pith. sign in

arxiv: 2606.25386 · v1 · pith:IODPC3SMnew · submitted 2026-06-24 · 💻 cs.RO

Commerge: Communication-Efficient, Robust, and Fast LiDAR Map Merging Framework for Multi-Robot Coordination in Resource-Constrained Scenarios

Hogyun Kim , Jiwon Choi , Juwon Kim , Geonmo Yang , Seokhwan Jeong , Hyungtae Lim , Younggun Cho This is my paper

Pith reviewed 2026-06-25 21:15 UTC · model grok-4.3

classification 💻 cs.RO
keywords LiDAR map mergingmulti-robot coordinationcommunication efficiencyscan selectionexchange graphcascaded optimizationmap alignment
0
0 comments X

The pith

Commerge reduces multi-robot LiDAR map merging communication by up to 5000x by exchanging only a carefully selected small subset of scans.

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

The paper introduces Commerge as a framework that solves the communication bottleneck in multi-robot LiDAR map merging. It claims that full alignment quality can be preserved by sharing only a small subset of scans rather than entire datasets. The subset is chosen through a three-stage cascaded optimization on an exchange graph whose vertices are robot keyframes and edges are possible inter-robot connections. This matters in resource-limited settings because full exchange often demands gigabytes of data that exceed available bandwidth. If the claim holds, robot teams can maintain global map consistency while exchanging only megabytes of data.

Core claim

Commerge reduces inter-robot communication by up to 5,000x while maintaining alignment accuracy by exchanging only a small subset of carefully selected scans formulated as a three-stage cascaded optimization on an exchange graph.

What carries the argument

The exchange graph with vertices as robot keyframes and edges as candidate inter-robot loops, processed by three cascade stages that enforce sequential overlap, balanced transmission cost, and geometric-perceptual optimality.

If this is right

  • Map merging becomes practical in low-bandwidth environments such as caves and planetary analogs by dropping data volume from gigabytes to megabytes.
  • Alignment accuracy stays comparable to full-data methods across indoor, outdoor, and embedded-to-desktop platforms.
  • Multi-robot exploration and area coverage can proceed without high-bandwidth links as the communication bottleneck is removed.
  • The same selection logic applies to the five public and four in-house datasets tested, covering diverse terrain types.

Where Pith is reading between the lines

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

  • The graph-based selection might extend to camera or radar data if analogous overlap and perceptual criteria can be defined for those modalities.
  • Larger robot teams would see proportionally larger savings because the per-robot exchange stays bounded by the selected subset size.
  • In rapidly changing scenes the optimization might need periodic re-execution to keep the selected subset current.

Load-bearing premise

A small, sequentially overlapped, balanced, and geometrically-perceptually optimal subset of scans is sufficient to preserve full alignment quality without degradation.

What would settle it

Direct measurement of alignment error on the HeLiPR dataset when merging with the full scan set versus only the selected subset; if error rises beyond the reported comparable level the claim fails.

Figures

Figures reproduced from arXiv: 2606.25386 by Geonmo Yang, Hogyun Kim, Hyungtae Lim, Jiwon Choi, Juwon Kim, Seokhwan Jeong, Younggun Cho.

Figure 1
Figure 1. Figure 1: Overall pipeline of Commerge, a communication-efficient, robust, and fast multi-robot LiDAR map merging framework. (a) Each robot ( ) runs intra-robot SLAM (Section 4.1) to generate SOLiD descriptors, poses, and scans, producing local maps. (b) The server ( ) receives SOLiD descriptors and poses via communication ( ) and constructs an affinity matrix for inter￾robot place recognition (Section 4.2). (c) Seq… view at source ↗
Figure 2
Figure 2. Figure 2: Submaps and associated exchange data sizes across modules in our framework. (a) Without selection, each robot requires approximately 7,000 MB of exchanged data from intra-robot SLAM. The cyan boxes highlight sequentially overlapping regions that are selected in step (A). (b) After sequential matching, i.e., step (A), the exchange volume is reduced to 60.0 MB for robot α and 50.1 MB for robot β. (c) With ba… view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison of LiDAR map reconstruction on the Mulran dataset, which is acquired by an Ouster OS1-64 sensor (Kim et al. 2020). (a) Without loop closure. Accumulated drift leads to noticeable misalignments and structural inconsistencies (red box). (b) With loop closure. The global map exhibits improved structural alignment and reduced distortion, with well-aligned overlapping regions (green box).… view at source ↗
Figure 4
Figure 4. Figure 4: Visualization of inter-robot place recognition results and the corresponding affinity matrix (see Section 4.2). (a) Detected inter-robot loop closures between trajectories consisting of N nodes (robot α) and M nodes (robot β), where green edges represent true positives and red edges represent false positives. (b) N × M affinity matrix A constructed from SOLiD descriptor distances, where the N columns corre… view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of submaps generated under (a) standard minimum vertex cover (MVC) and (b) balanced vertex cover (BVC). The magenta and cyan frames indicate scans from robot α and robot β, respectively, with the transmitted data size shown in MB. While standard MVC achieves lower total transmission volume (2.2 + 47.9 = 50.1 MB vs 25.9 + 26.9 = 52.8 MB), it produces submaps with reduced overlap despite being tak… view at source ↗
Figure 6
Figure 6. Figure 6: (a), C△ assigns each matrix cell a unified pairwise compatibility weight by combining positional consistency κϑ1,ϑ2 and perceptual consistency κϑ, while infeasible pairs are masked out by M (set to zero). For loop selection, we could apply a standard maximum clique approach to identify the largest set of mutually com￾patible loops. However, similar to the limitation discussed in Section 4.4, this approach … view at source ↗
Figure 7
Figure 7. Figure 7: Results before and after applying Commerge for multi-robot map merging in the Multi-Campus Dataset (MCD) acquired with a Livox Mid-70 LiDAR sensor. The magenta box highlights a building structure, and the cyan box highlights trees, both of which appear well-aligned after merging without distortion.  Rˆ α,β init ˆt α,β init 0 ⊤ 1  ∈ SE(3) as follows: Rˆ α,β init , ˆt α,β init = arg min R∈SO(3) t∈R 3 X (p,… view at source ↗
Figure 8
Figure 8. Figure 8: Real-world robot deployments for in-house multi-robot dataset acquisition. (a) Subterranean lava cave with a custom rover. (b) Planetary-analog terrain with a legged robot. (c) Indoor corridor with a wheeled robot. (d) Nighttime campus with a wheeled robot. 0 10m (a) INHA-Cave 0 20m (b) INHA-Planetary 0 10m (c) INHA-Indoor 0 60m (d) INHA-Outdoor [PITH_FULL_IMAGE:figures/full_fig_p015_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Multi-robot map merging results produced by Commerge on each in-house dataset, where different colors indicate maps acquired by different robots. (a) Cave. (b) Planetary-analog terrain. (c) Indoor corridor. (d) Outdoor campus, overlaid on a satellite map for reference [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Point clouds from robot α (left, orange) and robot β (right, blue) in the INHA-Outdoor sequence. Red boxes indicate dynamic pedestrians. where Nsuccess is the number of registration attempts that converge within acceptable error thresholds (i.e., RTE < 2 m and RRE < 5 ◦ ), Ntotal is the total number of registration attempts, tn,GT and ˆtn are the ground truth and estimated translations, and Rn,GT and Rˆ n… view at source ↗
Figure 11
Figure 11. Figure 11: Field communication setup used in our experi￾ments. A central server (i.e., industrial mini PC IPC-6000) communicates with robots through two Omada WiFi Access Points (left/right), forming a WLAN for outdoor trials. This infrastructure was used to exchange global descriptors and LiDAR scans via ROS/ZeroMQ and to measure communication time (tcomm) and total exchange volume (dexch). 7. Experimental Results … view at source ↗
Figure 12
Figure 12. Figure 12: Qualitative comparison between Scan Context (Kim and Kim 2018) and SOLiD (Kim et al. 2024) descriptors. (a) Top view of a query scan (blue) and its matched candidate scan (orange) from NTU 01-10. (b) Scan Context descriptors of the query (left) and candidate retrieved from database (right). (c) SOLiD descriptors of the query (left) and candidate retrieved from database (right). (a) LT-Mapper (Kim and Kim … view at source ↗
Figure 13
Figure 13. Figure 13: Multi-robot map merging results on the Kimera-Multi-Outdoor sequence acquired with an Ouster OS1-64 LiDAR sensor. (a)–(d) Baseline methods exhibit substantial misalignment and structural inconsistencies. (e) KISS-Matcher and (f) our method yield tighter alignment in the overlap regions, with zoom-in views highlighting the resulting structural consistency. 7.1. Map Merging Performance Analysis with State-o… view at source ↗
Figure 14
Figure 14. Figure 14: (a)–(c) Multi-robot map merging performance comparison between baseline approaches and ours for three robots (α, β, γ) on the Town (top) and Roundabout (bottom) sequences acquired with an Ouster OS2-128 LiDAR sensor. As highlighted in zoom boxes, note that our method demonstrates high inter-robot alignment quality. to others. The rightmost columns, together with [PITH_FULL_IMAGE:figures/full_fig_p018_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Multi-robot map merging in geometrically degenerate indoor environments using the Kimera-Multi-Tunnel dataset. (a) Input: Two locally consistent robot maps built from ground truth trajectories with deliberately injected yaw bias to simulate misalignment. (b) Output: Successfully merged map showing tight color co-registration between robots. The cyan and magenta boxes highlight regions where even ground tr… view at source ↗
Figure 16
Figure 16. Figure 16: STD descriptor’s false positive matches in repetitive indoor environments across four Kimera-Multi’s Tunnel sequences: (a) A01-A02, (b) A01-AP01, (c) A01-H01, (d) A01-S02. Black and blue trajectories represent the two robots’ paths, respectively. Cyan and magenta boxes highlight locations that STD incorrectly identifies as the same place despite being at different positions along the trajectories. The zoo… view at source ↗
Figure 17
Figure 17. Figure 17: Visualization of map merging results on the WildPlaces dataset (Knights et al. 2023). (a) Full merged map with improved color scheme. (b) Zoomed-in view of the merged result. (c)–(e) Point-to-point corresponding distances between robot pairs: V02-V03 (6 month gap), V03-V04 (8 month gap), and V02-V04 (14 month gap). Elliptical regions in (d) and (e) highlight areas where V04 exhibits larger corresponding d… view at source ↗
Figure 18
Figure 18. Figure 18: Multi-robot map merging results on the Botanic Garden dataset acquired with a Livox Avia LiDAR sensor. (a–f) Mapping results from different methods, where colored point clouds represent different robots. The red and purple boxes (cyan boxes) highlight critical overlap regions that should align if inter-robot alignment is successful. Prepared using sagej.cls [PITH_FULL_IMAGE:figures/full_fig_p020_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Exchange data over time for each merging round on the (a) NTU 01-10 (Nguyen et al. 2024), (b) Town 01-03 (Jung et al. 2024), and (c) V02-V03-V04 (Knights et al. 2023) sequences. Exchange Data (MB) 800 600 400 200 161284 Time step (min) (a) 8.0 6.0 4.0 2.0 161284 Time step (min) (b) Robot ! Robot " Robot # Robot $ Robot % [PITH_FULL_IMAGE:figures/full_fig_p021_19.png] view at source ↗
Figure 21
Figure 21. Figure 21: Comparison of peak RAM usage (x-axis) and total exchange data sent to the server (y-axis, both in log scale) for different map merging methods on Intel-i7 across two datasets: (a) Kimera-Multi-Outdoor (VLP-16) and (b) HeLiPR (OS2-128). In (b), several baseline methods are not shown as they encountered OOM failures due to the GB-level data scale exceeding available memory capacity. 4.0 3.0 2.0 1.0 0.3 RAM … view at source ↗
Figure 22
Figure 22. Figure 22: RAM consumption of each method across merging rounds on the (a) Kimera-Multi-Outdoor (Tian et al. 2023) and (b) Town (Jung et al. 2024) sequences. all sequences without time out (T/O), while the minimal memory footprint enables deployment even on embedded platforms such as Jet-Nano, where all baseline methods fail due to Out-of-Memory (OOM). Specifically, as shown in Figs. 21 and 22, the minimal exchange … view at source ↗
Figure 23
Figure 23. Figure 23: Qualitative comparison with Multi-Proxy (Wang et al. 2026) under network degradation simulated via NetEm (Hemminger 2005) on the planetary terrain dataset. (a)–(b) Network delay (10 ms vs. 2,000 ms), (c)–(d) bandwidth limitation (≤20 Mbit/s vs. ≤2 Mbit/s), (e)–(f) packet loss (10% vs. 50%). Green and red boxes indicate successful and failed merging. Prepared using sagej.cls [PITH_FULL_IMAGE:figures/full_… view at source ↗
Figure 24
Figure 24. Figure 24: Multi-robot map merging results on a subterranean cave dataset under various network conditions. (a) Before map merging. (b) Map merging without network degradation. (c) Map merging under 2000 ms delay. (d) Map merging under 2 Mbit/s bandwidth limit. (e) Map merging under 50% packet loss. 0 10m (a) 0 60m (b) Map by robot ! Map by robot " Map by robot # [PITH_FULL_IMAGE:figures/full_fig_p024_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: Multi-robot map merging results on the (a) INHA-Indoor and (b) INHA-Outdoor. Left: individual robot maps before merging. Right: merged map produced by Commerge. 0 10m (a) 0 60m (b) Map by robot ! Map by robot " Map by robot # [PITH_FULL_IMAGE:figures/full_fig_p024_25.png] view at source ↗
Figure 26
Figure 26. Figure 26: Multi-robot map merging results on the (a) INHA-Indoor and (b) INHA-Outdoor datasets, where each robot transmits its data to the server over a real wireless network and the server performs map merging using the received data. Left: individual robot maps before merging. Right: merged map produced by Commerge. The icon indicates the actual location of the server on each map, where signals are blocked by wal… view at source ↗
Figure 27
Figure 27. Figure 27: Sequential matching comparison on the HeLiPR dataset (Jung et al. 2024) between (a) AutoMerge (Yin et al. 2023) and (b) our proposed method. For each method, the left panel shows the affinity matrix where the highlighted box indicates the sequentially matched region, and the right panel shows the merged submap constructed from the submaps of robot α and robot β corresponding to the selected region [PITH_… view at source ↗
Figure 28
Figure 28. Figure 28: Box plot analysis of RMSE distribution across different pipeline configurations on (a) MCD NTU 01-10 and (b) HeLiPR Town 01-03 sequences. , , and represent the sequential matching, balanced minimum vertex cover, and maximum edge-weighted clique selection, respectively. The ∗∗∗∗ annotations indicate measurements with a p-value < 10−4 after a paired t-test. robot, corresponding to a 99.6% reduction in trans… view at source ↗
Figure 29
Figure 29. Figure 29: Submaps and associated exchange data sizes across modules in our framework. (a) From intra-robot SLAM, each robot would require approximately 100 MB–200 MB of exchanged data. The cyan boxes highlight sequentially overlapping regions that are selected in step (A). (b) After sequential matching, i.e., step (A), the exchange volume is reduced to 8.1 MB for robot α and 9.0 MB for robot β. (c) With balanced mi… view at source ↗
Figure 30
Figure 30. Figure 30: Validation of our three-stage optimization with Scan Context descriptors on Kimera-Multi-Outdoor partial overlap scenario to demonstrate the effectiveness of descriptor-agnosticism. (a) Affinity matrix constructed from Scan Context similarities, where sequential matching identifies the optimal cluster (cyan box) representing sequentially coherent inter-robot correspondences despite the less discriminative… view at source ↗
Figure 31
Figure 31. Figure 31: Temporal evolution of map merging quality on HeLiPR Town 01-03 sequence at different mission stages. Each scenario shows the merged global map constructed at the server (top) and the selected submaps transmitted from the robots to the server under the proposed exchange policy (bottom). (a) Both robots after covering the first quarter of the mission (10 min): successful alignment despite neither robot havi… view at source ↗
Figure 32
Figure 32. Figure 32: Execution time breakdown across hardware platforms (Intel-i7: Intel i7, IPC-6000: IPC-6000, Jet-Nano: Jetson Nano) and operational scales: (a) small-scale NTU 01-10 and (b) large-scale Town 01-03 sequences [PITH_FULL_IMAGE:figures/full_fig_p028_32.png] view at source ↗
read the original abstract

By maintaining global consistency across robot teams, multi-robot LiDAR map merging enables faster exploration and efficient area coverage. However, map merging requires exchanging massive sensor data between the server and robots, making communication the bottleneck, especially in communication-constrained environments. Therefore, we present Commerge, a communication-efficient map merging framework that achieves bandwidth reduction through graph-theoretic selective data exchange. By doing so, our Commerge reduces inter-robot communication by up to 5,000x while maintaining alignment accuracy. Our key insight is that only a small subset of carefully selected scans is sufficient for robust map merging. We formulate this as a three-stage cascaded optimization problem on an exchange graph, where vertices represent robot keyframes and edges denote candidate inter-robot loops. Through three cascade stages, we select a sequentially overlapped, balanced-transmission-cost, and geometrically-perceptually optimal scan subset that preserves alignment quality while reducing communication. Unlike existing approaches that either transmit whole scans, which require GB-scale data exchange, or employ naive downsampling, our approach exchanges only MB-scale data while achieving comparable alignment accuracy. Extensive evaluation on five public datasets and four in-house datasets covering cave, planetary-analog, indoor, and outdoor campus environments shows up to 99.98% reduction in data exchange (e.g., from 7,000MB to 1.3MB on the HeLiPR dataset), while maintaining alignment performance across embedded to desktop platforms. The supplementary materials are available at https://sparolab.github.io/research/commerge.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The paper presents Commerge, a framework for multi-robot LiDAR map merging that reduces communication by selecting a minimal subset of scans via three-stage cascaded optimization (sequential overlap, balanced cost, geometric-perceptual optimality) on an exchange graph. It claims up to 5,000x (99.98%) bandwidth reduction (e.g., 7 GB to 1.3 MB on HeLiPR) while maintaining alignment accuracy, validated on five public and four in-house datasets across cave, planetary, indoor, and outdoor settings.

Significance. If the accuracy-preservation claim holds, the result is significant for resource-constrained multi-robot coordination, enabling GB-to-MB data exchange without full-scan transmission or naive downsampling, directly addressing bandwidth bottlenecks in exploration and coverage tasks.

major comments (1)
  1. [Evaluation (datasets and quantitative results)] The load-bearing claim that the three-stage cascaded optimization on the exchange graph yields a subset preserving full alignment quality (no degradation relative to the complete scan set) is only partially supported by the reported 'comparable alignment accuracy.' The evaluation must explicitly demonstrate that the sequentially overlapped + balanced + geometrically-perceptually optimal subset captures all necessary loop closures and geometric constraints; otherwise the downstream pose-graph optimizer could silently lose critical edges. This requires ablation or statistical equivalence tests (e.g., ATE or map-overlap metrics with confidence intervals) comparing subset vs. full-set results on each dataset.
minor comments (2)
  1. [Method (three-stage optimization)] Clarify the exact definitions and weighting of the three optimality criteria in the cascaded stages; the abstract describes them at a high level but the formulation details (objective functions, thresholds) are needed for reproducibility.
  2. [Abstract and results] The 5,000x reduction figure and the specific 7,000 MB to 1.3 MB example should be tied to a table or figure showing per-dataset communication volumes and the corresponding alignment metrics side-by-side with baselines.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comment below and will strengthen the evaluation with additional analyses in the revised version.

read point-by-point responses

  1. Referee: The load-bearing claim that the three-stage cascaded optimization on the exchange graph yields a subset preserving full alignment quality (no degradation relative to the complete scan set) is only partially supported by the reported 'comparable alignment accuracy.' The evaluation must explicitly demonstrate that the sequentially overlapped + balanced + geometrically-perceptually optimal subset captures all necessary loop closures and geometric constraints; otherwise the downstream pose-graph optimizer could silently lose critical edges. This requires ablation or statistical equivalence tests (e.g., ATE or map-overlap metrics with confidence intervals) comparing subset vs. full-set results on each dataset.

    Authors: We agree that explicit verification of constraint preservation is required to substantiate the no-degradation claim. While the manuscript reports comparable accuracy on nine datasets, we acknowledge that direct ablations and statistical tests would more rigorously confirm retention of all critical loop closures. In revision we will add: (1) stage-wise ablation quantifying preserved inter-robot edges, and (2) ATE and map-overlap metrics with 95% confidence intervals comparing subset versus full-set results on every dataset. These additions will directly address potential silent loss of geometric constraints. revision: yes

Circularity Check

0 steps flagged

No circularity: independent formulation of cascaded graph optimization

full rationale

The paper formulates the subset selection directly as a three-stage cascaded optimization on an exchange graph whose vertices and edges are defined from the input keyframes and candidate loops. No equations reduce a fitted parameter to a renamed prediction, no self-citation supplies a load-bearing uniqueness theorem, and no ansatz is smuggled in. The central claim (MB-scale exchange preserves alignment) is presented as an empirical outcome of the stated optimality criteria rather than a definitional identity. This is the normal self-contained case.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the central claim rests on the domain assumption that selective scan subsets suffice for alignment; no explicit free parameters, invented entities, or additional axioms are detailed.

axioms (1)
  • domain assumption A small subset of scans selected via cascaded graph optimization is sufficient for robust map merging without loss of alignment quality
    This premise underpins the communication reduction claim and is invoked in the description of the three-stage process.

pith-pipeline@v0.9.1-grok · 5841 in / 1328 out tokens · 28780 ms · 2026-06-25T21:15:25.415828+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

98 extracted references · 1 canonical work pages

  1. [1]

    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Ao S, Hu Q, Wang H, Xu K and Guo Y (2023) BUFFER: Balancing accuracy, efficiency, and generalizability in point cloud registration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 1255--1264

    2023

  2. [2]

    In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Arandjelovic R, Gronat P, Torii A, Pajdla T and Sivic J (2016) NetVLAD: CNN architecture for weakly supervised place recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 5297--5307

    2016

  3. [3]

    Science Robotics 8(80): eade9548

    Arm P, Waibel G, Preisig J, Tuna T, Zhou R, Bickel V, Ligeza G, Miki T, Kehl F, Kolvenbach H and others (2023) Scientific exploration of challenging planetary analog environments with a team of legged robots. Science Robotics 8(80): eade9548

    2023

  4. [4]

    IEEE Robotics and Automation Letters 6(2): 855--862

    Bernreiter L, Ott L, Nieto J, Siegwart R and Cadena C (2021) PHASER: A robust and correspondence‑free global pointcloud registration. IEEE Robotics and Automation Letters 6(2): 855--862

    2021

  5. [5]

    In: Sensor Fusion IV: Control Paradigms and Data Structures, volume 1611, pp

    Besl PJ and McKay ND (1992) Method for registration of 3-D shapes. In: Sensor Fusion IV: Control Paradigms and Data Structures, volume 1611, pp. 586--606

    1992

  6. [6]

    The International Journal of Robotics Research 43(4): 485--512

    Best G, Garg R, Keller J, Hollinger GA and Scherer S (2024) Multi-robot, multi-sensor exploration of multifarious environments with full mission aerial autonomy. The International Journal of Robotics Research 43(4): 485--512

    2024

  7. [7]

    IEEE Transactions on Robotics 21(3): 376--386

    Burgard W, Moors M, Stachniss C and Schneider FE (2005) Coordinated multi-robot exploration. IEEE Transactions on Robotics 21(3): 376--386

    2005

  8. [8]

    In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), pp

    Cao Y, Wang Y and Chen H (2025) Real-time LiDAR point cloud compression and transmission for resource-constrained robots. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), pp. 12789--12795

    2025

  9. [9]

    IEEE Robotics and Automation Letters 7(4): 9175--9182

    Chang Y, Ebadi K, Denniston CE, Ginting MF, Rosinol A, Reinke A, Palieri M, Shi J, Chatterjee A, Morrell B and others (2022) 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

  10. [10]

    IEEE Transactions on Instrumentation and Measurement

    Chang G, Shen H, Hui Y, Zhang X, Lu J, Lu H and Tian B (2025) LiDAR-based decentralized collaborative localization and mapping for multi-UAV with stable triangle descriptor. IEEE Transactions on Instrumentation and Measurement

    2025

  11. [11]

    IEEE Robotics and Automation Letters 2(2): 640--647

    Cieslewski T and Scaramuzza D (2017) Efficient decentralized visual place recognition using a distributed inverted index. IEEE Robotics and Automation Letters 2(2): 640--647

    2017

  12. [12]

    In: Proceedings of the 2018 IEEE International Conference on Robotics and Automation (ICRA), pp

    Cieslewski T, Choudhary S and Scaramuzza D (2018) Data-efficient decentralized visual SLAM. In: Proceedings of the 2018 IEEE International Conference on Robotics and Automation (ICRA), pp. 2466--2473

    2018

  13. [13]

    SICE Journal of Control, Measurement, and System Integration 10(6): 495--503

    Cortés J and Egerstedt M (2017) Coordinated control of multi-robot systems: A survey. SICE Journal of Control, Measurement, and System Integration 10(6): 495--503

    2017

  14. [14]

    John Wiley & Sons

    Cover TM (1999) Elements of Information Theory. John Wiley & Sons

    1999

  15. [15]

    Journal of Intelligent & Robotic Systems 110(1): 22

    Damigos G, Stathoulopoulos N, Koval A, Lindgren T, Nikolakopoulos G (2024) Communication-Aware Control of Large Data Transmissions via Centralized Cognition and 5G Networks for Multi-Robot Map Merging. Journal of Intelligent & Robotic Systems 110(1): 22

    2024

  16. [16]

    IEEE Transactions on Image Processing 25(8): 3947--3956

    De Queiroz RL and Chou PA (2016) Compression of 3D point clouds using a region-adaptive hierarchical transform. IEEE Transactions on Image Processing 25(8): 3947--3956

    2016

  17. [17]

    Technical Report GT-RIM-CP&R-2012-002, Georgia Institute of Technology

    Dellaert F (2012) Factor graphs and GTSAM: A hands-on introduction. Technical Report GT-RIM-CP&R-2012-002, Georgia Institute of Technology

    2012

  18. [18]

    Communications of the ACM 24(6): 381--395

    Fischler MA and Bolles RC (1981) Random sample consensus: A paradigm for model fitting with applications to image analysis and automated cartography. Communications of the ACM 24(6): 381--395

    1981

  19. [19]

    In: Proceedings of the 19th IEEE International Conference on Factory Communication Systems (WFCS), pp

    Formis G, Scanzio S, Cena G, Valenzano A (2023) Predicting wireless channel quality by means of moving averages and regression models. In: Proceedings of the 19th IEEE International Conference on Factory Communication Systems (WFCS), pp. 1--8

    2023

  20. [20]

    In: Proceedings of the 2018 IEEE International Conference on Robotics and Automation (ICRA), pp

    Giamou M, Khosoussi K and How JP (2018) Talk resource-efficiently to me: Optimal communication planning for distributed loop closure detection. In: Proceedings of the 2018 IEEE International Conference on Robotics and Automation (ICRA), pp. 3841--3848

    2018

  21. [21]

    Current Robotics Reports 3(4): 213--225

    Gielis J, Shankar A, Prorok A (2022) A Critical Review of Communications in Multi-Robot Systems. Current Robotics Reports 3(4): 213--225

    2022

  22. [22]

    arXiv preprint arXiv:2502.13452

    Gil H, Lee D, Kim G and Kim A (2025) Ephemerality meets LiDAR-based lifelong mapping. arXiv preprint arXiv:2502.13452

    arXiv 2025

  23. [23]

    APSIPA Transactions on Signal and Information Processing 9: e13

    Graziosi D, Nakagami O, Kuma S, Zaghetto A, Suzuki T and Tabatabai A (2020) An overview of ongoing point cloud compression standardization activities: Video-based (V-PCC) and geometry-based (G-PCC). APSIPA Transactions on Signal and Information Processing 9: e13

    2020

  24. [24]

    Available at: https://github.com/MichaelGrupp/evo

    Grupp M (2017) evo: Python package for the evaluation of odometry and SLAM. Available at: https://github.com/MichaelGrupp/evo

    2017

  25. [25]

    arXiv preprint arXiv:2102.03804

    He D, Xu W, Zhang F (2021) Kalman filters on differentiable manifolds. arXiv preprint arXiv:2102.03804

    arXiv 2021

  26. [26]

    Journal of Field Robotics

    He K, Jia R, Hong H, Wang N and Hu Y (2025) LDG-CSLAM: Multi-robot collaborative SLAM based on curve analysis, normal distribution, and factor graph optimization. Journal of Field Robotics

    2025

  27. [27]

    In: Proceedings of Linux Conf Au, vol

    Hemminger S (2005) Network emulation with NetEm. In: Proceedings of Linux Conf Au, vol. 5

    2005

  28. [28]

    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Huang S, Gojcic Z, Usvyatsov M, Wieser A and Schindler K (2021) Predator: Registration of 3D point clouds with low overlap. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 4267--4276

    2021

  29. [29]

    IEEE Robotics and Automation Letters 7(2): 1150--1157

    Huang Y, Shan T, Chen F and Englot B (2021) Disco-SLAM: Distributed scan context-enabled multi-robot LiDAR SLAM with two-stage global-local graph optimization. IEEE Robotics and Automation Letters 7(2): 1150--1157

    2021

  30. [30]

    arXiv preprint arXiv:2507.17130

    Jeong S, Kim H and Cho Y (2025) MARSCalib: Multi-robot, Automatic, Robust, Spherical Target-based Extrinsic Calibration in Field and Extraterrestrial Environments. arXiv preprint arXiv:2507.17130

    arXiv 2025

  31. [31]

    The International Journal of Robotics Research 43(12): 1867--1883

    Jung M, Yang W, Lee D, Gil H, Kim G and Kim A (2024) HeLiPR: Heterogeneous LiDAR dataset for inter-LiDAR place recognition under spatiotemporal variations. The International Journal of Robotics Research 43(12): 1867--1883

    2024

  32. [32]

    IEEE Transactions on Intelligent Vehicles

    Kang G, Kim H, Choi B, Jeong S, Shin Y-S and Cho Y (2025) Uni-Mapper: Unified Mapping Framework for Multi-modal LiDARs in Complex and Dynamic Environments. IEEE Transactions on Intelligent Vehicles

    2025

  33. [33]

    Journal of the Operational Research Society 49(3): 237--252

    Kelly FP, Maulloo AK, Tan DKH (1998) Rate Control for Communication Networks: Shadow Prices, Proportional Fairness and Stability. Journal of the Operational Research Society 49(3): 237--252

    1998

  34. [34]

    In: Cooperative Robots and Sensor Networks 2015, pp

    Khamis A, Hussein A and Elmogy A (2015) Multi-robot task allocation: A review of the state-of-the-art. In: Cooperative Robots and Sensor Networks 2015, pp. 31--51

    2015

  35. [35]

    In: Proceedings of the 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp

    Kim G and Kim A (2018) Scan Context: Egocentric spatial descriptor for place recognition within 3D point cloud map. In: Proceedings of the 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 4802--4809

    2018

  36. [36]

    In: Proceedings of the 2020 IEEE International Conference on Robotics and Automation (ICRA), pp

    Kim G, Park YS, Cho Y, Jeong J and Kim A (2020) Mulran: Multimodal range dataset for urban place recognition. In: Proceedings of the 2020 IEEE International Conference on Robotics and Automation (ICRA), pp. 6246--6253

    2020

  37. [37]

    In: Proceedings of the 2022 IEEE International Conference on Robotics and Automation (ICRA), pp

    Kim G and Kim A (2022) LT-Mapper: A modular framework for LiDAR-based lifelong mapping. In: Proceedings of the 2022 IEEE International Conference on Robotics and Automation (ICRA), pp. 7995--8002

    2022

  38. [38]

    In: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), pp

    Kim J, Kim H, Jeong S, Shin Y, Cho Y (2025) DiTer++: Diverse terrain and multi-modal dataset for multi-robot SLAM in multi-session environments. In: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), pp. 12187--12193

    2025

  39. [39]

    IEEE Robotics and Automation Letters

    Kim H, Choi J, Sim T, Kim G and Cho Y (2024) Narrowing your FOV with SOLiD: Spatially organized and lightweight global descriptor for FOV-constrained LiDAR place recognition. IEEE Robotics and Automation Letters

    2024

  40. [40]

    arXiv preprint arXiv:2505.08230

    Kim H, Choi J, Kim J, Yang G, Cho D, Lim H and Cho Y (2025) SKiD-SLAM: Robust, Lightweight, and Distributed Multi-Robot LiDAR SLAM in Resource-Constrained Field Environments. arXiv preprint arXiv:2505.08230

    arXiv 2025

  41. [41]

    In: Proceedings of the 2023 IEEE International Conference on Robotics and Automation (ICRA), pp

    Knights J, Vidanapathirana K, Ramezani M, Sridharan S, Fookes C and Moghadam P (2023) WILD-Places: A large-scale dataset for LiDAR place recognition in unstructured natural environments. In: Proceedings of the 2023 IEEE International Conference on Robotics and Automation (ICRA), pp. 11322--11328

    2023

  42. [42]

    In: Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2003), Vol

    Ko J, Stewart B, Fox D, Konolige K and Limketkai B (2003) A practical, decision-theoretic approach to multi-robot mapping and exploration. In: Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2003), Vol. 4, pp. 3232--3238

    2003

  43. [43]

    In: 2021 IEEE International Conference on Robotics and Automation (ICRA), pp

    Koide K, Yokozuka M, Oishi S and Banno A (2021) Voxelized GICP for fast and accurate 3D point cloud registration. In: 2021 IEEE International Conference on Robotics and Automation (ICRA), pp. 11054--11059

    2021

  44. [44]

    Journal of Open Source Software 9(100): 6948

    Koide K (2024) small\_gicp: Efficient and parallel algorithms for point cloud registration. Journal of Open Source Software 9(100): 6948. doi:10.21105/joss.06948 https://doi.org/10.21105/joss.06948

    work page doi:10.21105/joss.06948 2024

  45. [45]

    IEEE Transactions on Vehicular Technology 69(1): 443--456

    Kulkarni A, Seetharam A, Ramesh A, Herath JD (2019) DeepChannel: Wireless Channel Quality Prediction Using Deep Learning. IEEE Transactions on Vehicular Technology 69(1): 443--456

    2019

  46. [46]

    Field Robotics 2: 971--1000

    Lajoie P - Y, Ramtoula B, Wu F, Beltrame G (2022) Towards Collaborative Simultaneous Localization and Mapping: A Survey of the Current Research Landscape. Field Robotics 2: 971--1000

    2022

  47. [47]

    IEEE Robotics and Automation Letters 9(1): 475--482

    Lajoie P-Y and Beltrame G (2023) Swarm-SLAM: Sparse decentralized collaborative simultaneous localization and mapping framework for multi-robot systems. IEEE Robotics and Automation Letters 9(1): 475--482

    2023

  48. [48]

    IEEE Robotics and Automation Letters

    Lajoie P-Y, Ramtoula B, De Martini D and Beltrame G (2025) 3D foundation model-based loop closing for decentralized collaborative SLAM. IEEE Robotics and Automation Letters

    2025

  49. [49]

    IEEE Transactions on Field Robotics

    Lajoie P-Y, Soma K, Bong HM, Lemieux-Bourque A, Zhang R, Varadharajan VS and Beltrame G (2025) Multi-robot decentralized collaborative SLAM in planetary analogue environments: Dataset, challenges, and lessons learned. IEEE Transactions on Field Robotics

    2025

  50. [50]

    arXiv preprint arXiv:2409.15615

    Lim H, Kim D, Shin G, Shi J, Vizzo I, Myung H, Park J and Carlone L (2024) KISS-Matcher: Fast and robust point cloud registration revisited. arXiv preprint arXiv:2409.15615

    arXiv 2024

  51. [51]

    arXiv preprint arXiv:2511.00635

    Lim H, Kim D, Myung H (2025) Multi-Mapcher: Loop Closure Detection-Free Heterogeneous LiDAR Multi-Session SLAM Leveraging Outlier-Robust Registration for Autonomous Vehicles. arXiv preprint arXiv:2511.00635

    arXiv 2025

  52. [52]

    IEEE Robotics and Automation Letters 9(3): 2798--2805

    Liu Y, Fu Y, Qin M, Xu Y, Xu B, Chen F, Goossens B, Sun PZ, Yu H, Liu C and others (2024) BotanicGarden: A high-quality dataset for robot navigation in unstructured natural environments. IEEE Robotics and Automation Letters 9(3): 2798--2805

    2024

  53. [53]

    IEEE Robotics and Automation Letters

    Liu J, Xu X, Lu S, Wang S, Wang C, Xiong R and Wang Y (2025) Sparse hierarchical LiDAR bundle adjustment for online collaborative localization and mapping. IEEE Robotics and Automation Letters

    2025

  54. [54]

    IEEE Transactions on Robotics

    Luo L, Cao S-Y, Li X, Xu J, Ai R, Yu Z and Chen X (2025) BEVPlace++: Fast, robust, and lightweight LiDAR global localization for unmanned ground vehicles. IEEE Transactions on Robotics

    2025

  55. [55]

    IEEE Robotics and Automation Letters 7(3): 6958--6965

    Ma J, Zhang J, Xu J, Ai R, Gu W and Chen X (2022) OverlapTransformer: An efficient and yaw-angle-invariant transformer network for LiDAR-based place recognition. IEEE Robotics and Automation Letters 7(3): 6958--6965

    2022

  56. [56]

    IEEE Transactions on Industrial Informatics 20(3): 4039--4048

    Ma J, Xiong G, Xu J and Chen X (2023) CVTNet: A cross-view transformer network for LiDAR-based place recognition in autonomous driving environments. IEEE Transactions on Industrial Informatics 20(3): 4039--4048

    2023

  57. [57]

    Expert Systems with Applications 173: 114660

    Madridano A, Al-Kaff A, Martín D and De La Escalera A (2021) Trajectory planning for multi-robot systems: Methods and applications. Expert Systems with Applications 173: 114660

    2021

  58. [58]

    In: Proceedings of the 2018 IEEE International Conference on Robotics and Automation (ICRA), pp

    Mangelson JG, Dominic D, Eustice RM and Vasudevan R (2018) Pairwise consistent measurement set maximization for robust multi-robot map merging. In: Proceedings of the 2018 IEEE International Conference on Robotics and Automation (ICRA), pp. 2916--2923

    2018

  59. [59]

    In: Proceedings of the 2012 IEEE International Conference on Robotics and Automation (ICRA), pp

    Milford MJ and Wyeth GF (2012) SeqSLAM: Visual route-based navigation for sunny summer days and stormy winter nights. In: Proceedings of the 2012 IEEE International Conference on Robotics and Automation (ICRA), pp. 1643--1649

    2012

  60. [60]

    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Nguyen T-M, Yuan S, Nguyen TH, Yin P, Cao H, Xie L, Wozniak M, Jensfelt P, Thiel M, Ziegenbein J and others (2024) MCD: Diverse large-scale multi-campus dataset for robot perception. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 22304--22313

    2024

  61. [61]

    Robotics and Autonomous Systems 112: 13--21

    Oishi S, Inoue Y, Miura J and Tanaka S (2019) SeqSLAM++: View-based robot localization and navigation. Robotics and Autonomous Systems 112: 13--21

    2019

  62. [62]

    Sensors 23(7): 3625

    Orr J and Dutta A (2023) Multi-agent deep reinforcement learning for multi-robot applications: A survey. Sensors 23(7): 3625

    2023

  63. [63]

    Autonomous Robots 34(3): 133--148

    Pomerleau F, Colas F, Siegwart R and Magnenat S (2013) Comparing ICP variants on real-world data sets: Open-source library and experimental protocol. Autonomous Robots 34(3): 133--148

    2013

  64. [64]

    Autonomous Robots 11(2): 117--136

    Roy N and Dudek G (2001) Collaborative robot exploration and rendezvous: Algorithms, performance bounds and observations. Autonomous Robots 11(2): 117--136

    2001

  65. [65]

    In: Proceedings of the IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct), pp

    Schmuck P, Ziegler T, Karrer M, Perraudin J and Chli M (2021) COVINS: Visual-inertial SLAM for centralized collaboration. In: Proceedings of the IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct), pp. 171--176

    2021

  66. [66]

    Algorithms and Combinatorics, vol

    Schrijver A (2003) Combinatorial Optimization: Polyhedra and Efficiency. Algorithms and Combinatorics, vol. 24

    2003

  67. [67]

    In: Robotics: Science and Systems (RSS)

    Segal A, Haehnel D and Thrun S (2009) Generalized-ICP. In: Robotics: Science and Systems (RSS). Seattle, WA, USA, pp. 435

    2009

  68. [68]

    In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp

    Seo M, Lim H, Lee K, Carlone L and Park J (2025) BUFFER-X: Towards zero-shot point cloud registration in diverse scenes. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp. 3851--3862

    2025

  69. [69]

    The Bell System Technical Journal 27(3): 379--423

    Shannon CE (1948) A Mathematical Theory of Communication. The Bell System Technical Journal 27(3): 379--423

    1948

  70. [70]

    In: Proceedings of the 2021 IEEE International Conference on Robotics and Automation (ICRA), pp

    Shi J, Yang H and Carlone L (2021) ROBIN: A graph-theoretic approach to reject outliers in robust estimation using invariants. In: Proceedings of the 2021 IEEE International Conference on Robotics and Automation (ICRA), pp. 13820--13827

    2021

  71. [71]

    In: Proceedings of the 2017 IEEE International Conference on Robotics and Automation (ICRA), pp

    Siam SM and Zhang H (2017) Fast-SeqSLAM: A fast appearance based place recognition algorithm. In: Proceedings of the 2017 IEEE International Conference on Robotics and Automation (ICRA), pp. 5702--5708

    2017

  72. [72]

    arXiv preprint arXiv:2301.09213

    Stathoulopoulos N, Koval A, Agha-mohammadi A-A and Nikolakopoulos G (2023) Frame: Fast and robust autonomous 3D point cloud map-merging for egocentric multi-robot exploration. arXiv preprint arXiv:2301.09213

    arXiv 2023

  73. [73]

    IEEE Transactions on Field Robotics

    Stathoulopoulos N, Lindqvist B, Koval A, Agha-Mohammadi AA and Nikolakopoulos G (2024) FRAME: A modular framework for autonomous map-merging: Advancements in the field. IEEE Transactions on Field Robotics

    2024

  74. [74]

    Expert Systems with Applications 237: 121508

    Stathoulopoulos N, Koval A and Nikolakopoulos G (2024) 3DEG: Data-driven descriptor extraction for global re-localization in subterranean environments. Expert Systems with Applications 237: 121508

    2024

  75. [75]

    IEEE Robotics and Automation Letters 11(1): 738--745

    Stathoulopoulos N, Kanellakis C and Nikolakopoulos G (2025) A minimal subset approach for informed keyframe sampling in large-scale SLAM. IEEE Robotics and Automation Letters 11(1): 738--745

    2025

  76. [76]

    arXiv preprint arXiv:2304.04362

    Tian Y, Chang Y, Quang L, Schang A, Nieto-Granda C, How JP, Carlone L (2023) Resilient and Distributed Multi-Robot Visual SLAM: Datasets, Experiments, and Lessons Learned. arXiv preprint arXiv:2304.04362

    arXiv 2023

  77. [77]

    arXiv preprint arXiv:1806.00188

    Tian Y, Khosoussi K, Giamou M, How JP and Kelly J (2018) Near-optimal budgeted data exchange for distributed loop closure detection. arXiv preprint arXiv:1806.00188

    Pith/arXiv arXiv 2018

  78. [78]

    The International Journal of Robotics Research 40(10--11): 1212--1233

    Tian Y, Khosoussi K and How JP (2021) A resource-aware approach to collaborative loop-closure detection with provable performance guarantees. The International Journal of Robotics Research 40(10--11): 1212--1233

    2021

  79. [79]

    IEEE Transactions on Pattern Analysis and Machine Intelligence 13(4): 376--380

    Umeyama S (1991) Least-squares estimation of transformation parameters between two point patterns. IEEE Transactions on Pattern Analysis and Machine Intelligence 13(4): 376--380

    1991

  80. [80]

    In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Uy MA and Lee GH (2018) PointNetVLAD: Deep point cloud based retrieval for large-scale place recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 4470--4479

    2018

Showing first 80 references.