Pith. sign in

REVIEW 3 major objections 5 minor 45 references

BETTY Dataset: A Multi-modal Dataset for Full-Stack Autonomy

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The BETTY dataset records every input and output of an autonomous racing vehicle—perception, state estimation, dynamics, actuation, and metadata—so researchers can train and test the full autonomy stack, not just one module.

desk verdict BETTY is a genuine, publicly hosted full-stack autonomous racing dataset that fills a real gap; the auto-generated perception labels need quantitative validation before the ground-truth claim is accepted. read the letter →

arxiv 2505.07266 v1 pith:OE6YBD65 submitted 2025-05-12 cs.RO

Micah Nye , Ayoub Raji , Andrew Saba , Eidan Erlich , Robert Exley , Aragya Goyal , Alexander Matros , Ritesh Misra
show 4 more authors
Matthew Sivaprakasam Marko Bertogna Deva Ramanan Sebastian Scherer
This is my paper · ORCID
classification cs.RO
keywords autonomousracingmulti-modaldatasetfull-stackautonomyvehicledynamicsstateestimationmotionforecastingsensorfusionself-supervisedlearning
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

The paper introduces BETTY, a large multi-modal dataset collected over four years from autonomous racing vehicles, spanning 13 hours and 32 TB across six tracks. It aims to fill a gap: existing autonomy datasets focus on perception, planning, or dynamics separately, whereas BETTY records all sensor inputs plus the outputs of the software stack, along with semantic metadata and ground-truth annotations. If the dataset delivers on its claims, researchers without access to a racecar can train and evaluate complete autonomy pipelines—perception, state estimation, dynamics modeling, motion forecasting, and control—under high-speed, high-acceleration conditions that previous datasets do not cover. The paper also claims that BETTY is the most diverse autonomous racing dataset to date in environment count, acceleration envelope, and sensor modalities.

What carries the argument

The object that carries the argument is the dataset itself, structured as synchronized bag files plus plain image, point-cloud, and CSV files. The load-bearing mechanism is the joint capture of all sensor streams and all autonomy-stack outputs on the same clock, with GPS/PTP synchronization (average offset 2.8 ms) and per-sensor calibration, so that perception, state estimation, dynamics, and control can be evaluated both independently and chained end-to-end. A second mechanism is the annotation pipeline: 2D boxes from a text-prompted open-set detector, 3D boxes from a point-cloud detector refined by template alignment initialized near the opponent's RTK-GNSS position, with manual verification of every accepted label.

What would settle it

Take a random sample of frames, have independent annotators re-label opponent vehicles from scratch, and compare with the provided labels. If agreement is low—especially on the 3D LiDAR boxes—or if the reported GNSS-to-box-center acceptance threshold is frequently violated, then the ground-truth labels are too noisy to support supervised training, and the paper's central utility claim collapses.

Watch

Extended reading notes

Core claim

The central claim is that full-stack autonomy for high-speed racing can be studied from a single open dataset. BETTY pairs exteroceptive sensors (cameras, LiDAR, radar) with proprioceptive and vehicle-state sensors (dual GNSS/IMUs, tire temperature and pressure, suspension, wheel speed and torque, gear, and commanded versus actual throttle, brake, and steering) and with autonomy outputs such as planned trajectories and opponent states. The authors argue this combination is unique: earlier racing datasets offer either perception only or dynamics and control only, while general autonomous-driving datasets lack the extreme dynamics, tire-state data, and raw GNSS observables at racing speeds. They further claim that the annotations, calibrated poses, HD maps, and temporal metadata make BETTY usable for supervised and self-supervised learning, and for module-level or end-to-end evaluation.

Load-bearing premise

The dataset's value as ground truth rests on the assumption that the automatically generated 2D and 3D labels, filtered by confidence thresholds and manually checked, are accurate enough to train supervised perception and state-estimation models; the paper does not report quantitative accuracy for these labels.

Editorial extensions

If this is right

  • A researcher can train perception, state estimation, and dynamics modules on the same vehicle and then measure how errors propagate across the stack, which no previous racing dataset supports.
  • Control and dynamics work can be tested against extreme events in the data, including 63 m/s crashes, loss of tire traction, and operation at the stability limit.
  • Raw GNSS observables at speeds above 30 m/s, paired with IMUs and LiDAR, enable reliability studies for localization under jamming or signal dropout.
  • Semantic metadata such as track conditions, race flags, run intent, and single- versus multi-agent status can act as supervisory signals for self-supervised and behavioral learning.
  • The modular release format lets users download only the modalities they need while keeping all streams and labels compatible, so the dataset can serve as a common benchmark across vision, SLAM, and controls communities.

Reading between the lines

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

  • The same corpus could serve as a sim-to-real calibration set: the high-rate vehicle-state and tire data could validate or tune racing simulators used for reinforcement learning.
  • Because tire temperature, pressure, and slip angle are recorded alongside extreme maneuvers, the dataset may support data-driven tire models that go beyond empirical curve fits, with potential spillover to road-car safety.
  • The temporal metadata—attacker or defender role, race flags, run intent, and success—could be turned into behavioral cloning or language-conditioned driving supervision, an extension the paper only gestures at.
  • With raw GNSS observables and map data, the road-course portions could become a challenging GNSS-denied localization benchmark under bridges and occlusions.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper introduces the BETTY dataset, a multi-modal autonomous-racing dataset collected over four years on multiple Dallara AV-21 vehicles from the Indy Autonomous Challenge. It reports over 13 hours and 32 TB of data spanning six environments, including ovals, a road course, and a hillclimb, with cameras, LiDAR, radar, IMU/GNSS with RTK corrections, vehicle-dynamics sensors, tire-state sensors, and raw autonomy-stack outputs (throttle, brake, steering, gear, planned trajectories). The dataset is publicly released in ROS2/mcap and ROS-agnostic formats, and it includes metadata, HD maps, calibration information, and auto-generated plus manually verified 2D and 3D opponent labels. The authors argue that BETTY uniquely covers the full autonomy loop, supporting perception, state estimation, dynamics modeling, motion forecasting, and control in highly dynamic regimes that existing datasets do not capture.

Significance. If the dataset is as described, it would be a valuable open resource. It is one of very few public datasets combining high-speed racing dynamics with rich perception data, and it appears to be the only one offering raw GNSS observables, tire-state channels, and non-ego opponent odometry together. The paper includes detailed platform and synchronization descriptions, a public data URL, and comparison tables against KITTI, NuScenes, Argoverse 2, Waymo, TartanDrive 2.0, and RACECAR. The most impressive strength is the breadth of cross-modal data, which could enable researchers to evaluate chained autonomy modules (perception to estimation to control) that prior datasets cannot support. The main risk is the unvalidated quality of the auto-generated perception labels, which are a load-bearing component for the supervised-perception use case.

major comments (3)
  1. [III-D2] The paper's central claim of providing 'ground-truth annotated labels' for supervised perception is not quantitatively supported. Section III-D2 states that 2D labels come from Grounding DINO and 3D labels from PointPillars with ICP template refinement, with acceptance based on an unspecified threshold between the GNSS position and box center, followed by manual verification. However, no precision, recall, IoU, center-error, or inter-annotator agreement statistics are reported. Manual verification is a process step, not a quality measurement. Because RTK GNSS provides centimeter-level opponent positions, the authors can directly report the distribution of box-to-GNSS deviations and the acceptance threshold; they should also provide per-environment and per-speed breakdowns, especially at the claimed 63 m/s operation. Without such numbers, the claim that BETTY supports supervised perception training is unsubstantiated.
  2. [III-F and III-G] Calibration and synchronization accuracy are load-bearing for a multi-modal full-stack dataset, yet the paper only reports that calibration was performed by a commercial tool and that the average synchronization offset was 2.8 ms. Section III-F gives no reprojection error, no extrinsic or intrinsic accuracy metric, and no evaluation of the LiDAR-map alignment used for the HD maps. Section III-G reports only an average offset, not its standard deviation, maximum, or per-sensor distribution. The authors should report quantitative calibration residuals and synchronization statistics across the dataset, and state whether any runs were excluded due to calibration or timing failures.
  3. [Contribution 3 and Section IV-C] The contribution statement that BETTY is 'the most diverse autonomous racing dataset with the largest environment diversity, highest accelerations, and most sensor modalities, by far' is not backed by the quantitative analysis in Section IV-C. Figure 5 shows a qualitative G-G plot comparing only the top 10 runs against RACECAR, with no numerical table of maximum longitudinal/lateral acceleration, no per-environment statistics, and no comparison against other racing datasets such as the simulation dataset of [12]. The 'by far' qualifier requires concrete metrics; otherwise this contribution claim should be softened to a descriptive comparison.
minor comments (5)
  1. [II-4] The claim that BETTY is 'the only dataset providing raw GNSS observables at speeds greater than 30 m/s' cites a Kaggle competition page, which does not support the exclusivity claim. This assertion needs a systematic survey citation or should be reworded as 'to the best of our knowledge' with a more appropriate reference.
  2. [Abstract and III-B] The abstract mentions '63 m/s crashes' and 'loss of tire traction,' but the data-collection section does not specify which runs contain these events or whether they are distributed across environments. A table or metadata note identifying the extreme-dynamics runs would help users select relevant data.
  3. [IV-C, Fig. 5] The G-G plot would be more informative with a clear legend distinguishing BETTY from RACECAR points, since the caption refers to two datasets but the plot style is not described.
  4. [III-D2] The term 'ground-truth annotated labels' is misleading for labels that are auto-generated with a confidence threshold and manual verification; consider using 'auto-labeled with manual verification' or 'pseudo-labels' in the dataset documentation, reserving 'ground truth' for the GNSS-derived states.
  5. [III-E] The paper states that all camera and LiDAR annotations are provided in KITTI format, but KITTI format has specific coordinate conventions and label semantics; the authors should document any extensions or deviations, such as velocity/heading fields for 3D boxes, on the dataset website.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: BETTY is an empirical dataset collection with no fitted predictions or derivation chain that reduces to its own inputs.

full rationale

The paper makes no predictive claim derived from fitted parameters, and no quantity is defined in terms of a target outcome it then claims to explain. Its central contribution is the release of raw multi-modal sensor data, autonomy-stack outputs, metadata, and annotations collected from autonomous racing vehicles. The auto-labeling pipeline (Grounding DINO for 2D, PointPillars plus ICP template refinement for 3D, with manual verification) is a data-processing procedure, not a derivation whose result is guaranteed by its inputs; even though the 3D label acceptance check uses GNSS position as a reference, the paper explicitly presents GNSS as the ground-truth source rather than as a predicted quantity. The comparative claim in Contribution 3 rests on the dataset contents described in Tables I and II, and on the G-G plot in Fig. 5, which are empirical characterizations rather than consequences of an assumed model. The absence of quantitative label-accuracy statistics is a legitimate correctness and usability concern, but it is not circularity. Citations to the authors' prior racing-software papers are contextual and illustrative, not load-bearing proofs of the dataset's properties. The dataset is self-contained as an empirical resource, so the appropriate circularity score is 0.

Assumptions & free parameters 0 free parameters · 2 assumptions · 0 invented entities

No free parameters are introduced. The dataset relies on two domain assumptions: that the auto-generated labels are accurate enough to be used as ground truth, and that the sensor calibration and timing are consistent across three vehicles and teams. Both are stated in the paper but not quantitatively validated.

assumptions (2)
  • domain assumption Auto-generated 2D and 3D labels are accurate enough to serve as ground truth after manual verification.
    The paper relies on these labels for supervised tasks but provides no quantitative accuracy metrics (Section III-D2).
  • domain assumption The calibration and synchronization of all sensors are correct and consistent across the three AV-21 vehicles.
    The dataset combines data from multiple teams; calibration was outsourced to MSA and time sync had an average 2.8 ms offset (Sections III-F and III-G). No cross-vehicle consistency validation is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of BETTY Dataset: A Multi-modal Dataset for Full-Stack Autonomy." pith.science (2026). https://pith.science/paper/OE6YBD65

@misc{pith2026250507266,
  author       = {Pith},
  title        = {Pith review of: BETTY Dataset: A Multi-modal Dataset for Full-Stack Autonomy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OE6YBD65}},
  note         = {Machine review of arXiv:2505.07266}
}
read the original abstract

We present the BETTY dataset, a large-scale, multi-modal dataset collected on several autonomous racing vehicles, targeting supervised and self-supervised state estimation, dynamics modeling, motion forecasting, perception, and more. Existing large-scale datasets, especially autonomous vehicle datasets, focus primarily on supervised perception, planning, and motion forecasting tasks. Our work enables multi-modal, data-driven methods by including all sensor inputs and the outputs from the software stack, along with semantic metadata and ground truth information. The dataset encompasses 4 years of data, currently comprising over 13 hours and 32TB, collected on autonomous racing vehicle platforms. This data spans 6 diverse racing environments, including high-speed oval courses, for single and multi-agent algorithm evaluation in feature-sparse scenarios, as well as high-speed road courses with high longitudinal and lateral accelerations and tight, GPS-denied environments. It captures highly dynamic states, such as 63 m/s crashes, loss of tire traction, and operation at the limit of stability. By offering a large breadth of cross-modal and dynamic data, the BETTY dataset enables the training and testing of full autonomy stack pipelines, pushing the performance of all algorithms to the limits. The current dataset is available at https://pitt-mit-iac.github.io/betty-dataset/.

Figures

Figures reproduced from arXiv: 2505.07266 by the authors.

Figure 1
Figure 1. We provide exteroceptive sensors (camera, LiDAR, and radar), proprioceptive sensors (tire temperature sensor, slip [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. AV-21 sensors and their respective locations. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The BETTY dataset was collected in six environments. Each track presents unique characteristics, which present [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Auto-labelling pipeline. Grounding DINO is used as [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: G-G plot of the top 10 most dynamic runs from [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 39 canonical work pages

  1. [12]

    A simulation bench- mark for autonomous racing with large-scale human data

    Adrian Remonda, Nicklas Hansen, Ayoub Raji, Nicola Musiu, Marko Bertogna, Eduardo Veas, and Xiaolong Wang. A simulation bench- mark for autonomous racing with large-scale human data. In A. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet, J. Tomczak, and C. Zhang, editors,Advances in Neural Information Processing Systems, volume 37, pages 102078–10210...

    work page 2024

  2. [1]

    Vision meets robotics: The kitti dataset.International Journal of Robotics Research (IJRR), 2013

    Andreas Geiger, Philip Lenz, Christoph Stiller, and Raquel Urtasun. Vision meets robotics: The kitti dataset.International Journal of Robotics Research (IJRR), 2013

    2013

  3. [2]

    Nuplan: A closed-loop ml-based planning benchmark for autonomous vehicles, 2022

    Holger Caesar, Juraj Kabzan, Kok Seang Tan, Whye Kit Fong, Eric Wolff, Alex Lang, Luke Fletcher, Oscar Beijbom, and Sammy Omari. Nuplan: A closed-loop ml-based planning benchmark for autonomous vehicles, 2022

    work page 2022

  4. [3]

    Common- road: Composable benchmarks for motion planning on roads

    Matthias Althoff, Markus Koschi, and Stefanie Manzinger. Common- road: Composable benchmarks for motion planning on roads. In2017 IEEE Intelligent V ehicles Symposium (IV), pages 719–726, 2017

    work page 2017

  5. [4]

    Tartanair: A dataset to push the limits of visual slam.2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020

    Wenshan Wang, Delong Zhu, Xiangwei Wang, Yaoyu Hu, Yuheng Qiu, Chen Wang, Yafei Hu, Ashish Kapoor, and Sebastian Scherer. Tartanair: A dataset to push the limits of visual slam.2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020

    work page 2020

  6. [5]

    Tartandrive 2.0: More modalities and better infrastructure to further self-supervised learning research in off- road driving tasks, 2024

    Matthew Sivaprakasam, Parv Maheshwari, Mateo Guaman Castro, Samuel Triest, Micah Nye, Steve Willits, Andrew Saba, Wenshan Wang, and Sebastian Scherer. Tartandrive 2.0: More modalities and better infrastructure to further self-supervised learning research in off- road driving tasks, 2024

    work page 2024

  7. [6]

    Harbott John C

    Lene K. Harbott John C. Kegelman and J. Christian Gerdes. Insights into vehicle trajectories at the handling limits: analysing open data from race car drivers.V ehicle System Dynamics, 55(2):191–207, 2017

    work page 2017

  8. [7]

    Test results at ford lommel proving ground esr 11, June 30 2019

    Cyrano Vaseur and Sebastiaan Van Aalst. Test results at ford lommel proving ground esr 11, June 30 2019

    work page 2019

Show all 45 references
  1. [8]

    Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom

    Holger Caesar, Varun Bankiti, Alex H. Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom. nuscenes: A multimodal dataset for autonomous driving.arXiv preprint arXiv:1903.11027, 2019

    1903 arXiv

  2. [9]

    Argoverse 2: Next generation datasets for self-driving perception and forecasting

    Benjamin Wilson, William Qi, Tanmay Agarwal, John Lambert, Jagjeet Singh, Siddhesh Khandelwal, Bowen Pan, Ratnesh Kumar, Andrew Hartnett, Jhony Kaesemodel Pontes, Deva Ramanan, Peter Carr, and James Hays. Argoverse 2: Next generation datasets for self-driving perception and fo...

    2021

  3. [10]

    Scalability in perception for autonomous driving: Waymo open dataset, 2020

    Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, Aurelien Chouard, Vijaysai Patnaik, Paul Tsui, James Guo, Yin Zhou, Yuning Chai, Ben- jamin Caine, Vijay Vasudevan, Wei Han, Jiquan Ngiam, Hang Zhao, Aleksei Timofeev, Scott Ettinger, Maxim Krivokon, Amy Gao, Aditya Joshi, Sheng Z...

    2020

  4. [11]

    Racecar - the dataset for high-speed autonomous racing

    Amar Kulkarni, John Chrosniak, Emory Ducote, Florian Sauerbeck, Andrew Saba, Utkarsh Chirimar, John Link, Madhur Behl, and Mar- cello Cellina. Racecar - the dataset for high-speed autonomous racing. In2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IR...

    2023

  5. [13]

    Guess the drift with lop-ukf: Lidar odometry and pacejka model for real- time racecar sideslip estimation

    Alessandro Toschi, Nicola Musiu, Francesco Gatti, Ayoub Raji, Francesco Amerotti, Micaela Verucchi, and Marko Bertogna. Guess the drift with lop-ukf: Lidar odometry and pacejka model for real- time racecar sideslip estimation. In2024 IEEE Intelligent V ehicles Symposium (IV), ...

    2024

  6. [14]

    Argoverse: 3d tracking and forecast- ing with rich maps

    Ming-Fang Chang, John W Lambert, Patsorn Sangkloy, Jagjeet Singh, Slawomir Bak, Andrew Hartnett, De Wang, Peter Carr, Simon Lucey, Deva Ramanan, and James Hays. Argoverse: 3d tracking and forecast- ing with rich maps. InConference on Computer Vision and Pattern Recognition (CV...

    2019

  7. [15]

    Panoptic nuscenes: A large-scale benchmark for lidar panoptic segmentation and tracking.arXiv preprint arXiv:2109.03805, 2021

    Whye Kit Fong, Rohit Mohan, Juana Valeria Hurtado, Lubing Zhou, Holger Caesar, Oscar Beijbom, and Abhinav Valada. Panoptic nuscenes: A large-scale benchmark for lidar panoptic segmentation and tracking.arXiv preprint arXiv:2109.03805, 2021

    2021 arXiv

  8. [16]

    One million scenes for autonomous driving: Once dataset.CoRR, abs/2106.11037, 2021

    Jiageng Mao, Minzhe Niu, Chenhan Jiang, Xiaodan Liang, Yamin Li, Chaoqiang Ye, Wei Zhang, Zhenguo Li, Jie Yu, Chunjing Xu, et al. One million scenes for autonomous driving: Once dataset.CoRR, abs/2106.11037, 2021

    2021 arXiv

  9. [17]

    1 Year, 1000km: The Oxford RobotCar Dataset.The International Journal of Robotics Research (IJRR), 36(1):3–15, 2017

    Will Maddern, Geoff Pascoe, Chris Linegar, and Paul Newman. 1 Year, 1000km: The Oxford RobotCar Dataset.The International Journal of Robotics Research (IJRR), 36(1):3–15, 2017

    2017

  10. [18]

    A commute in data: The comma2k19 dataset, 2018

    Harald Schafer, Eder Santana, Andrew Haden, and Riccardo Biasini. A commute in data: The comma2k19 dataset, 2018

    2018

  11. [19]

    Textual explanations for self-driving vehicles.Proceedings of the European Conference on Computer Vision (ECCV), 2018

    Jinkyu Kim, Anna Rohrbach, Trevor Darrell, John Canny, and Zeynep Akata. Textual explanations for self-driving vehicles.Proceedings of the European Conference on Computer Vision (ECCV), 2018

    2018

  12. [20]

    Jakob Geyer, Yohannes Kassahun, Mentar Mahmudi, Xavier Ricou, Rupesh Durgesh, Andrew S. Chung, Lorenz Hauswald, Viet Hoang Pham, Maximilian M ¨uhlegg, Sebastian Dorn, Tiffany Fernandez, Martin J ¨anicke, Sudesh Mirashi, Chiragkumar Savani, Martin Sturm, Oleksandr V orobiov, Ma...

    2020

  13. [21]

    Mona: The munich motion dataset of natural driving

    Luis Gressenbuch, Klemens Esterle, Tobias Kessler, and Matthias Althoff. Mona: The munich motion dataset of natural driving. In2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC), pages 2093–2100, 2022

    2022

  14. [22]

    Rogers, David Han, and Heesung Kwon

    Maggie Wigness, Sungmin Eum, John G. Rogers, David Han, and Heesung Kwon. A rugd dataset for autonomous navigation and visual perception in unstructured outdoor environments. In2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 5000–5007, 2019

    2019

  15. [23]

    Rellis-3d dataset: Data, benchmarks and analysis, 2022

    Peng Jiang, Philip Osteen, Maggie Wigness, and Srikanth Saripalli. Rellis-3d dataset: Data, benchmarks and analysis, 2022

    2022

  16. [24]

    Barfoot, and Franc ¸ois Pomerleau

    Mat ˇej Boxan, Alexander Krawciw, Effie Daum, Xinyuan Qiao, Sven Lilge, Timothy D. Barfoot, and Franc ¸ois Pomerleau. Fomo: A proposal for a multi-season dataset for robot navigation in for ˆet montmorency, 2024

    2024

  17. [25]

    Subt-mrs dataset: Pushing slam towards all- weather environments, 2024

    Shibo Zhao, Yuanjun Gao, Tianhao Wu, Damanpreet Singh, Rushan Jiang, Haoxiang Sun, Mansi Sarawata, Yuheng Qiu, Warren Whittaker, Ian Higgins, Yi Du, Shaoshu Su, Can Xu, John Keller, Jay Karhade, Lucas Nogueira, Sourojit Saha, Ji Zhang, Wenshan Wang, Chen Wang, and Sebastian Sc...

    2024

  18. [26]

    A benchmark for the evaluation of rgb-d slam systems

    J ¨urgen Sturm, Nikolas Engelhard, Felix Endres, Wolfram Burgard, and Daniel Cremers. A benchmark for the evaluation of rgb-d slam systems. In2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, pages 573–580, 2012

    2012

  19. [27]

    (Accessed on Sept

    Google Smartphone Decimeter Challenge 2022 — kag- gle.com.https://www.kaggle.com/competitions/ smartphone-decimeter-2022/discussion/326764. (Accessed on Sept. 15, 2024)

    2022

  20. [28]

    Fast and modular autonomy software for autonomous racing vehicles.Field Robotics, 4(1):1–45, January 2024

    Andrew Saba, Aderotimi Adetunji, Adam Johnson, Aadi Kothari, Matthew Sivaprakasam, Joshua Spisak, Prem Bharatia, Arjun Chauhan, Brendan Duff, Noah Gasparro, Charles King, Ryan Larkin, Brian Mao, Micah Nye, Anjali Parashar, Joseph Attias, Aurimas Balciunas, Austin Brown, Chris ...

    2024

  21. [29]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection.arXiv preprint arXiv:2303.05499, 2023

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Chunyuan Li, Jianwei Yang, Hang Su, Jun Zhu, et al. Grounding dino: Marrying dino with grounded pre-training for open-set object detection.arXiv preprint arXiv:2303.05499, 2023

    2023 arXiv

  22. [30]

    Lang, Sourabh V ora, Holger Caesar, Lubing Zhou, Jiong Yang, and Oscar Beijbom

    Alex H. Lang, Sourabh V ora, Holger Caesar, Lubing Zhou, Jiong Yang, and Oscar Beijbom. Pointpillars: Fast encoders for object detection from point clouds, 2018

    2018

  23. [31]

    Main Street Autonomy.https://mainstreetautonomy. com/. (Accessed Aug. 23, 2024)

    2024

  24. [32]

    Learn to see fast: Lessons learned from autonomous racing on how to develop perception systems.IEEE Access, 11:44034–44050, 2023

    Florian Sauerbeck, Sebastian Huch, Felix Fent, Phillip Karle, Dominik Kulmer, and Johannes Betz. Learn to see fast: Lessons learned from autonomous racing on how to develop perception systems.IEEE Access, 11:44034–44050, 2023

    2023

  25. [33]

    Development and testing of a software stack for an autonomous racing vehicle

    Andrew Saba. Development and testing of a software stack for an autonomous racing vehicle. Master’s thesis, Carnegie Mellon University, Pittsburgh, PA, December 2023

    2023

  26. [34]

    er.autopilot 1.1: A software stack for autonomous racing on oval and road course tracks.IEEE Transactions on Field Robotics, 1:332–359, 2024

    Ayoub Raji, Danilo Caporale, Francesco Gatti, Alessandro Toschi, Nicola Musiu, Micaela Verucchi, Francesco Prignoli, Davide Malat- esta, Andr ´e Fialho Jesus, Andrea Finazzi, Francesco Amerotti, Fabio Bagni, Eugenio Mascaro, Pietro Musso, and Marko Bertogna. er.autopilot 1.1: ...

    2024

  27. [35]

    Tobias Betz, Phillip Karle, Frederik Werner, and Johannes Betz. An analysis of software latency for a high-speed autonomous race car—a case study in the indy autonomous challenge.SAE International Journal of Connected and Automated V ehicles, 6, 02 2023

    2023

  28. [36]

    Towards streaming perception

    Mengtian Li, Yuxiong Wang, and Deva Ramanan. Towards streaming perception. InECCV, 2020

    2020

  29. [37]

    Harley, Zhaoyuan Fang, Jie Li, Rares Ambrus, and Katerina Fragkiadaki

    Adam W. Harley, Zhaoyuan Fang, Jie Li, Rares Ambrus, and Katerina Fragkiadaki. Simple-BEV: What really matters for multi-sensor bev perception? InIEEE International Conference on Robotics and Automation (ICRA), 2023

    2023

  30. [38]

    Bevfusion: Multi-task multi- sensor fusion with unified bird’s-eye view representation

    Zhijian Liu, Haotian Tang, Alexander Amini, Xingyu Yang, Huizi Mao, Daniela Rus, and Song Han. Bevfusion: Multi-task multi- sensor fusion with unified bird’s-eye view representation. InIEEE International Conference on Robotics and Automation (ICRA), 2023

    2023

  31. [39]

    Velociraptor: Leveraging visual foundation models for label-free, risk-aware off-road navigation

    Samuel Triest, Matthew Sivaprakasam, Shubhra Aich, David Fan, Wenshan Wang, and Sebastian Scherer. Velociraptor: Leveraging visual foundation models for label-free, risk-aware off-road navigation. In8th Annual Conference on Robot Learning, 2024

    2024

  32. [40]

    Mixnet: Physics constrained deep neural motion prediction for autonomous racing.IEEE Access, 11:85914–85926, 2023

    Phillip Karle, Ferenc T ¨or¨ok, Maximilian Geisslinger, and Markus Lienkamp. Mixnet: Physics constrained deep neural motion prediction for autonomous racing.IEEE Access, 11:85914–85926, 2023

    2023

  33. [41]

    A tube-mpc approach to autonomous multi-vehicle racing on high-speed ovals.IEEE Transactions on Intelligent V ehicles, 8(1):368–378, 2023

    Alexander Wischnewski, Thomas Herrmann, Frederik Werner, and Boris Lohmann. A tube-mpc approach to autonomous multi-vehicle racing on high-speed ovals.IEEE Transactions on Intelligent V ehicles, 8(1):368–378, 2023

    2023

  34. [42]

    Model parameter identification via a hyperparameter optimization scheme for autonomous racing systems.IEEE Control Systems Letters, 7:1652–1657, 2023

    Hyunki Seong, Chanyoung Chung, and David Hyunchul Shim. Model parameter identification via a hyperparameter optimization scheme for autonomous racing systems.IEEE Control Systems Letters, 7:1652–1657, 2023

    2023

  35. [43]

    Spisak, A

    J. Spisak, A. Saba, N. Suvarna, B. Mao, C. Zhang, C. Chang, S. A. Scherer, and D. Ramanan. Robust modeling and controls for racing on the edge.arXiv preprint arXiv:2205.10841, 2022

    2022 arXiv

  36. [44]

    A tricycle model to accurately control an autonomous racecar with locked differential

    Ayoub Raji, Nicola Musiu, Alessandro Toschi, Francesco Prignoli, Eu- genio Mascaro, Pietro Musso, Francesco Amerotti, Alexander Liniger, Silvio Sorrentino, and Marko Bertogna. A tricycle model to accurately control an autonomous racecar with locked differential. In2023 IEEE 11...

    2023

  37. [45]

    Three- dimensional vehicle dynamics state estimation for high-speed race cars under varying signal quality, 2024

    Sven Goblirsch, Marcel Weinmann, and Johannes Betz. Three- dimensional vehicle dynamics state estimation for high-speed race cars under varying signal quality, 2024

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.