse or search-and-rescue vehicles, wherever high-fidelity sensing and custom dynamics both matter.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper presents R-CARLA, a CARLA-based simulation framework for autonomous racing that combines custom vehicle dynamics, a digital-twin pipeline built from real LiDAR point clouds, and opponent simulation to enable full-stack perception-to-control testing. The central claim is that R-CARLA reduces the Sim-to-Real gap by 42% for vehicle dynamics and by 81-82% for sensor simulation. The evaluation uses two real racing platforms (a 1:10 F1TENTH car and a 1:2 gokart) and compares lap-time and lateral-error differences in simulation versus real-world laps, plus a SLAM localization accuracy experiment in a digital-twin map.

Significance. The framework addresses a real need: no existing open simulator for autonomous racing simultaneously provides high-fidelity sensor rendering and user-defined vehicle dynamics. The modular dynamics interface, open-source release, and the demonstration that CARLA's default dynamics cannot operate at 1:10 scale are valuable contributions. However, the headline quantitative claims are not adequately supported by the experiments as presented, and the paper's strongest numbers appear in the abstract rather than in a reproducible experimental result.

major comments (4)
  1. [Section IV-C, Table IV] The claimed 81-82% sensor-simulation Sim-to-Real reduction is not measured by the experiments described here. Both rows of Table IV use synthetic LiDAR data for localization: the 'Synthetic' row builds and localizes in the same synthetic map, while the 'Real' row localizes the same synthetic scans in a map built from real LiDAR. No real LiDAR data is ever passed through the SLAM localization pipeline, so the 1.74 cm difference is a measure of map-transfer consistency, not of sensor fidelity. This does not support the abstract's claim of an 82% reduction in the sensor-simulation gap.
  2. [Abstract, Introduction, Section IV, Conclusions] The headline numbers are internally inconsistent: the abstract reports 82% for sensor simulation, the introduction reports 81% 'by testing holistically', Section IV-A reports an 81% lap-time reduction on one gokart track, Section IV-B reports an average 81% lap-time reduction for F1TENTH, and the conclusion attributes 81% to digital-twin/sensor simulation. With the quantities defined differently in different places, the 82% figure cannot be traced to a single calculation and should be restated with a clear, consistent definition.
  3. [Section III-C1 and Section IV-A] The proposed dynamics model requires Pacejka tire parameters (B, C, D, E) per vehicle, but the paper does not report their values, the system-identification procedure, or the data used to fit them. Without this information, the 42% average reduction relative to CARLA dynamics is a comparison between a calibrated model and an uncalibrated one, and it is not possible to judge how much of the improvement is due to the model architecture. The authors should report the fitting procedure and an independent validation split.
  4. [Section IV-A, Table II] The 42% average reduction for car dynamics is computed only from the gokart row, because the CARLA baseline is marked n/a for the F1TENTH. The 'across various testing scenarios' phrasing therefore overstates the evidence: the dynamics comparison covers one vehicle platform, one track, and three metrics. This single-platform limitation should be stated explicitly, and per-metric results should be reported with lap-to-lap variability.
minor comments (5)
  1. [Section IV-C, Table IV] The sentence describing 'a discrepancy of only 1.74 cm' should be rephrased: 1.74 cm is the difference between two RMSE values, not the position accuracy of either system.
  2. [Section III-B] The parameters of the outlier rejection, Poisson-disk sampling, and ball-pivoting steps are not reported; adding them would improve reproducibility.
  3. [Section III-C1] Equation (2) is attributed to [33] (CommonRoad), but a reference to the original single-track model derivation would be more appropriate; also, the sign convention of the v_y terms in the first two rows should be checked.
  4. [Section IV-A] Only single-lap comparisons are reported, so it is unclear whether the reported differences are within normal lap-to-lap variability; adding repeated runs or variance estimates would strengthen the claims.
  5. [Figures 2 and 3] The subfigures in Figures 2 and 3 appear to overlap or duplicate content; the captions should describe each panel distinctly.

Circularity Check

2 steps flagged · score 4.0 of 10

The sensor-simulation Sim-to-Real reduction is partially self-referential: Section IV-C compares synthetic LiDAR against a real-built map, never against real LiDAR, and the 81/82% 'sensor' figure is a downstream lap-time reduction relabelled as a sensor gap.

  1. renaming known result [Section IV-C, Table IV]
    "This configuration represents a best-case scenario for the SLAM system, as the data used for mapping and localization are identical. The accuracy of this setup was evaluated by computing the RMSE ... Additionally, the synthetic data was used to perform localization in a map created from the original real-world data. The errors resulting from this process are labeled as real."

    Both rows of Table IV feed the identical synthetic CARLA LiDAR into the SLAM system; they differ only in whether the map is built from synthetic or real scans. No real sensor data is ever localized. The 'Synthetic' condition is explicitly defined by using identical data for mapping and localization, making it a self-consistency upper bound. Because the synthetic scans are rendered from a digital-twin mesh reconstructed from the same real-world point cloud that generated the 'real' map, the 1.74 cm closeness is a closed-loop reconstruction-consistency check. Presenting it as validating sensor simulation against real data reduces an external-fidelity claim to an internal self-comparison.

  2. renaming known result [Section IV-B, last paragraph; Conclusions]
    "the differences in lap time are significantly lower with a reduction of 69%, 92%, and 87% over all F1TENTH racetracks resulting in an average of 81%. This shows that simulating the entire pipeline holistically is meaningful ... Moreover, it highlights the realism of the simulated sensor data, as any significant divergence from real-life data would have resulted in an increased difference in the used metrics."

    The headline 'sensor simulation' reduction of 81%/82% is this lap-time reduction between two simulated scenarios ('Ours no SE' using ground-truth state and 'Ours' using full state estimation on synthetic sensors). It is a downstream closed-loop metric, not a sensor-to-real-sensor fidelity measurement. The conclusion relabels this downstream improvement as the sensor Sim-to-Real gap, so the central sensor claim is an inference from a proxy rather than an independent measurement.

full rationale

The dynamics contribution retains independent content: the gokart comparison against standard CARLA dynamics is a real closed-loop experiment, even though the Pacejka parameters are not reported. The full-stack/digital-twin engineering is also not circular in itself. However, the sensor-fidelity validation in Section IV-C is self-referential: both 'synthetic' and 'real' conditions use the same generated LiDAR, and the 'real' map is built from the same source point cloud used to construct the digital twin, so the 1.74 cm agreement is a consistency check, not an external validation against real sensor output. The 81%/82% abstract figure is likewise a renamed downstream lap-time improvement, not a direct sensor Sim-to-Real measurement. These issues do not make the whole framework circular, but they make the central sensor-simulation claim partially reduce to its own construction, hence the moderate score.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claims rest on a handful of fitted parameters (tire model, mesh processing) and on the unstated assumption that CARLA's sensors are realistic enough and that the simplified digital-twin mesh behaves like the real track. No new physical entities are introduced.

free parameters (4)
  • Pacejka tire force parameters (B, C, D, E) for each vehicle = Not reported
    Used in the dynamic single-track model of Section III-C1; required to compute tire forces F_f and F_r. The paper states the model requires Pacejka parameters for adaptation but does not report how they were identified or the fitted values.
  • Outlier rejection radius r and minimum neighbors m = Not reported
    Chosen thresholds in the sphere-based outlier rejection step of the digital-twin pipeline (Section III-B).
  • Poisson-disk sampling density / simplified point count = 25898 points from 1750112 after simplification
    The simplification step in Section III-B reduces the point cloud; the target density is chosen by hand.
  • Ball-pivoting algorithm radius = Not reported
    Chosen radius for mesh reconstruction in Section III-B; affects map geometry and therefore sensor simulation outcomes.
assumptions (5)
  • domain assumption The dynamic single-track model from CommonRoad [33] is an accurate representation of vehicle motion for racing platforms.
    Invoked in Section III-C1 to derive the state update equations; no validation is provided for the gokart or F1TENTH vehicles.
  • domain assumption The Pacejka tire model [34] captures tire forces adequately at the operating conditions of the two test vehicles.
    Used in Section III-C1 to compute F_f and F_r; the paper does not justify this choice for the specific tires used.
  • domain assumption CARLA's sensor simulation faithfully reproduces real LiDAR, IMU, and camera behavior.
    The entire sensor-simulation claim relies on CARLA's sensors being realistic; this is assumed throughout Section III-A and IV.
  • domain assumption Cartographer SLAM produces accurate maps and localization estimates in both real and synthetic environments.
    Cartographer is used to create maps from real data (Section III-B) and to evaluate state estimation accuracy (Section IV-C).
  • domain assumption A mesh generated by Poisson-disk sampling and ball-pivoting preserves the geometry needed for racing and LiDAR simulation.
    The digital-twin pipeline in Section III-B assumes that the simplified mesh is a faithful enough environment for the subsequent experiments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of R-CARLA: High-Fidelity Sensor Simulations with Interchangeable Dynamics for Autonomous Racing." pith.science (2026). https://pith.science/paper/6YFEA265

@misc{pith2026250609629,
  author       = {Pith},
  title        = {Pith review of: R-CARLA: High-Fidelity Sensor Simulations with Interchangeable Dynamics for Autonomous Racing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6YFEA265}},
  note         = {Machine review of arXiv:2506.09629}
}
read the original abstract

Autonomous racing has emerged as a crucial testbed for autonomous driving algorithms, necessitating a simulation environment for both vehicle dynamics and sensor behavior. Striking the right balance between vehicle dynamics and sensor accuracy is crucial for pushing vehicles to their performance limits. However, autonomous racing developers often face a trade-off between accurate vehicle dynamics and high-fidelity sensor simulations. This paper introduces R-CARLA, an enhancement of the CARLA simulator that supports holistic full-stack testing, from perception to control, using a single system. By seamlessly integrating accurate vehicle dynamics with sensor simulations, opponents simulation as NPCs, and a pipeline for creating digital twins from real-world robotic data, R-CARLA empowers researchers to push the boundaries of autonomous racing development. Furthermore, it is developed using CARLA's rich suite of sensor simulations. Our results indicate that incorporating the proposed digital-twin framework into R-CARLA enables more realistic full-stack testing, demonstrating a significant reduction in the Sim-to-Real gap of car dynamics simulation by 42% and by 82% in the case of sensor simulation across various testing scenarios.

Figures

Figures reproduced from arXiv: 2506.09629 by the authors.

Figure 1
Figure 1. Overview of the 5 main modules of Racing CARLA (R-CARLA). On the left, the digital twin creation module is shown which creates new environments from real sensor data. The vehicle drive inputs u from an autonomy stack are passed through an interface to the dynamics simulator, which computes the next state of the racecar. Together with the pose of the opponents coming from the opponent simulator, this is passed to CAR… view at source ↗
Figure 2
Figure 2. Overview of different sensors and the rendering of a [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. A comparison of the data of different sensors avail [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Comparison of different steps in the digital-twin creation process. From left to right: The original point cloud, [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Top: Different parts of the real environment. Bottom: The resulting digital twin recreated from real-world sensor [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

34 extracted references · 28 canonical work pages

  1. [1]

    Indy autonomous challenge-autonomous race cars at the handling limits,

    A. Wischnewski, M. Geisslinger, J. Betz, T. Betz, F. Fent, A. Heilmeier, L. Hermansdorfer, T. Herrmann, S. Huch, P. Karleet al., “Indy autonomous challenge-autonomous race cars at the handling limits,” in12th International Munich Chassis Symposium 2021: chassis. tech plus. Springer, 2022, pp. 163–182

    2021

  2. [2]

    Emergent ecosystem for radical innovation: Entrepreneurial probing at formula e,

    S. Jarvenpaa and W. Standaert, “Emergent ecosystem for radical innovation: Entrepreneurial probing at formula e,” in Proceedings of the 50th HICSS, 2017

    work page 2017

  3. [3]

    From accelerated advertising to fanboost: mediatized motorsport,

    M. Finn, “From accelerated advertising to fanboost: mediatized motorsport,”Sport in Society, vol. 24, no. 6, pp. 937–953, 2021. [Online]. Available: https://doi.org/10.1080/17430437.2019.1710131

    work page arXiv 2021

  4. [4]

    Autonomous vehicles on the edge: A survey on autonomous vehicle racing

    J. Betz, H. Zheng, A. Liniger, U. Rosolia, P. Karle, M. Behl, V . Krovi, and R. Mangharam, “Autonomous vehicles on the edge: A survey on autonomous vehicle racing.” [Online]. Available: https://arxiv.org/abs/2202.07008

    arXiv

  5. [5]

    Tum autonomous motorsport: An autonomous racing software for the indy autonomous challenge,

    J. Betz, T. Betz, F. Fent, M. Geisslinger, A. Heilmeier, L. Hermansdorfer, T. Herrmann, S. Huch, P. Karle, M. Lienkamp, B. Lohmann, F. Nobis, L. ¨Ogretmen, M. Rowold, F. Sauerbeck, T. Stahl, R. Trauth, F. Werner, and A. Wischnewski, “Tum autonomous motorsport: An autonomous racing software for the indy autonomous challenge,”Journal of Field Robotics, vol....

    work page doi:10.1002/rob.22153 2023

  6. [6]

    er.autopilot 1.1: A software stack for autonomous racing on oval and road course tracks,

    A. Raji, D. Caporale, F. Gatti, A. Toschi, N. Musiu, M. Veruc- chi, F. Prignoli, D. Malatesta, A. F. Jesus, A. Finazzi, F. Amerotti, F. Bagni, E. Mascaro, P. Musso, and M. Bertogna, “er.autopilot 1.1: A software stack for autonomous racing on oval and road course tracks,”IEEE Transactions on Field Robotics, vol. 1, pp. 332–359, 2024

    work page 2024

  7. [7]

    An autonomous racing system: Design, implementation, and analysis; team KAIST at the IAC,

    C. Jung, A. Finazzi, H. Seong, D. Lee, S. Lee, B. Kim, G. Kang, and H. Shim, “An autonomous racing system: Design, implementation, and analysis; team KAIST at the IAC,”Field Robotics, vol. 3, no. 1, pp. 766–800, Jan. 2023. [Online]. Available: https://doi.org/10.55417/fr.2023024

    work page doi:10.55417/fr.2023024 2023

  8. [8]

    AMZ driverless: The full autonomous racing system,

    J. Kabzan, M. I. Valls, V . J. F. Reijgwart, H. F. C. Hendrikx, C. Ehmke, M. Prajapat, A. B ¨uhler, N. Gosala, M. Gupta, R. Sivanesan, A. Dhall, E. Chisari, N. Karnchanachari, S. Brits, M. Dangel, I. Sa, R. Dub ´e, A. Gawel, M. Pfeiffer, A. Liniger, J. Lygeros, and R. Siegwart, “AMZ driverless: The full autonomous racing system,”J. field robot., no. rob.2...

    work page 2020

Show all 34 references
  1. [9]

    Forzaeth race stack—scaled autonomous head- to-head racing on fully commercial off-the-shelf hardware,

    N. Baumann, E. Ghignone, J. K ¨uhne, N. Bastuck, J. Becker, N. Imholz, T. Kr¨anzlin, T. Y . Lim, M. L¨otscher, L. Schwarzen- bachet al., “Forzaeth race stack—scaled autonomous head- to-head racing on fully commercial off-the-shelf hardware,” Journal of Field Robotics, 2024

    2024

  2. [10]

    F1tenth: An open-source evaluation environment for contin- uous control and reinforcement learning,

    M. O’Kelly, H. Zheng, D. Karthik, and R. Mangharam, “F1tenth: An open-source evaluation environment for contin- uous control and reinforcement learning,” inNeurIPS 2019 Competition and Demonstration Track. PMLR, 2020, pp. 77–89

    2019

  3. [11]

    How simulation helps autonomous driving: A survey of sim2real, digital twins, and parallel intelligence,

    X. Hu, S. Li, T. Huang, B. Tang, R. Huai, and L. Chen, “How simulation helps autonomous driving: A survey of sim2real, digital twins, and parallel intelligence,”IEEE Transactions on Intelligent Vehicles, vol. 9, no. 1, pp. 593–612, 2024

    2024

  4. [12]

    Siegwart and I

    R. Siegwart and I. R. Nourbakhsh,Introduction to autonomous mobile robots. MIT Press, 2004, vol. 1

    2004

  5. [13]

    What lies in the shadows? safe and computation-aware motion planning for autonomous vehicles using intent-aware dynamic shadow re- gions,

    Y . Nager, A. Censi, and E. Frazzoli, “What lies in the shadows? safe and computation-aware motion planning for autonomous vehicles using intent-aware dynamic shadow re- gions,” in2019 International Conference on Robotics and Automation (ICRA), 2019, pp. 5800–5806

    2019

  6. [14]

    CARLA: An open urban driving simulator,

    A. Dosovitskiy, G. Ros, F. Codevilla, A. Lopez, and V . Koltun, “CARLA: An open urban driving simulator,” inProceedings of the 1st Annual Conference on Robot Learning, 2017, pp. 1–16

    2017

  7. [15]

    End-to-end autonomous driving: Challenges and frontiers,

    L. Chen, P. Wu, K. Chitta, B. Jaeger, A. Geiger, and H. Li, “End-to-end autonomous driving: Challenges and frontiers,” IEEE Transactions on Pattern Analysis and Machine Intelli- gence, vol. 46, no. 12, pp. 10 164–10 183, 2024

    2024

  8. [16]

    Anti-carla: An adversarial testing framework for autonomous vehicles in carla,

    S. Ramakrishna, B. Luo, C. B. Kuhn, G. Karsai, and A. Dubey, “Anti-carla: An adversarial testing framework for autonomous vehicles in carla,” in2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC), 2022, pp. 2620– 2627

    2022

  9. [17]

    Carla-gear: A dataset generator for a systematic eval- uation of adversarial robustness of deep learning vision mod- els,

    F. Nesti, G. Rossolini, G. D’Amico, A. Biondi, and G. But- tazzo, “Carla-gear: A dataset generator for a systematic eval- uation of adversarial robustness of deep learning vision mod- els,”IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 8, pp. 9840–9851, 2024

    2024

  10. [18]

    Ad perdevkit: An autonomous driving perception development kit using carla simulator and ros,

    J. de la Pe ˜na, L. M. Bergasa, M. Antunes, F. Arango, C. G´omez-Hu´elamo, and E. L ´opez-Guill´en, “Ad perdevkit: An autonomous driving perception development kit using carla simulator and ros,” in2022 IEEE 25th International Confer- ence on Intelligent Transportation Systems...

    2022

  11. [19]

    Multimodal end-to-end autonomous driving,

    Y . Xiao, F. Codevilla, A. Gurram, O. Urfalioglu, and A. M. L´opez, “Multimodal end-to-end autonomous driving,”IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 1, pp. 537–547, 2022

    2022

  12. [20]

    Carla-autoware-bridge: Facilitating autonomous driv- ing research with a unified framework for simulation and module development,

    G. Kaljavesi, T. Kerbl, T. Betz, K. Mitkovskii, and F. Dier- meyer, “Carla-autoware-bridge: Facilitating autonomous driv- ing research with a unified framework for simulation and module development,” in2024 IEEE Intelligent Vehicles Sym- posium (IV), 2024, pp. 224–229

    2024

  13. [21]

    A simulation benchmark for autonomous racing with large-scale human data,

    A. Remonda, N. Hansen, A. Raji, N. Musiu, M. Bertogna, E. E. Veas, and X. Wang, “A simulation benchmark for autonomous racing with large-scale human data,” 2024

    2024

  14. [22]

    A super-human vision-based rein- forcement learning agent for autonomous racing in Gran Turismo,

    M. Vasco, T. Seno, K. Kawamoto, K. Subramanian, P. R. Wurman, and P. Stone, “A super-human vision-based rein- forcement learning agent for autonomous racing in Gran Turismo,”Reinforcement Learning Journal, vol. 4, pp. 1674– 1710, 2024

    2024

  15. [23]

    Super-human performance in gran turismo sport using deep reinforcement learning,

    F. Fuchs, Y . Song, E. Kaufmann, D. Scaramuzza, and P. D ¨urr, “Super-human performance in gran turismo sport using deep reinforcement learning,”IEEE Robotics and Automation Let- ters, vol. 6, no. 3, pp. 4257–4264, 2021

    2021

  16. [24]

    Deepracing: A framework for au- tonomous racing,

    T. Weiss and M. Behl, “Deepracing: A framework for au- tonomous racing,” in2020 Design, Automation and Test in Europe Conference and Exhibition (DATE), 2020, pp. 1163– 1168

    2020

  17. [25]

    Autodrive simulator: A simulator for scaled autonomous vehicle research and education,

    T. V . Samak, C. V . Samak, and M. Xie, “Autodrive simulator: A simulator for scaled autonomous vehicle research and education,” in2021 2nd International Conference on Control, Robotics and Intelligent System, ser. CCRIS’21. New York, NY , USA: Association for Computing Machin...

    2021 doi

  18. [26]

    f1tenth. dev-an open-source ros based f1/10 autonomous racing simulator,

    V . S. Babu and M. Behl, “f1tenth. dev-an open-source ros based f1/10 autonomous racing simulator,” in2020 IEEE 16th International Conference on Automation Science and Engineering (CASE). IEEE, 2020, pp. 1614–1620

    2020

  19. [27]

    Learning dynamics models for velocity estimation in au- tonomous racing,

    J. Wegrzynowski, G. Czechmanowski, P. Kicki, and K. Walas, “Learning dynamics models for velocity estimation in au- tonomous racing,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024, pp. 972–979

    2024

  20. [28]

    First, learn what you don’t know: Active information gathering for driving at the limits of handling,

    A. Davydov, F. Djeumou, M. Greiff, M. Suminaka, M. Thompson, J. Subosits, and T. Lew, “First, learn what you don’t know: Active information gathering for driving at the limits of handling,” 2024. [Online]. Available: https://arxiv.org/abs/2411.00107

    2024 arXiv

  21. [29]

    Anycar to anywhere: Learning universal dynamics model for agile and adaptive mobility,

    W. Xiao, H. Xue, T. Tao, D. Kalaria, J. M. Dolan, and G. Shi, “Anycar to anywhere: Learning universal dynamics model for agile and adaptive mobility,” 2024. [Online]. Available: https://arxiv.org/abs/2409.15783

    2024 arXiv

  22. [30]

    Real-time loop closure in 2d lidar slam,

    W. Hess, D. Kohler, H. Rapp, and D. Andor, “Real-time loop closure in 2d lidar slam,” inProceedings - IEEE International Conference on Robotics and Automation, vol. 2016-June, 2016

    2016

  23. [31]

    Dithering with blue noise,

    R. A. Ulichney, “Dithering with blue noise,”Proceedings of the IEEE, vol. 76, no. 1, pp. 56–79, 1988

    1988

  24. [32]

    The ball-pivoting algorithm for surface recon- struction,

    F. Bernardini, J. Mittleman, H. Rushmeier, C. Silva, and G. Taubin, “The ball-pivoting algorithm for surface recon- struction,”IEEE Transactions on Visualization and Computer Graphics, vol. 5, no. 4, pp. 349–359, 1999

    1999

  25. [33]

    Commonroad: Composable benchmarks for motion planning on roads,

    M. Althoff, M. Koschi, and S. Manzinger, “Commonroad: Composable benchmarks for motion planning on roads,” in 2017 IEEE Intelligent Vehicles Symposium (IV), 2017, pp. 719–726

    2017

  26. [34]

    Shear force development by pneumatic tyres in steady state conditions: a review of modelling aspects,

    H. B. Pacejka and R. S. Sharp, “Shear force development by pneumatic tyres in steady state conditions: a review of modelling aspects,”Vehicle system dynamics, vol. 20, no. 3-4, pp. 121–175, 1991

    1991

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.