Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

CU-Multi: A Dataset for Multi-Robot Data Association

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 CU-Multi is a multi-robot dataset with designed trajectory overlap and dense LiDAR labels for data-association benchmarks.

desk verdict A genuinely new multi-robot dataset with controlled overlap and semantic LiDAR labels, but the load-bearing ground-truth quality is not quantitatively validated. read the letter →

arxiv 2505.17576 v2 pith:OXB7WOAC submitted 2025-05-23 cs.RO

Doncey Albin , Miles Mena , Annika Thomas , Harel Biggie , Xuefei Sun , Dusty Woods , Steve McGuire , Christoffer Heckman This is my paper
classification cs.RO
keywords multi-robotSLAMdataassociationdatasetLiDARsemanticannotationtrajectoryoverlapcollaborativeloopclosuregeospatialalignment
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

This paper introduces CU-Multi, a real-world multi-robot dataset collected at two outdoor campus sites. Its central claim is that current evaluation practice — splitting a single-robot trajectory into segments to fake multiple robots — fails to capture the pose-dependent variation real multi-robot teams produce, and that a purpose-built dataset with controlled overlap fills that gap. The dataset provides four synchronized runs per environment with deliberately varied trajectory overlap, plus dense semantic labels for every LiDAR scan and geospatially aligned poses. A sympathetic reader would care because these are the ingredients needed to benchmark inter-robot loop closure, map merging, and collaborative SLAM under field-relevant conditions.

What carries the argument

The load-bearing device is the trajectory-overlap structure of Equation (1), which turns a dataset into a parametric testbed: users can pick subsets with high, medium, or sparse cross-robot redundancy. Supporting that is a two-stage annotation pipeline that applies the CENet LiDAR segmentation network zero-shot and then prunes labels using OSM ground truth, and a geospatial alignment stage that couples LIO-SAM with RTK measurements and a fixed starting calibration.

What would settle it

Segment a small subset of CU-Multi scans, hand-label them, and compare against the supplied semantic labels: if the agreement on OSM-filtered classes is not high, the "ground truth" annotations fail. Similarly, align the overlapping scans of robots 1 and 2 at the rendezvous point and measure the residual point-cloud distance under the supplied poses; centimeter-level disagreement would indicate the geospatial ground truth is not accurate enough for inter-robot association benchmarks.

Watch

Extended reading notes

Core claim

The dataset's core claim is that it provides a controlled and structured degree of trajectory overlap across four robots, summarized by $T_1 \approx T_2$, $(T_1 \cup T_2) \subseteq T_3$, and $(T_1 \cup T_2 \cup T_3) \subseteq T_4$: robots 1 and 2 share much of their paths from different viewpoints, robot 3 covers the union of both, and robot 4 covers everything. On top of these runs it supplies RGB-D, GPS with RTK heading, IMU, and LiDAR scans whose labels follow the SemanticKITTI taxonomy, produced by zero-shot CENet inference and then filtered against OpenStreetMap ground truth. A geospatial alignment pipeline based on LIO-SAM with RTK GPS, plus a wheel-chock calibration to fix the starting pose of every run, is offered as ground truth for inter-robot comparison.

Load-bearing premise

The dataset's usefulness as ground truth rests on the zero-shot LiDAR labels and the RTK-aligned poses being accurate enough to benchmark against, but the paper reports no quantitative error metrics for either.

Editorial extensions

If this is right

  • Researchers can evaluate inter-robot loop closure and map merging under explicitly varied overlap instead of hand-split single-robot trajectories.
  • The rendezvous design lets methods be tested on viewpoint variation of the same physical scene, isolating pose-dependent observation effects.
  • Dense semantic LiDAR labels enable evaluation of semantics-based data association and place recognition pipelines.
  • The hierarchical overlap (robot 4 covers all runs) allows principled ablation of the amount of shared observations in multi-robot SLAM experiments.

Reading between the lines

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

  • The controlled overlap fractions could be turned into a quantitative benchmark curve — association accuracy versus overlap percentage — giving a standardized way to compare methods across the community.
  • Because runs were collected at different times with the same platform, the dataset isolates pose-dependent observation variation from platform-dependent variation; a multi-platform extension would be needed to study the latter.
  • The OSM-filtering step assumes the scene matches OpenStreetMap semantics; in areas where it does not, the labels are only as good as CENet's zero-shot predictions, so verifying those filtered regions separately would increase confidence.
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 introduces CU-Multi, a multi-robot dataset collected with a single ground platform over multiple days in two outdoor campus environments. The dataset provides four trajectories per environment with deliberately varied overlap, formalized qualitatively in Eq. (1) as T1 ≈ T2, (T1 ∪ T2) ⊆ T3, and (T1 ∪ T2 ∪ T3) ⊆ T4. Sensor data includes 64-beam LiDAR, RGB-D, IMU, and GPS/RTK, along with LIO-SAM-derived poses, OSM alignment, and per-scan semantic labels produced by zero-shot CENet inference followed by OSM filtering. The paper argues that existing evaluation practice for multi-robot data association relies too heavily on splitting single-robot trajectories and that CU-Multi provides a controlled-overlap alternative with dense semantic annotations. No quantitative validation of the poses or the semantic labels is provided, and Section V explicitly defers benchmark baselines to future work.

Significance. If the dataset is made available with the claimed features, it addresses a real gap in multi-robot SLAM and data association evaluation: controlled, realistic trajectory overlap with dense LiDAR semantics. The authors correctly identify that trajectory splitting from single-robot datasets can create unrealistic shared viewpoints, and CU-Multi's design with explicit overlap levels and a rendezvous endpoint is a valuable idea. The paper also ships support code and conversion scripts for ROS/ROS 2/MCAP, which is a practical strength. However, the scientific value of the dataset as ground truth depends on the accuracy of the poses and labels, and the paper currently provides no quantitative evidence for either. The central claims of 'accurate geospatial alignment' and 'semantically annotated LiDAR' are therefore unsubstantiated at the level required for a benchmark resource.

major comments (4)
  1. [III-C, Table IV, Eq. (1)] The paper claims 'accurate geospatial alignment' and 'precise odometry estimation' using LIO-SAM with RTK GPS, but reports no quantitative pose accuracy metrics (e.g., ATE, RPE, or comparison with raw RTK fixes) for any of the eight trajectories. This is load-bearing because the overlap quantities in Table IV and the containment relationships in Eq. (1) are computed directly from these poses. Without a measure of drift or alignment error, a user cannot know whether the controlled-overlap design holds at the stated 1-meter scale. Please add pose-error statistics for each run, as well as a description of how the wheel-chock calibration was used in the optimization.
  2. [III-D, Table II] The semantic annotations are produced by zero-shot CENet inference followed by OSM filtering, yet they are presented as 'Annotated LiDAR' in Table II without any quantitative validation. No per-class IoU, overall accuracy, or even a sample-based inspection statistic is reported. If the zero-shot labels are noisy or the OSM filter mislabels points, any downstream evaluation of semantics-based data association will be biased in ways invisible to the user. Since the dataset is intended as a benchmark with dense point-wise semantic labels, please report a validation study on a held-out subset (e.g., manually labeled scans or comparison against SemanticKITTI-like labels) and clearly state the expected label accuracy.
  3. [V] Section V explicitly lists 'illustrative baseline benchmarks' as future work, so the paper provides no end-to-end evidence that the dataset behaves as intended for data association tasks. For a dataset paper, at least one sanity-check experiment is important: for example, run a well-known place-recognition or scan-context method on the overlapping trajectory pairs and show that it detects the intended overlap, or run a map-merging pipeline and report success/failure. Without such evidence, the central claim that CU-Multi is 'compelling' and 'ideal' for evaluating these methods is not supported. Please add at least one baseline demonstration in the revised version.
  4. [Table IV] The overlap metric is defined only by a note saying 'the source robot's pose being within 1 meter radius of the target robot's pose,' with no justification for the 1 m threshold and no sensitivity analysis. Since controlled trajectory overlap is a headline feature, the threshold choice should be explained and its influence on the reported overlap distances quantified (e.g., report values for several radii).
minor comments (5)
  1. [III-D, Table II] The column header 'Annotated LiDAR' in Table II is misleading because the provided labels are zero-shot outputs rather than human-verified ground truth. Please rename it to something like 'Semantic labels (zero-shot)' or add a footnote clarifying that the annotations are automatic and unvalidated.
  2. [Eq. (1)] The notation T_i is used as both a trajectory path and a set of poses, and the symbol '≈' is informal. Please define the set representation explicitly (e.g., the set of poses visited by robot i) and state the tolerance for the approximate equality.
  3. [III-C] Please clarify whether the wheel-chock starting-point calibration was used only as an initial pose constraint or as a residual factor in the LIO-SAM graph, and describe how the two GPS antennas were used to derive geospatial heading.
  4. [IV] The text alternates between 'CU-Multi' and 'CU-MULTI'; please standardize the hyphenation and capitalization.
  5. [Fig. 1] The overhead map in Figure 1 labels the two environments but does not identify which trajectory corresponds to robot1 through robot4; adding a legend or matching the color key would make the overlap structure easier to read.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: CU-Multi is a dataset paper with no fitted parameters or derived predictions; its construction choices are self-contained, though ground-truth accuracy is unvalidated (a soundness concern, not circularity).

full rationale

The paper is a dataset description rather than a derivation, and no circular step is present. The claimed contributions—controlled trajectory overlap (Eq. 1), dense LiDAR semantic annotations, and geospatial pose alignment—are constructed from data collection choices (single platform, wheel-chock starting calibration, LIO-SAM with RTK GPS, zero-shot CENet inference filtered by OSM). None of these quantities is fitted to a target prediction, and none is justified by a self-citation chain; LIO-SAM [32], CENet [8], and CoPeD [45] are external references, not the authors' own prior results. The absence of quantitative validation for the semantic labels and poses (Sections III-C and III-D, Table II, with Section V deferring baseline benchmarks) is a correctness and validation limitation, not circularity: the labels and poses are inputs to the dataset, not predictions derived from them. The overlap metric in Table IV is computed from the same estimated poses used to define the dataset, but this is a measurement of self-consistency rather than a derived result that reduces to its input by construction. Therefore, the circularity score is 0.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The paper contributes a dataset, not a theoretical derivation. The central assumptions are domain-level: repeated single-platform runs emulate multi-robot realism, OSM-filtered zero-shot labels are accurate enough for ground truth, and LIO-SAM plus RTK yields trustworthy poses. The only hand-chosen setting is the 1 m overlap threshold. No invented entities are introduced.

free parameters (1)
  • 1 m radius overlap threshold
    Used in Table IV to define trajectory overlap as the source robot's pose being within 1 meter of the target robot's pose. This is a hand-chosen definition rather than a fitted value, and no sensitivity analysis is provided.
assumptions (3)
  • domain assumption Single-platform repeated runs faithfully represent multi-robot observation diversity
    The premise of the dataset is that four runs of the same platform at different times emulate four independent robots. This assumes that viewpoint, lighting, and timing differences capture realistic multi-robot conditions and that inter-run sensor biases are negligible.
  • domain assumption OSM ground truth is sufficient to filter semantic labels
    Section III-D states that inferred semantics are filtered by 'ground truth labeled OSM'. This assumes OSM labels are accurate and complete for the campus and that the alignment between LiDAR and OSM coordinates is correct.
  • domain assumption LIO-SAM with RTK GPS provides accurate ground truth odometry
    Section III-C claims 'precise odometry estimation' using a tightly-coupled LIO-SAM configuration with RTK GPS, but no quantitative accuracy metrics are reported. If the odometry drifts, the geospatial alignment and overlap metrics are affected.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CU-Multi: A Dataset for Multi-Robot Data Association." pith.science (2026). https://pith.science/paper/OXB7WOAC

@misc{pith2026250517576,
  author       = {Pith},
  title        = {Pith review of: CU-Multi: A Dataset for Multi-Robot Data Association},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OXB7WOAC}},
  note         = {Machine review of arXiv:2505.17576}
}
read the original abstract

Multi-robot systems (MRSs) are valuable for tasks such as search and rescue due to their ability to coordinate over shared observations. A central challenge in these systems is aligning independently collected perception data across space and time, i.e., multi-robot data association. While recent advances in collaborative SLAM (C-SLAM), map merging, and inter-robot loop closure detection have significantly progressed the field, evaluation strategies still predominantly rely on splitting a single trajectory from single-robot SLAM datasets into multiple segments to simulate multiple robots. Without careful consideration to how a single trajectory is split, this approach will fail to capture realistic pose-dependent variation in observations of a scene inherent to multi-robot systems. To address this gap, we present CU-Multi, a multi-robot dataset collected over multiple days at two locations on the University of Colorado Boulder campus. Using a single robotic platform, we generate four synchronized runs with aligned start times and deliberate percentages of trajectory overlap. CU-Multi includes RGB-D, GPS with accurate geospatial heading, and semantically annotated LiDAR data. By introducing controlled variations in trajectory overlap and dense lidar annotations, CU-Multi offers a compelling alternative for evaluating methods in multi-robot data association. Instructions on accessing the dataset, support code, and the latest updates are publicly available at https://arpg.github.io/cumulti

Figures

Figures reproduced from arXiv: 2505.17576 by the authors.

Figure 1
Figure 1. Overhead view of the all paths overlaid on map from the Main [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 3
Figure 3. Our platform situated on wheel chocks before data collection (top). [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Example of geospatially-aligned poses through annotated lidar map [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (2 more)
Figure 6
Figure 6. Figure 6: Visualization of the rendezvous point in the [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 5
Figure 5. Figure 5: File structure for CU-Multi dataset. For users interested in testing algorithms through data playback, we provide Python scripts to convert the dataset into various formats, including ROS bags, ROS 2 bags, and MCAP files [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. COSMO-Bench: A Benchmark for Collaborative SLAM Optimization

    cs.RO 2025-08 conditional novelty 7.0 of 10

    A new suite of 24 realistic collaborative SLAM optimization benchmarks derived from real LiDAR data with simulated inter-robot communication.

  2. Learning Adaptive Solvers for Distributed Factor Graph Optimization on Matrix Lie Groups

    cs.RO 2026-07 conditional novelty 6.0 of 10

    A learned feedback policy replaces manual parameter tuning in distributed Riemannian optimization over matrix Lie groups, achieving lower objective values on multi-robot mapping benchmarks.

Reference graph

Works this paper leans on

46 extracted references · 41 canonical work pages · cited by 2 Pith papers

  1. [1]

    Ford multi-av seasonal dataset

    Siddharth Agarwal, Ankit V ora, Gaurav Pandey, Wayne Williams, Helen Kourous, and James McBride. Ford multi-av seasonal dataset. The International Journal of Robotics Research , 39(12):1367–1376, 2020

    work page 2020

  2. [2]

    Lusk, Annika Thomas, and Jonathan P

    Jacqueline Ankenbauer, Parker C. Lusk, Annika Thomas, and Jonathan P. How. Global localization in unstructured environments using semantic object maps built from various viewpoints. In 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 1358–1365, 2023

    work page 2023

  3. [3]

    Semantickitti: A dataset for semantic scene understanding of lidar sequences

    Jens Behley, Martin Garbade, Andres Milioto, Jan Quenzel, Sven Behnke, Cyrill Stachniss, and Jurgen Gall. Semantickitti: A dataset for semantic scene understanding of lidar sequences. In Proceedings of the IEEE/CVF international conference on computer vision , pages 9297–9307, 2019

    work page 2019

  4. [4]

    The m ´alaga urban dataset: High-rate stereo and lidar in a realistic urban scenario

    Jos ´e-Luis Blanco-Claraco, Francisco-Angel Moreno-Duenas, and Javier Gonz´alez-Jim´enez. The m ´alaga urban dataset: High-rate stereo and lidar in a realistic urban scenario. The International Journal of Robotics Research, 33(2):207–214, 2014

    work page 2014

  5. [5]

    The euroc micro aerial vehicle datasets

    Michael Burri, Janosch Nikolic, Pascal Gohl, Thomas Schneider, Joern Rehder, Sammy Omari, Markus W Achtelik, and Roland Siegwart. The euroc micro aerial vehicle datasets. The International Journal of Robotics Research, 35(10):1157–1163, 2016

    work page 2016

  6. [6]

    Multi-robot object slam using distributed variational inference

    Hanwen Cao, Sriram Shreedharan, and Nikolay Atanasov. Multi-robot object slam using distributed variational inference. IEEE Robotics and Automation Letters, 2024

    work page 2024

  7. [7]

    Lamp 2.0: A robust multi-robot slam system for operation in challenging large-scale underground environments

    Yun Chang, Kamak Ebadi, Christopher E Denniston, Muhammad Fad- hil Ginting, Antoni Rosinol, Andrzej Reinke, Matteo Palieri, Jingnan Shi, Arghya Chatterjee, Benjamin Morrell, et al. Lamp 2.0: A robust multi-robot slam system for operation in challenging large-scale underground environments. IEEE Robotics and Automation Letters , 7(4):9175–9182, 2022

    work page 2022

  8. [8]

    Cenet: To- ward concise and efficient lidar semantic segmentation for autonomous driving

    Hui-Xian Cheng, Xian-Feng Han, and Guo-Qiang Xiao. Cenet: To- ward concise and efficient lidar semantic segmentation for autonomous driving. In 2022 IEEE International Conference on Multimedia and Expo (ICME) , pages 01–06. IEEE, 2022

    work page 2022

Show all 46 references
  1. [9]

    Openstreetmap-based lidar global localization in urban environment without a prior lidar map

    Younghun Cho, Giseop Kim, Sangmin Lee, and Jee-Hwan Ryu. Openstreetmap-based lidar global localization in urban environment without a prior lidar map. IEEE Robotics and Automation Letters , 7(2):4999–5006, 2022

    2022

  2. [10]

    Data- efficient decentralized visual slam

    Titus Cieslewski, Siddharth Choudhary, and Davide Scaramuzza. Data- efficient decentralized visual slam. In 2018 IEEE international conference on robotics and automation (ICRA) , pages 2466–2473. IEEE, 2018

    2018

  3. [11]

    Efficient decentralized visual place recognition using a distributed inverted index

    Titus Cieslewski and Davide Scaramuzza. Efficient decentralized visual place recognition using a distributed inverted index. IEEE Robotics and Automation Letters , 2(2):640–647, 2017

    2017

  4. [12]

    Airmu- seum: a heterogeneous multi-robot dataset for stereo-visual and inertial simultaneous localization and mapping

    Rodolphe Dubois, Alexandre Eudes, and Vincent Fr ´emont. Airmu- seum: a heterogeneous multi-robot dataset for stereo-visual and inertial simultaneous localization and mapping. In 2020 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems (...

    2020

  5. [13]

    The mit stata center dataset

    Maurice Fallon, Hordur Johannsson, Michael Kaess, and John J Leonard. The mit stata center dataset. The International Journal of Robotics Research , 32(14):1695–1699, 2013

    2013

  6. [14]

    S3e: A multi-robot multi- modal dataset for collaborative slam

    Dapeng Feng, Yuhua Qi, Shipeng Zhong, Zhiqiang Chen, Qiming Chen, Hongbo Chen, Jin Wu, and Jun Ma. S3e: A multi-robot multi- modal dataset for collaborative slam. IEEE Robotics and Automation Letters, 2024

    2024

  7. [15]

    Multi s- graphs: an efficient distributed semantic-relational collaborative slam

    Miguel Fernandez-Cortizas, Hriday Bavle, David Perez-Saura, Jose Luis Sanchez-Lopez, Pascual Campoy, and Holger V oos. Multi s- graphs: an efficient distributed semantic-relational collaborative slam. IEEE Robotics and Automation Letters , 2024

    2024

  8. [16]

    Vision meets robotics: The kitti dataset

    Andreas Geiger, Philip Lenz, Christoph Stiller, and Raquel Urtasun. Vision meets robotics: The kitti dataset. The international journal of robotics research, 32(11):1231–1237, 2013

    2013

  9. [17]

    Talk resource-efficiently to me: Optimal communication planning for dis- tributed loop closure detection

    Matthew Giamou, Kasra Khosoussi, and Jonathan P How. Talk resource-efficiently to me: Optimal communication planning for dis- tributed loop closure detection. In 2018 IEEE International Conference on Robotics and Automation (ICRA) , pages 3841–3848. IEEE, 2018

    2018

  10. [18]

    Collaborative dynamic 3d scene graphs for automated driving

    Elias Greve, Martin B ¨uchner, Niclas V ¨odisch, Wolfram Burgard, and Abhinav Valada. Collaborative dynamic 3d scene graphs for automated driving. In 2024 IEEE International Conference on Robotics and Automation (ICRA) , pages 11118–11124. IEEE, 2024

    2024

  11. [19]

    Disco- slam: Distributed scan context-enabled multi-robot lidar slam with two-stage global-local graph optimization

    Yewei Huang, Tixiao Shan, Fanfei Chen, and Brendan Englot. Disco- slam: Distributed scan context-enabled multi-robot lidar slam with two-stage global-local graph optimization. IEEE Robotics and Au- tomation Letters , 7(2):1150–1157, 2021

    2021

  12. [20]

    Diter++: Diverse terrain and multi-modal dataset for multi-robot slam in multi-session environments

    Juwon Kim, Hogyun Kim, Seokhwan Jeong, Youngsik Shin, and Younggun Cho. Diter++: Diverse terrain and multi-modal dataset for multi-robot slam in multi-session environments. arXiv preprint arXiv:2412.05839, 2024

    2024 arXiv

  13. [21]

    Multi-robot multi-room exploration with geometric cue extraction and circular decomposition

    Seungchan Kim, Micah Corah, John Keller, Graeme Best, and Sebas- tian Scherer. Multi-robot multi-room exploration with geometric cue extraction and circular decomposition. IEEE Robotics and Automation Letters, 9(2):1190–1197, 2023

    2023

  14. [22]

    Segment anything

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C Berg, Wan-Yen Lo, et al. Segment anything. In Proceedings of the IEEE/CVF international conference on computer vision , pages 4015– 4026, 2023

    2023

  15. [23]

    Swarm-slam: Sparse decentralized collaborative simultaneous localization and mapping framework for multi-robot systems

    Pierre-Yves Lajoie and Giovanni Beltrame. Swarm-slam: Sparse decentralized collaborative simultaneous localization and mapping framework for multi-robot systems. IEEE Robotics and Automation Letters, 9(1):475–482, 2023

    2023

  16. [24]

    Door-slam: Distributed, online, and outlier resilient slam for robotic teams

    Pierre-Yves Lajoie, Benjamin Ramtoula, Yun Chang, Luca Carlone, and Giovanni Beltrame. Door-slam: Distributed, online, and outlier resilient slam for robotic teams. IEEE Robotics and Automation Letters, 5(2):1656–1663, 2020

    2020

  17. [25]

    Towards collaborative simultaneous localization and mapping: a survey of the current research landscape

    Pierre-Yves Lajoie, Benjamin Ramtoula, Fang Wu, and Giovanni Bel- trame. Towards collaborative simultaneous localization and mapping: a survey of the current research landscape. Field Robotics, 2:971–1000, 2022

    2022

  18. [26]

    The utias multi-robot cooperative localization and mapping dataset

    Keith YK Leung, Yoni Halpern, Timothy D Barfoot, and Hugh HT Liu. The utias multi-robot cooperative localization and mapping dataset. The International Journal of Robotics Research , 30(8):969–974, 2011

    2011

  19. [27]

    Kitti-360: A novel dataset and benchmarks for urban scene understanding in 2d and 3d

    Yiyi Liao, Jun Xie, and Andreas Geiger. Kitti-360: A novel dataset and benchmarks for urban scene understanding in 2d and 3d. IEEE Trans- actions on Pattern Analysis and Machine Intelligence , 45(3):3292– 3310, 2022

    2022

  20. [28]

    Grounding dino: Marrying dino with grounded pre-training for open- set object detection

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Qing Jiang, Chunyuan Li, Jianwei Yang, Hang Su, et al. Grounding dino: Marrying dino with grounded pre-training for open- set object detection. In European Conference on Computer Vision , pages 38–55. Spring...

    2024

  21. [29]

    Slideslam: Sparse, lightweight, decentralized metric-semantic slam for multi- robot navigation

    Xu Liu, Jiuzhou Lei, Ankit Prabhu, Yuezhan Tao, Igor Spasojevic, Pratik Chaudhari, Nikolay Atanasov, and Vijay Kumar. Slideslam: Sparse, lightweight, decentralized metric-semantic slam for multi- robot navigation. arXiv preprint arXiv:2406.17249 , 2024

    2024

  22. [30]

    Multi-robot assembly strategies and metrics

    Jeremy A Marvel, Roger Bostelman, and Joe Falco. Multi-robot assembly strategies and metrics. ACM Computing Surveys (CSUR) , 51(1):1–32, 2018

    2018

  23. [31]

    Lego-loam: Lightweight and ground-optimized lidar odometry and mapping on variable terrain

    Tixiao Shan and Brendan Englot. Lego-loam: Lightweight and ground-optimized lidar odometry and mapping on variable terrain. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 4758–4765. IEEE, 2018

    2018

  24. [32]

    Lio-sam: Tightly-coupled lidar inertial odometry via smoothing and mapping

    Tixiao Shan, Brendan Englot, Drew Meyers, Wei Wang, Carlo Ratti, and Rus Daniela. Lio-sam: Tightly-coupled lidar inertial odometry via smoothing and mapping. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 5135–5142. IEEE, 2020

    2020

  25. [33]

    Frame: A modular framework for autonomous map-merging: Advancements in the field

    Nikolaos Stathoulopoulos, Bj ¨orn Lindqvist, Anton Koval, Ali-Akbar Agha-Mohammadi, and George Nikolakopoulos. Frame: A modular framework for autonomous map-merging: Advancements in the field. IEEE Transactions on Field Robotics , 2024

    2024

  26. [34]

    Lusk, Kota Kondo, and Jonathan P

    Annika Thomas, Jouko Kinnari, Parker C. Lusk, Kota Kondo, and Jonathan P. How. Sos-match: Segmentation for open-set robust corre- spondence search and robot localization in unstructured environments. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (...

    2024

  27. [35]

    Kimera-multi: Robust, distributed, dense metric-semantic slam for multi-robot systems

    Yulun Tian, Yun Chang, Fernando Herrera Arias, Carlos Nieto- Granda, Jonathan P How, and Luca Carlone. Kimera-multi: Robust, distributed, dense metric-semantic slam for multi-robot systems. IEEE Transactions on Robotics , 38(4), 2022

    2022

  28. [36]

    How, and Luca Carlone

    Yulun Tian, Yun Chang, Long Quang, Arthur Schang, Carlos Nieto- Granda, Jonathan P. How, and Luca Carlone. Resilient and distributed multi-robot visual slam: Datasets, experiments, and lessons learned. 2023

    2023

  29. [37]

    Wi-closure: Reliable and efficient search of inter- robot loop closures using wireless sensing

    Weiying Wang, Anne Kemmeren, Daniel Son, Javier Alonso-Mora, and Stephanie Gil. Wi-closure: Reliable and efficient search of inter- robot loop closures using wireless sensing. In 2023 IEEE International Conference on Robotics and Automation (ICRA) , pages 2069–2075. IEEE, 2023

    2023

  30. [38]

    Rdc-slam: A real-time distributed cooperative slam system based on 3d lidar

    Yuting Xie, Yachen Zhang, Long Chen, Hui Cheng, Wei Tu, Dongpu Cao, and Qingquan Li. Rdc-slam: A real-time distributed cooperative slam system based on 3d lidar. IEEE Transactions on Intelligent Transportation Systems, 23(9):14721–14730, 2021

    2021

  31. [39]

    Track anything: Segment anything meets videos

    Jinyu Yang, Mingqi Gao, Zhe Li, Shang Gao, Fangjing Wang, and Feng Zheng. Track anything: Segment anything meets videos. arXiv preprint arXiv:2304.11968, 2023

    2023 arXiv

  32. [40]

    Automerge: A framework for map assembling and smoothing in city-scale environments

    Peng Yin, Shiqi Zhao, Haowen Lai, Ruohai Ge, Ji Zhang, Howie Choset, and Sebastian Scherer. Automerge: A framework for map assembling and smoothing in city-scale environments. IEEE Transac- tions on Robotics , 39(5):3686–3704, 2023

    2023

  33. [41]

    A review on map-merging methods for typical map types in multiple-ground- robot slam solutions

    Shuien Yu, Chunyun Fu, Amirali K Gostar, and Minghui Hu. A review on map-merging methods for typical map types in multiple-ground- robot slam solutions. Sensors, 20(23):6988, 2020

    2020

  34. [42]

    Recognize anything: A strong image tagging model

    Youcai Zhang, Xinyu Huang, Jinyu Ma, Zhaoyang Li, Zhaochuan Luo, Yanchun Xie, Yuzhuo Qin, Tong Luo, Yaqian Li, Shilong Liu, et al. Recognize anything: A strong image tagging model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1724...

    2024

  35. [43]

    A review of multi-robot collaborative simultaneous localization and mapping

    Xianzhe Zhao, Shiliang Shao, Ting Wang, Chuxi Fang, Jin Zhang, and Hai Zhao. A review of multi-robot collaborative simultaneous localization and mapping. In 2023 IEEE International Conference on Unmanned Systems (ICUS) , pages 900–905. IEEE, 2023

    2023

  36. [44]

    Dcl-slam: A distributed collaborative lidar slam framework for a robotic swarm

    Shipeng Zhong, Yuhua Qi, Zhiqiang Chen, Jin Wu, Hongbo Chen, and Ming Liu. Dcl-slam: A distributed collaborative lidar slam framework for a robotic swarm. IEEE sensors journal , 24(4):4786–4797, 2023

    2023

  37. [45]

    Coped-advancing multi-robot collaborative perception: A comprehensive dataset in real-world environments

    Yang Zhou, Long Quang, Carlos Nieto-Granda, and Giuseppe Loianno. Coped-advancing multi-robot collaborative perception: A comprehensive dataset in real-world environments. IEEE Robotics and Automation Letters , 2024

    2024

  38. [46]

    Graco: A multimodal dataset for ground and aerial cooperative localization and mapping

    Yilin Zhu, Yang Kong, Yingrui Jie, Shiyou Xu, and Hui Cheng. Graco: A multimodal dataset for ground and aerial cooperative localization and mapping. IEEE Robotics and Automation Letters , 8(2):966–973, 2023. Presented at the 2025 IEEE ICRA Workshop on Field Robotics

    2023

Pith tools

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