pith. sign in

arxiv: 2606.29851 · v2 · pith:ULX2RKP4new · submitted 2026-06-29 · 💻 cs.RO

TACO: A Test and Check Framework for Robust Pose Graph Optimization

Emilio Olivastri , Alberto Pretto , Tobias Fischer This is my paper

Pith reviewed 2026-07-01 06:50 UTC · model grok-4.3

classification 💻 cs.RO
keywords robust pose graph optimizationoutlier rejectionSLAMincremental consensusswitchable constraintsonline optimizationloop closure verification
0
0 comments X

The pith

TACO approximates the maximally consistent set of measurements in pose graph optimization incrementally by testing loop closures online and sanitizing errors with switchable constraints.

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

The paper presents TACO as a framework to make pose graph optimization robust against outliers caused by incorrect loop closures in SLAM. Rather than labeling measurements as inliers or outliers upfront, it builds an approximation to the largest consistent set through two parts that run together. The test part checks the consistency of each new loop closure as it arrives. The check part then periodically removes any inconsistent measurements that were wrongly accepted. If the approach holds, it delivers offline-level robustness while meeting the speed needs of online mapping.

Core claim

TACO finds an approximation to the maximally consistent set of measurements incrementally through two complementary components: the Incremental Probabilistic Consensus algorithm evaluates the consistency of each incoming loop closure online, and Switchable Outlier Sanitization leverages existing switchable constraints to periodically sanitize any inconsistent measurements that the first component may have included by mistake.

What carries the argument

The dual test-and-check structure of Incremental Probabilistic Consensus for online consistency evaluation paired with Switchable Outlier Sanitization for periodic cleanup of the accepted set.

If this is right

  • Supports online robust pose graph optimization in both 2D and 3D SLAM with success rates above 90 percent in 2D and 83 percent in 3D at outlier rates up to 50 percent.
  • Delivers mean convergence times of roughly 45 ms in 2D and 100 ms in 3D while retaining the efficiency needed for real-time use.
  • Achieves robustness comparable to state-of-the-art offline methods without requiring batch processing of all measurements at once.

Where Pith is reading between the lines

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

  • The incremental test-and-check pattern could be applied to other sequential estimation tasks that receive measurements one at a time and must tolerate occasional bad data.
  • Periodic sanitization might be tuned to different outlier arrival patterns, such as bursts during rapid motion or perceptual aliasing in repetitive environments.
  • Combining the approach with additional geometric checks on loop closures could further reduce the load on the sanitization step.

Load-bearing premise

That the Incremental Probabilistic Consensus algorithm can correctly evaluate the consistency of most incoming loop closures online and that switchable constraints can sanitize any remaining inconsistent measurements without degrading the optimization.

What would settle it

A sequence of loop closures where the Incremental Probabilistic Consensus algorithm accepts a large fraction of inconsistent measurements that the subsequent Switchable Outlier Sanitization step fails to remove, producing visibly degraded trajectory estimates compared with a ground-truth solution.

Figures

Figures reproduced from arXiv: 2606.29851 by Alberto Pretto, Emilio Olivastri, Tobias Fischer.

Figure 1
Figure 1. Figure 1: Overview of the TACO framework used inside a Visual SLAM system. The front-end generates both odometry and loop closures constraints from a sequence of images. IPC tests the consistency of candidate loop closures against their best loop matches. The output of IPC provides an online estimate of the trajectory (bottom right in blue). Each candidate loop closure, classified as an inlier, is then added to a se… view at source ↗
Figure 2
Figure 2. Figure 2: An example of a pose graph with nodes represented by light blue circles. Each error term is associated with the corresponding edge. The coordinate system is usually fixed in the first node x0. Light green and purple ovals represent examples of simple independent subgraphs (i.e., they only include one loop) while the dotted red oval represents more complex independent subgraphs that include both multiple lo… view at source ↗
Figure 3
Figure 3. Figure 3: This figure illustrates the switchable constraint formulation described in (26). The green edge represents the ternary factor involving the variables xa, xb, and φa,b, where φa,b serves as the switch variable, capable of activating or deactivating the constraint. The orange unary edge connected to φa,b represents the prior constraint on the loop closure edge. This expression defines the maximum admissible … view at source ↗
Figure 4
Figure 4. Figure 4: The left plot compares the number of variables of the full problem inG (red) and the trusted subgraph G T (blue) across SOS iterations, while the right plot reports the corresponding average node degree. Results are obtained on the FRH dataset with 50% injected outliers. Algorithm 2: SOS(G, Rc, Ur) // Set of mandatory vertices to visit 1 Vr ← ∅ 2 a = inf, b = 0 3 for ( each ei,j ∈ Ur ) do 4 a = min(a, i) 5… view at source ↗
Figure 6
Figure 6. Figure 6: Image representing two distortion effects that arise from incorrectly integrating outliers into the optimization process. A run is deemed successful if the estimated trajectory achieves an ATE below a defined threshold thAT E, otherwise, it is considered a failure due to trajectory distortion or insufficient drift correction. As a result, the SR reflects the likelihood of a robust optimizer converging to a… view at source ↗
Figure 5
Figure 5. Figure 5: Examples of Absolute Trajectory Error (ATE) and Relative Pose Error (RPE) scores for trajectories under high distortion levels. Although trajectories (a2) and (b2) appear qualitatively superior, the ATE and RPE metrics indicate the opposite. by outlier-induced distortions, complicating the selection and tuning of robust solvers [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: A comparison of the performance of TACO and state-of-the-art methods on 2D SLAM datasets as the outlier percentage varies. The metrics are presented in the following order from left to right: Success Rate, Runtime, and F1 score. Legend of the graphs: TACO( ), IPC( ), HUBER (29)( ), DCS (19)( ), SC (26)( ), MAXMIX (32)( ), RRR (35)( ), GNC (24)( ) [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: A comparison of the performance of TACO and state-of-the-art methods on 3D Visual SLAM datasets as the outlier percentage varies. The metrics are presented in the following order from top to bottom: Success Rate and Mean Convergence Time. Legend of the graphs: TACO( ), HUBER (29)( ), DCS (19)( ), SC (26)( ), MAXMIX (32)( ), RRR (35)( ), GNC (24)( ). 40% and a precision of 99.3% at 10% outliers to an SR of … view at source ↗
Figure 9
Figure 9. Figure 9: Comparison of IPC and TACO performance on 2D SLAM datasets with varying parameter s. Results are averaged across all datasets and outlier percentages. The metrics are presented from left to right: Success Rate, Precision, and Recall. Legend of the graphs: TACO( ), IPC( ). 10 20 30 40 50 Parameter M 0.0 0.2 0.4 0.6 0.8 1.0 Success Rate 10 20 30 40 50 Parameter M 0.040 0.042 0.044 0.046 0.048 0.050 Mean Conv… view at source ↗
Figure 10
Figure 10. Figure 10: Evaluation of TACO ( ) performance on 2D SLAM datasets with varying parameter M. Results are averaged across all datasets and outlier percentages. The metrics, presented from left to right, are Success Rate, Mean Convergence Time, Mean Recovery Time, and Recovery Operations. achieving values exceeding 90%, as well as by TACO, which consistently maintains SR above 83%. RRR and MAXMIX exhibit decreasing tre… view at source ↗
read the original abstract

Pose Graph Optimization (PGO) is one of the most widely adopted approaches for solving Simultaneous Localization and Mapping (SLAM) problems. However, PGO approaches are particularly sensitive to outliers, which can substantially degrade the quality of the estimated trajectories. These outliers arise from incorrect place recognition associations caused by perceptual aliasing in the environment. In this paper, we present TACO (short for Test And Check Optimization), a robust optimization framework designed to filter out outliers from PGO systems. Rather than explicitly modeling measurements as inliers or outliers, TACO finds an approximation to the maximally consistent set of measurements incrementally through two complementary components: (i) The test component, namely the Incremental Probabilistic Consensus (IPC) algorithm, evaluates the consistency of each incoming loop closure online. (ii) The check component dubbed Switchable Outlier Sanitization leverages the existing Switchable Constraints to periodically sanitize any inconsistent measurements from the consistent set that IPC may have mistakenly included. We evaluate TACO on 2D SLAM and 3D Visual SLAM datasets against several state-of-the-art methods. The results show robustness comparable to state-of-the-art offline methods while preserving the computational efficiency required for online deployment, achieving a success rate above 90% in 2D and 83% in 3D across outlier rates up to 50%, with mean convergence times of approximately 45 ms and 100 ms, respectively. We release an open-source implementation of our method with this paper.

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

3 major / 2 minor

Summary. The paper presents TACO, a robust PGO framework for SLAM that approximates the maximally consistent measurement set incrementally via two components: the Incremental Probabilistic Consensus (IPC) algorithm, which evaluates incoming loop-closure consistency online, and Switchable Outlier Sanitization, which periodically removes inconsistencies that IPC may have accepted. Experiments on 2D and 3D SLAM datasets report success rates above 90% (2D) and 83% (3D) at up to 50% outlier rates, with mean times of ~45 ms and ~100 ms, and an open-source implementation is released.

Significance. If validated, the test-and-check design offers a practical online alternative to explicit outlier modeling in PGO by leveraging existing switchable constraints, preserving efficiency while achieving robustness comparable to offline methods. The open-source release supports reproducibility.

major comments (3)
  1. [§3] §3 (IPC description): the claim that IPC correctly classifies most incoming loop closures online is load-bearing for the central incremental claim, yet no analysis is given of how an early false-positive acceptance shifts the pose-graph estimate and thereby corrupts subsequent consistency tests; the skeptic concern about a feedback loop is not addressed.
  2. [§4] §4 (Switchable Outlier Sanitization): the assertion that periodic sanitization corrects IPC mistakes is central to robustness, but the manuscript provides no convergence-rate analysis, required cycle count, or conditions under which the switch variables isolate remaining inconsistencies without destabilizing the incremental solver.
  3. [Experiments] Experiments (2D/3D results tables): success rates above 90%/83% are reported across outlier rates, but the text supplies no details on algorithm implementation, data handling, or error analysis, preventing verification that the data support the robustness claims.
minor comments (2)
  1. [Abstract] Abstract: the phrase 'mean convergence times' is used without defining what quantity is converging (e.g., solver iterations or full trajectory optimization).
  2. [§3] Notation: the probabilistic criterion inside IPC could be stated more explicitly with an equation to aid readers who have not studied the base IPC reference.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major point below and will revise the paper to incorporate additional analysis and details as outlined.

read point-by-point responses

  1. Referee: [§3] §3 (IPC description): the claim that IPC correctly classifies most incoming loop closures online is load-bearing for the central incremental claim, yet no analysis is given of how an early false-positive acceptance shifts the pose-graph estimate and thereby corrupts subsequent consistency tests; the skeptic concern about a feedback loop is not addressed.

    Authors: We agree that a dedicated analysis of how early false-positive acceptances in IPC could propagate errors and affect subsequent tests would strengthen the incremental claim. The manuscript's empirical results across multiple datasets with up to 50% outliers indicate that the overall framework remains robust, but this does not directly analyze the feedback-loop concern. We will add a new subsection to §3 providing a qualitative discussion of error propagation, supported by additional controlled experiments on synthetic graphs that isolate the effect of early misclassifications. revision: yes

  2. Referee: [§4] §4 (Switchable Outlier Sanitization): the assertion that periodic sanitization corrects IPC mistakes is central to robustness, but the manuscript provides no convergence-rate analysis, required cycle count, or conditions under which the switch variables isolate remaining inconsistencies without destabilizing the incremental solver.

    Authors: We acknowledge that the paper lacks a formal convergence-rate analysis or explicit conditions for the sanitization step. The current presentation relies on empirical evidence that periodic sanitization improves robustness. In revision we will expand §4 with observed cycle counts from the experiments, a brief discussion of stability conditions drawn from the switchable-constraints literature, and a note on when the periodic check is triggered to avoid destabilization of the incremental solver. revision: yes

  3. Referee: [Experiments] Experiments (2D/3D results tables): success rates above 90%/83% are reported across outlier rates, but the text supplies no details on algorithm implementation, data handling, or error analysis, preventing verification that the data support the robustness claims.

    Authors: We agree that additional implementation and data-handling details are required for verification. We will expand the Experiments section to describe the exact parameter settings for IPC and sanitization, the method used to inject outliers, the number of independent runs, and basic error statistics (means and standard deviations). The released open-source code already contains the full implementation; the revised text will reference the corresponding code sections. revision: yes

Circularity Check

0 steps flagged

No circularity: TACO is an algorithmic composition of prior components with empirical validation

full rationale

The paper describes TACO as an incremental framework that applies the existing Incremental Probabilistic Consensus (IPC) algorithm for online consistency testing and Switchable Constraints for periodic sanitization. No derivation chain, first-principles result, or prediction is claimed that reduces by construction to fitted parameters or self-referential definitions. The abstract and method explicitly build on established techniques without redefining them in terms of the new outputs. Evaluation consists of empirical success rates on external datasets rather than any closed-loop mathematical equivalence. No load-bearing self-citation or ansatz smuggling is present in the provided text.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

With only the abstract available, specific free parameters, axioms, and invented entities cannot be identified from the text. The method appears to build on existing switchable constraints from prior literature without introducing new entities.

pith-pipeline@v0.9.1-grok · 5796 in / 1280 out tokens · 44948 ms · 2026-07-01T06:50:13.887876+00:00 · methodology

Review history (2 revisions) →

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

55 extracted references · 10 canonical work pages

  1. [1]

    Past, present, and future of simultaneous localization and mapping: Towards the robust- perception age.IEEE Transactions on Robotics2016; 32(6): 1309–1332

    Cadena C, Carlone L, Carrillo H et al. Past, present, and future of simultaneous localization and mapping: Towards the robust- perception age.IEEE Transactions on Robotics2016; 32(6): 1309–1332

  2. [2]

    A tutorial on graph- based slam.IEEE Intelligent Transportation Systems Magazine 2010; 2(4): 31–43

    Grisetti G, K¨ummerle R, Stachniss C et al. A tutorial on graph- based slam.IEEE Intelligent Transportation Systems Magazine 2010; 2(4): 31–43. DOI:10.1109/MITS.2010.939925

    work page doi:10.1109/mits.2010.939925 2010

  3. [3]

    Visual odometry [tutorial]

    Scaramuzza D and Fraundorfer F. Visual odometry [tutorial]. IEEE Robotics and Automation Magazine2011; 18(4): 80–92. DOI:10.1109/MRA.2011.943233

    work page doi:10.1109/mra.2011.943233 2011

  4. [4]

    Lidar odometry survey: recent advancements and remaining challenges.Intelligent Service Robotics2024; 17(2): 95–118

    Lee D, Jung M, Yang W et al. Lidar odometry survey: recent advancements and remaining challenges.Intelligent Service Robotics2024; 17(2): 95–118

  5. [5]

    Visual place recognition: A tutorial [tutorial].IEEE Robotics and Automation Magazine 2024; 31(3): 139–153

    Schubert S, Neubert P, Garg S et al. Visual place recognition: A tutorial [tutorial].IEEE Robotics and Automation Magazine 2024; 31(3): 139–153. DOI:10.1109/MRA.2023.3310859

    work page doi:10.1109/mra.2023.3310859 2024

  6. [6]

    3d point cloud-based place recognition: a survey.Artificial Intelligence Review2024; 57(4): 83

    Luo K, Yu H, Chen X et al. 3d point cloud-based place recognition: a survey.Artificial Intelligence Review2024; 57(4): 83

  7. [7]

    g2o: A general framework for (hyper) graph optimization

    Grisetti G, K ¨ummerle R, Strasdat H et al. g2o: A general framework for (hyper) graph optimization. InIEEE International Conference on Robotics and Automation. pp. 9– 13

  8. [8]

    borglab/gtsam, 2022

    Dellaert F and Contributors G. borglab/gtsam, 2022. DOI: 10.5281/zenodo.5794541. URL https://github.com/ borglab/gtsam

    work page doi:10.5281/zenodo.5794541 2022

  9. [9]

    Ceres Solver,

    Agarwal S, Mierle K and Team TCS. Ceres Solver,

  10. [10]

    URL https://github.com/ceres-solver/ ceres-solver

  11. [11]

    A gauss—newton method for convex composite optimization.Mathematical Programming1995; 71(2): 179–194

    Burke JV and Ferris MC. A gauss—newton method for convex composite optimization.Mathematical Programming1995; 71(2): 179–194

  12. [12]

    The levenberg-marquardt algorithm: implementation and theory

    Mor´e JJ. The levenberg-marquardt algorithm: implementation and theory. InNumerical analysis: proceedings of the biennial Conference held at Dundee, June 28–July 1, 1977. pp. 105– 116

    1977

  13. [13]

    Springer, 1999

    Nocedal J and Wright SJ.Numerical optimization. Springer, 1999

    1999

  14. [14]

    Se-sync: A certifiably correct algorithm for synchronization over the special euclidean group.The International Journal of Robotics Research2019; 38(2-3): 95–125

    Rosen DM, Carlone L, Bandeira AS et al. Se-sync: A certifiably correct algorithm for synchronization over the special euclidean group.The International Journal of Robotics Research2019; 38(2-3): 95–125

  15. [15]

    Lsd-slam: Large-scale direct monocular slam

    Engel J, Sch ¨ops T and Cremers D. Lsd-slam: Large-scale direct monocular slam. In Fleet D, Pajdla T, Schiele B et al. (eds.)European Conference on Computer Vision. ISBN 978-3- 319-10605-2, pp. 834–849

  16. [16]

    Orb-slam3: An accurate open-source library for visual, visual–inertial, and multimap slam.IEEE Transactions on Robotics2021; 37(6): 1874–1890

    Campos C, Elvira R, Rodr ´ıguez JJG et al. Orb-slam3: An accurate open-source library for visual, visual–inertial, and multimap slam.IEEE Transactions on Robotics2021; 37(6): 1874–1890. DOI:10.1109/TRO.2021.3075644

    work page doi:10.1109/tro.2021.3075644 2021

  17. [17]

    Ldso: Direct sparse odometry with loop closure

    Gao X, Wang R, Demmel N et al. Ldso: Direct sparse odometry with loop closure. InIEEE/RSJ International Conference on Intelligent Robots and Systems. pp. 2198–2204. DOI: 10.1109/IROS.2018.8593376

    work page doi:10.1109/iros.2018.8593376 2018

  18. [18]

    Bimanual robot-assisted dressing: A spherical coordinate-based strategy for tight-fitting garments

    Guadagnino T, Mersch B, Gupta S et al. KISS-SLAM: A Simple, Robust, and Accurate 3D LiDAR SLAM System With Enhanced Generalization Capabilities. InIEEE/RSJ International Conference on Intelligent Robots and Systems. pp. 5363–5370. DOI:10.1109/IROS60139.2025.11246613

    work page doi:10.1109/iros60139.2025.11246613 2025

  19. [19]

    A general and adaptive robust loss function

    Barron JT. A general and adaptive robust loss function. InIEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 4331–4339

  20. [20]

    Robust map optimization using dynamic covariance scaling

    Agarwal P, Tipaldi GD, Spinello L et al. Robust map optimization using dynamic covariance scaling. InIEEE International Conference on Robotics and Automation. pp. 62–69

  21. [21]

    Randomized ransac with td, d test.Image and vision computing2004; 22(10): 837–842

    Matas J and Chum O. Randomized ransac with td, d test.Image and vision computing2004; 22(10): 837–842

  22. [22]

    Pairwise consistent measurement set maximization for robust multi- robot map merging

    Mangelson JG, Dominic D, Eustice RM et al. Pairwise consistent measurement set maximization for robust multi- robot map merging. InIEEE International Conference on Robotics and Automation. pp. 2916–2923

  23. [23]

    Clipper: A graph-theoretic framework for robust data association

    Lusk PC, Fathian K and How JP. Clipper: A graph-theoretic framework for robust data association. InIEEE International Conference on Robotics and Automation. pp. 13828–13834

  24. [24]

    Data association in stochastic mapping using the joint compatibility test.IEEE Transactions on Robotics and Automation2001; 17(6): 890–897

    Neira J and Tard´os JD. Data association in stochastic mapping using the joint compatibility test.IEEE Transactions on Robotics and Automation2001; 17(6): 890–897

  25. [25]

    Graduated non- convexity for robust spatial perception: From non-minimal solvers to global outlier rejection.IEEE Robotics and Automation Letters2020; 5(2): 1127–1134

    Yang H, Antonante P, Tzoumas V et al. Graduated non- convexity for robust spatial perception: From non-minimal solvers to global outlier rejection.IEEE Robotics and Automation Letters2020; 5(2): 1127–1134

  26. [26]

    Robust incremental smoothing and mapping (risam)

    McGann D, Rogers JG and Kaess M. Robust incremental smoothing and mapping (risam). InIEEE International Conference on Robotics and Automation. pp. 4157–4163. Prepared usingsagej.cls Olivastri et al. 15

  27. [27]

    Switchable constraints for robust pose graph slam

    S¨underhauf N and Protzel P. Switchable constraints for robust pose graph slam. InIEEE/RSJ International Conference on Intelligent Robots and Systems. pp. 1879–1884

  28. [28]

    IPC: Incremental probabilistic consensus-based consistent set maximization for SLAM backends

    Olivastri E and Pretto A. IPC: Incremental probabilistic consensus-based consistent set maximization for SLAM backends. InIEEE International Conference on Robotics and Automation. pp. 10283–10289. DOI:10.1109/ICRA57147. 2024.10611214

    work page doi:10.1109/icra57147 2024

  29. [29]

    Cambridge University Press, 2025

    Carlone L, Kim A, Barfoot T et al.SLAM handbook: From localization and mapping to spatial intelligence. Cambridge University Press, 2025

    2025

  30. [30]

    On the unification of line processes, outlier rejection, and robust statistics with applications in early vision.International Journal of Computer Vision1996; 19(1): 57–91

    Black MJ and Rangarajan A. On the unification of line processes, outlier rejection, and robust statistics with applications in early vision.International Journal of Computer Vision1996; 19(1): 57–91

  31. [31]

    Aeros: Adaptive robust least-squares for graph-based slam.Frontiers in Robotics and AI2022; 9: 789444

    Ramezani M, Mattamala M and Fallon M. Aeros: Adaptive robust least-squares for graph-based slam.Frontiers in Robotics and AI2022; 9: 789444

  32. [32]

    isam2: Incremental smoothing and mapping using the bayes tree.The International Journal of Robotics Research2012; 31(2): 216–235

    Kaess M, Johannsson H, Roberts R et al. isam2: Incremental smoothing and mapping using the bayes tree.The International Journal of Robotics Research2012; 31(2): 216–235

  33. [33]

    Inference on networks of mixtures for robust robot mapping.The International Journal of Robotics Research2013; 32(7): 826–840

    Olson E and Agarwal P. Inference on networks of mixtures for robust robot mapping.The International Journal of Robotics Research2013; 32(7): 826–840

  34. [34]

    Selecting good measurements via l1 relaxation: A convex approach for robust estimation over graphs

    Carlone L, Censi A and Dellaert F. Selecting good measurements via l1 relaxation: A convex approach for robust estimation over graphs. InIEEE/RSJ International Conference on Intelligent Robots and Systems. pp. 2667–2674

  35. [35]

    Robin: a graph-theoretic approach to reject outliers in robust estimation using invariants

    Shi J, Yang H and Carlone L. Robin: a graph-theoretic approach to reject outliers in robust estimation using invariants. InIEEE International Conference on Robotics and Automation. pp. 13820–13827

  36. [36]

    Robust loop closing over time for pose graph slam.The International Journal of Robotics Research2013; 32(14): 1611–1626

    Latif Y , Cadena C and Neira J. Robust loop closing over time for pose graph slam.The International Journal of Robotics Research2013; 32(14): 1611–1626

  37. [37]

    Cluster-based penalty scaling for robust pose graph optimization.IEEE Robotics and Automation Letters2020; 5(4): 6193–6200

    Wu F and Beltrame G. Cluster-based penalty scaling for robust pose graph optimization.IEEE Robotics and Automation Letters2020; 5(4): 6193–6200

  38. [38]

    Loam: Lidar odometry and mapping in real-time

    Zhang J, Singh S et al. Loam: Lidar odometry and mapping in real-time. InRobotics: Science and systems, volume 2. Berkeley, CA, pp. 1–9

  39. [39]

    Stereo dso: Large-scale direct sparse visual odometry with stereo cameras

    Wang R, Schworer M and Cremers D. Stereo dso: Large-scale direct sparse visual odometry with stereo cameras. InIEEE International Conference on Computer Vision. pp. 3903–3911

  40. [40]

    Bags of binary words for fast place recognition in image sequences.IEEE Transactions on Robotics2012; 28(5): 1188–1197

    Galvez-L´opez D and Tardos JD. Bags of binary words for fast place recognition in image sequences.IEEE Transactions on Robotics2012; 28(5): 1188–1197. DOI:10.1109/TRO.2012. 2197158

    work page doi:10.1109/tro.2012 2012

  41. [41]

    Kiss-matcher: Fast and robust point cloud registration revisited

    Lim H, Kim D, Shin G et al. Kiss-matcher: Fast and robust point cloud registration revisited. InIEEE International Conference on Robotics and Automation. pp. 11104–11111

  42. [42]

    Visual place recognition: A tutorial [tutorial].IEEE Robotics & Automation Magazine 2023; 31(3): 139–153

    Schubert S, Neubert P, Garg S et al. Visual place recognition: A tutorial [tutorial].IEEE Robotics & Automation Magazine 2023; 31(3): 139–153

    2023

  43. [43]

    Robust optimization of factor graphs by using condensed measurements

    Grisetti G, K ¨ummerle R and Ni K. Robust optimization of factor graphs by using condensed measurements. InIEEE/RSJ International Conference on Intelligent Robots and Systems. pp. 581–588

  44. [44]

    Hipe: Hierarchical initialization for pose graphs.IEEE Robotics and Automation Letters2021; 7(1): 287–294

    Guadagnino T, Di Giammarino L and Grisetti G. Hipe: Hierarchical initialization for pose graphs.IEEE Robotics and Automation Letters2021; 7(1): 287–294

  45. [45]

    The steiner cycle polytope.European Journal of Operational Research2003; 147(3): 671–679

    Salazar-Gonz´alez JJ. The steiner cycle polytope.European Journal of Operational Research2003; 147(3): 671–679

  46. [46]

    A fast and accurate approximation for planar pose graph optimization

    Carlone L, Aragues R, Castellanos JA et al. A fast and accurate approximation for planar pose graph optimization. The International Journal of Robotics Research2014; 33(7): 965–987

  47. [47]

    A note on two problems in connexion with graphs

    Dijkstra EW. A note on two problems in connexion with graphs. Numerische Mathematik1959; : 269–271

  48. [48]

    Evaluating the performance of map optimization algorithms

    Olson E and Kaess M. Evaluating the performance of map optimization algorithms. InRSS Workshop on Good Experimental Methodology in Robotics, volume 15. p. 35

  49. [49]

    Reliable graphs for slam.The International Journal of Robotics Research2019; 38(2-3): 260–298

    Khosoussi K, Giamou M, Sukhatme GS et al. Reliable graphs for slam.The International Journal of Robotics Research2019; 38(2-3): 260–298

  50. [50]

    A tutorial on quantitative trajectory evaluation for visual (-inertial) odometry

    Zhang Z and Scaramuzza D. A tutorial on quantitative trajectory evaluation for visual (-inertial) odometry. In IEEE/RSJ International Conference on Intelligent Robots and Systems. pp. 7244–7251

  51. [51]

    From angular manifolds to the integer lattice: Guaranteed orientation estimation with application to pose graph optimization.IEEE Transactions on Robotics2014; 30(2): 475–492

    Carlone L and Censi A. From angular manifolds to the integer lattice: Guaranteed orientation estimation with application to pose graph optimization.IEEE Transactions on Robotics2014; 30(2): 475–492

  52. [52]

    Are we ready for autonomous driving? the kitti vision benchmark suite

    Geiger A, Lenz P and Urtasun R. Are we ready for autonomous driving? the kitti vision benchmark suite. InIEEE Conference on Computer Vision and Pattern Recognition. pp. 3354–3361

  53. [53]

    A benchmark for the evaluation of rgb-d slam systems

    Sturm J, Engelhard N, Endres F et al. A benchmark for the evaluation of rgb-d slam systems. InIEEE/RSJ International Conference on Intelligent Robots and Systems. pp. 573–580

  54. [54]

    Vertigo (versatile extensions for robust inference using graph optimization) source code,

    Suenderhauf N and Protzel P. Vertigo (versatile extensions for robust inference using graph optimization) source code,

  55. [55]

    URL https:// openslam-org.github.io/vertigo.html

    DOI:10.25912/5dcb33f1400ec. URL https:// openslam-org.github.io/vertigo.html. Prepared usingsagej.cls

    work page doi:10.25912/5dcb33f1400ec