Pith. sign in

REVIEW 3 major objections 5 minor 77 references

Fault-Tolerant Multi-Modal Localization of Multi-Robots on Matrix Lie Groups

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A Lie-group EKF with newly defined stochastic operations fuses IMU, wheel odometry, and fiducial-marker pseudo-poses, keeping multi-robot teams localized where plain IMU/odometry fusion diverges.

desk verdict A plausible multi-robot Lie-group localization framework, but the manuscript has a load-bearing dimension mismatch and an invalid averaging identity; it needs major revision before the results can be trusted. read the letter →

arxiv 2505.00842 v1 pith:YYAAYA4R submitted 2025-05-01 cs.RO cs.SYeess.SYmath.GR

Mahboubeh Zarei , Robin Chhabra This is my paper · ORCID
classification cs.ROcs.SYeess.SYmath.GR MSC 93E1168T40
keywords multi-robotlocalizationmatrixLiegroupsextendedKalmanfilterstochasticoperationsonfaultdetectionfiducialmarkersmulti-modalsensorfusioncollaborative
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 proposes a fault-tolerant, multi-modal localization framework for teams of wheeled robots, built on new stochastic operations for matrix Lie groups. Its central claim is that by representing each robot's pose, velocity, and IMU biases as a single element of the extended-pose group $\mathbb{SE}_2(3)\times\mathbb{R}^3\times\mathbb{R}^3$, and by constructing pseudo-pose measurements from fiducial markers through Lie-group composition, one can fuse IMU, wheel odometry, and relative-pose data in a single EKF that scales across a leader-follower network. A squared-Mahalanobis-distance fault-detection gate admits only trustworthy marker updates. The paper reports real-time experiments in which this IMU/WO/ArUco/FD filter recovers ground-truth trajectories where IMU-only or IMU-plus-odometry fusion diverges, especially during turns, sharp maneuvers, and uneven terrain.

What carries the argument

The central object is the matrix Lie group $G_X=\mathbb{SE}_2(3)\times\mathbb{R}^3\times\mathbb{R}^3$ for prediction (an extended pose group whose elements carry rotation, position, velocity, and IMU bias terms) together with the measurement group $G_Z=\mathbb{SE}(3)\times\mathbb{R}^3$. The argument is carried by newly defined stochastic operations: composition, inverse, difference, averaging, and fusion, in which Lie-algebra uncertainties $\zeta\sim\mathcal{N}(0,P)$ are mapped through the Adjoint $\mathrm{Ad}_X$ and the right Jacobian $J_r$, so covariance propagation is done entirely on the Lie algebra. A squared Mahalanobis distance $D^2(k)$ on the innovation gates whether a pseudo-pose update is admitted.

What would settle it

Compute the intrinsic dimension of $G_X=\mathbb{SE}_2(3)\times\mathbb{R}^3\times\mathbb{R}^3$ and compare it with the dimension of $Q$ in Eq. (52): if the filter uses the 18-dimensional matrix in the 15-dimensional covariance update, the predicted covariance is mathematically undefined on the stated state manifold, so the method cannot be implemented exactly as written. A separate behavioral falsifier: drive a follower through a sharp turn while its marker is fully occluded for a known interval; if the gated filter's error is not significantly smaller than the ungated filter's error, the fault-detection claim fails.

Watch

Extended reading notes

Core claim

On its own terms, the discovery is that every operation a filter needs—composition, differencing, inversion, averaging, and fusion—can be defined for correlated and uncorrelated stochastic members of a matrix Lie group by pushing Gaussian uncertainties through the Adjoint map and the right Jacobian, and that these operations make it possible to synthesize pseudo-pose measurements that propagate through a multi-robot network. The prediction step runs on $G_X=\mathbb{SE}_2(3)\times\mathbb{R}^3\times\mathbb{R}^3$, where each robot carries rotation, position, velocity, and IMU biases; the update consumes body velocity plus pseudo-pose on $G_Z=\mathbb{SE}(3)\times\mathbb{R}^3$. The paper claims that this combination runs in real time, remains accurate as leader, follower type 1, and follower type 2 estimates are cascaded, and recovers ground-truth trajectories in cases where the IMU/wheel-odometry filter alone fails.

Load-bearing premise

The argument assumes the filter equations are dimensionally consistent on the stated Lie groups: the prediction group $\mathbb{SE}_2(3)\times\mathbb{R}^3\times\mathbb{R}^3$ has 15 degrees of freedom, while the noise covariance $Q$, the adjoint, and the right Jacobian used in prediction are written as $18\times18$, and the measurement group $\mathbb{SE}(3)\times\mathbb{R}^3$ has 9 degrees of freedom while the measurement noise matrices are $12\times12$; without an unstated correction, the equations are not well-defined as printed.

Editorial extensions

If this is right

  • A team of robots can maintain globally consistent localization with only one stationary fiducial marker for the leader while followers observe mobile markers on their neighbors, with no GPS required.
  • A robot whose proprioceptive IMU/odometry filter has begun to diverge can be pulled back to the ground-truth trajectory whenever a chain of trusted relative-pose measurements reaches it.
  • Because the stochastic operations are defined on an arbitrary matrix Lie group, the same EKF machinery transfers to other product-manifold state spaces and other relative-pose sensors.
  • Fusing two pseudo-pose measurements at a follower yields an estimate that the paper reports as better than either individual estimate and better than the IMU/odometry filter.

Reading between the lines

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

  • The same stochastic operations should apply to cooperative localization with UWB, LiDAR, or camera-relative measurements, since the update only needs a relative-pose or relative-velocity model on a Lie group; the paper demonstrates only fiducial markers.
  • A controlled fault-injection study—occluding a follower's marker for known intervals during a sharp turn—would let practitioners measure the gate's detection latency and false-alarm rate, quantities the paper does not report.
  • The Lie-group fusion rule may also serve object-level multi-robot SLAM, where landmarks and robots both carry SE(3) estimates with cross-correlations, because the fusion operation is designed for correlated pose estimates.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. This paper develops stochastic operations on matrix Lie groups—composition, inversion, difference, averaging, fusion, and constrained fusion—for correlated and uncorrelated estimates, and embeds them in a multi-robot EKF localization framework. The prediction step runs on G_X = SE_2(3) × R^3 × R^3 using IMU measurements; updates fuse wheel-odometry velocity and pseudo-pose measurements constructed from ArUco markers on G_Z = SE(3) × R^3. A Mahalanobis-distance fault detection module gates pseudo-pose updates for leader, follower type 1, and follower type 2 robots. The paper reports experiments with three AgileX LIMO robots and a PhaseSpace motion tracker, claiming real-time, scalable, fault-tolerant localization and showing that IMU/WO fusion failures are recovered by ArUco/FD updates.

Significance. If the proposed operations and filter are correct, the paper offers a unified Lie-group treatment of multi-robot sensor fusion: it generalizes prior pose-only operations to correlated multi-estimate composition and fusion, and it demonstrates the framework on physical hardware with three robot types rather than in simulation. The fault detection module addresses a practical issue in marker-based updates, and the experiments use a motion-capture ground truth. However, part of the theoretical contribution is located in the authors' prior paper [44], and the dimension inconsistencies described below must be resolved before the reported experiments can be reliably tied to the stated mathematics.

major comments (3)
  1. [§IV-A, Appendix A, §V-D1 (Eqs. 50–53, 62)] The prediction filter is not well-defined on the stated 15-dimensional state group. Section IV-A defines G_X := SE_2(3) × R^3 × R^3, which has dimension 15 (9 for pose/velocity plus 6 for biases), but the prediction uses f ∈ R^18 in Eq. (51), w ∼ N(0_18, Q) with Q ∈ R^{18×18} in Eq. (52), and Eq. (53) propagates this 18-dimensional noise. Appendix A derives Ad_X and J_r for the 18-dimensional group G = SE_2(3) × SE_2(3) and states that G_X is the special case where the second block's rotation equals I_3; the explicit X in Eq. (50) indeed fixes that rotation. The paper never states that the three covariance directions corresponding to that fixed rotation are zero. The experiments set P(0|0) = 10^{-2} I_18 in Section V-D1, which assigns nonzero uncertainty in directions that leave G_X. As written, Eqs. (13)–(18) are therefore not defined on the claimed manifold, and the trajectories cannot be reproduced from the stated theory. The authors should either define the filter on the full 18-dimensional group G, or explicitly state the 15-dimensional parameterization with zero-noise/zero-covariance constraints on the fixed rotation and correct Eq. (53) and P(0|0) accordingly.
  2. [§IV-C1–C3 (measurement model and R_l, R_f1, R_f2)] The measurement update has the same dimension inconsistency. The text states G_Z := SE(3) × R^3, which is 9-dimensional, but defines R_l = diag([R_pl, ε I_3, R_vl]) ∈ R^{12×12}, with analogous R_f1 and R_f2. The ε I_3 block corresponds to the rotation coordinate of the second block, which is fixed to I_3 in the measurement members Z_l, Z_f1, and Z_f2. The general EKF in Section II-C is written for a q-dimensional group, so q is inconsistent between the stated manifold and the matrices. If the measurement model is intended to live in SE(3) × SE(3) (12-dimensional), the text should say so and remove the placeholder ε block; if it is intended to be SE(3) × R^3, the innovation and noise must be 9-dimensional with the fixed rotation component removed. The authors need to state which objects are correct and adjust the equations and experiments accordingly.
  3. [§III-D, §III-E, §V-D4] The averaging and fusion operations are not self-contained, despite being central to the paper's claimed contribution (i) and to the fusion experiments in Section V-D4. The averaging operation is introduced via X_i^{α_i} = \bar{X}_i^{α_i} exp(α_i [ζ_i]^∧) without proof; fractional powers of a noisy Lie-group element are not canonical, so this equality needs a precise definition (e.g., through the exponential of α_i times the logarithm) and a consistency argument before the covariance formula (33) is valid. The fusion operation's closed-form solution, Eqs. (35)–(38), is taken directly from the authors' prior paper [44] with only the final result reproduced, and the cross-covariance recursion is also quoted from [44] without derivation. Since the overall filter consistency and the fusion experiments depend on these steps, the manuscript should provide self-contained proofs or clearly state which parts of the claimed novelty are delegated to [44].
minor comments (5)
  1. [§I-B] The contribution list jumps from (ii) to (iv); the missing item (iii) should be restored or the list renumbered.
  2. [§IV-A, Eq. (52)] Equation (52) first writes w with −w_g and −w_a and then states an equivalence to +w_g and +w_a; because these noises are zero-mean, the distributions coincide, but the double definition is confusing and should be clarified.
  3. [Appendix A] The appendix says the proof for G_Z := SE(3) × R^3 follows similarly and is left to the reader; given that the dimension of this group is central to the issues above, the explicit group law, logarithm, and Jacobians for G_Z should be included.
  4. [§V-D1, Fig. 9] Figure 9 is described qualitatively as showing better or worse localization for different P(0|0), but no error metric or axis definition is given; please state what quantity is plotted so the claim that the proposed method is insensitive to P_0 can be checked.
  5. [General notation] The group SE_2(3) is used both as the 9-dimensional pose-velocity group and as a factor of the block-diagonal group G_X; the embedding used in Eq. (50) should be defined once and used consistently throughout.

Circularity Check

1 steps flagged · score 4.0 of 10

Partial self-citation dependency: the fusion operation and cross-covariance recursion are imported from the authors' prior paper [44] without derivation, while the main EKF localization chain is independently specified.

  1. self citation load bearing [Section III-E, Eqs. (34)-(38) and the cross-covariance recursion after Eq. (38); reference [44]]
    "The optimal fused member can be obtained from the following minimization problem [44]: ... The closed-form solution to this minimization problem is presented in [44]. For brevity, we only present the final derivation, as follows:"

    The paper lists fusion among its newly proposed stochastic operations, but the closed-form fused mean/covariance in Eqs. (35)-(38) and the cross-covariance recursion for correlated estimates are not derived or proved in this manuscript; they are attributed only to [44], whose authors are the same two researchers. The claimed 'novel' fusion operation therefore reduces to a self-citation for its mathematical content. The main localization filter, however, is built on independently stated Lie-group EKF equations and sensor models, so this is a partial load-bearing dependency rather than a fully circular derivation.

full rationale

The central localization derivation is not circular: the inertial prediction, velocity update, and pseudo-pose pose updates use standard Lie-group EKF equations (13)-(21) and physically stated sensor models (43)-(54), and the experimental claims are checked against external PhaseSpace ground truth. The only real circularity-adjacent step is the fusion operation and correlated-estimate cross-covariance recursion, which are taken from the authors' prior work [44] without derivation; this makes one claimed contribution depend on a self-citation while the fault-tolerant multi-robot localization results stand on independent content. The paper also has a consistency defect, not a circularity: the state manifold GX = SE2(3) × R3 × R3 is 15-dimensional, yet Eqs. (51)-(53), Q, Appendix A, and P(0|0) = 10^-2 I18 use 18-dimensional objects, and the measurement group SE(3) × R3 is 9-dimensional while R_l is 12-dimensional; this makes the filter as written ill-defined on the claimed manifolds, but it does not make any predicted quantity equal to a fitted input or to the paper's assumptions by construction.

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

The filter rests on first-order BCH truncation, Gaussian noise assumptions, an unproved fractional-power identity for averaging, and a state dimension mismatch between the group definitions and the filter equations. The fault detection threshold and noise covariances are user-selected inputs rather than fitted to a benchmark.

free parameters (4)
  • Mahalanobis fault detection threshold = 40
    Section V-D1 sets the threshold to 40 for all experiments; no sensitivity analysis or selection procedure is given, though it directly controls which marker updates are accepted.
  • Small covariance epsilon = epsilon << 1, unspecified
    Introduced in Section IV-C1 (Rl = diag([Rpl, epsilon I3, Rvl])) to avoid inverting a zero block; the value is not specified and affects the conditioning of the measurement update.
  • Sensor noise covariances = Qg=Qbg=1e-6 I3, Qa=Qba=I3, Rvl=1.5e-4 I3, Rml=diag([0.0075 I3 rad^2, 0.005 I3 m^2]), Rmf1=diag([0.005 I3 rad^2, 0.0016…
    Chosen values reported in Section V-D1; the filter requires them, but no calibration or inference procedure is described.
  • Initial covariance P(0|0) = 1e-2 I18
    Set in Section V-D1; sensitivity to this value is tested, but the value itself is a user choice.
assumptions (5)
  • standard math BCH formula and its first-order truncation are valid for the covariance propagation steps in Section III.
    Composition, difference, and averaging covariances use the first-order BCH approximation (Appendix C); the approximation error is not quantified, which is normal but unstated in the main text.
  • domain assumption IMU and wheel encoder measurements are corrupted by zero-mean Gaussian white noise and random-walk biases.
    Used in Section IV-A, Eqs. (43)-(49). Reasonable for high-quality sensors but the experiments use a low-cost MPU6050.
  • domain assumption ArUco marker pose measurement errors are zero-mean Gaussian in the Lie algebra.
    Section IV-C assumes Z_ClMl = Zbar_ClMl exp([zeta]^wedge) with zeta ~ N(0,Rml); marker corner detection noise is not exactly Gaussian.
  • ad hoc to paper X_i^{alpha_i} = Xbar_i^{alpha_i} exp(alpha_i [zeta_i]^wedge) for fractional powers on SE2(3).
    Invoked in Section III-D to derive the averaging mean and covariance; this identity is generally false for non-abelian matrix Lie groups, so the averaging formulas are at best an approximation without proof.
  • ad hoc to paper The filter can treat the state and measurement Lie algebras as 18- and 12-dimensional while the stated groups are 15- and 9-dimensional.
    Eq. (51)-(53) and Appendix A use R^18 for f, w, Q, and Ad; Section IV-C1 uses R^12 for measurement covariance. The extra dimensions are never given dynamics or noise, leaving the EKF equations formally undefined on the stated manifolds.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fault-Tolerant Multi-Modal Localization of Multi-Robots on Matrix Lie Groups." pith.science (2026). https://pith.science/paper/YYAAYA4R

@misc{pith2026250500842,
  author       = {Pith},
  title        = {Pith review of: Fault-Tolerant Multi-Modal Localization of Multi-Robots on Matrix Lie Groups},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YYAAYA4R}},
  note         = {Machine review of arXiv:2505.00842}
}
abstract

Consistent localization of cooperative multi-robot systems during navigation presents substantial challenges. This paper proposes a fault-tolerant, multi-modal localization framework for multi-robot systems on matrix Lie groups. We introduce novel stochastic operations to perform composition, differencing, inversion, averaging, and fusion of correlated and non-correlated estimates on Lie groups, enabling pseudo-pose construction for filter updates. The method integrates a combination of proprioceptive and exteroceptive measurements from inertial, velocity, and pose (pseudo-pose) sensors on each robot in an Extended Kalman Filter (EKF) framework. The prediction step is conducted on the Lie group $\mathbb{SE}_2(3) \times \mathbb{R}^3 \times \mathbb{R}^3$, where each robot's pose, velocity, and inertial measurement biases are propagated. The proposed framework uses body velocity, relative pose measurements from fiducial markers, and inter-robot communication to provide scalable EKF update across the network on the Lie group $\mathbb{SE}(3) \times \mathbb{R}^3$. A fault detection module is implemented, allowing the integration of only reliable pseudo-pose measurements from fiducial markers. We demonstrate the effectiveness of the method through experiments with a network of wheeled mobile robots equipped with inertial measurement units, wheel odometry, and ArUco markers. The comparison results highlight the proposed method's real-time performance, superior efficiency, reliability, and scalability in multi-robot localization, making it well-suited for large-scale robotic systems.

Figures

Figures reproduced from arXiv: 2505.00842 by the authors.

Figure 1
Figure 1. Uncertainty on Lie group GX and its Lie algebra gX III. STOCHASTIC OPERATIONS ON LIE GROUPS Here, we develop necessary mechanisms to perform fun￾damental operations on stochastic members of a matrix Lie group. The operations include composition, inverse, difference, averaging, and fusion, which are crucial for the development of our proposed localization algorithm. As shown in [PITH_FULL_IMAGE:figures/full_fig_p004… view at source ↗
Figure 2
Figure 2. Illustration of stochastic operations: (a) inverse, (b) composition, (c) [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. A leader robot’s (a) coordinate frames, and (b) estimation framework. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Follower type 1 robot, (a) coordinate frames, (b) estimation framework. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Follower type 2 robot, (a) coordinate frames, (b) estimation framework. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: The robot platform and experimental setup. [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Effects of lighting conditions, marker size, and robot speed on the [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Leader localization performance; (a) circular motion, (b) cornering [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 10
Figure 10. Figure 10: Performance of the Leader-Follower localization framework. [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: Fig.11. In both cases, we observe that uncertainty propagates [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]
Figure 12
Figure 12. Figure 12: Performance of fusion in the multi-robot localization framework. [PITH_FULL_IMAGE:figures/full_fig_p013_12.png]
Figure 11
Figure 11. Figure 11: Performance of the multi-robot localization framework. [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

77 extracted references · 66 canonical work pages

  1. [44]

    Consistent fusion of correlated pose estimates on matrix lie groups,

    M. Zarei and R. Chhabra, “Consistent fusion of correlated pose estimates on matrix lie groups,” IEEE Robotics and Automation Letters , vol. 9, no. 7, pp. 6584–6591, 2024

    work page 2024

  2. [1]

    Design and motion planning of a two-module collaborative indoor pipeline inspection robot,

    Y .-S. Kwon and B.-J. Yi, “Design and motion planning of a two-module collaborative indoor pipeline inspection robot,” IEEE Transactions on Robotics, vol. 28, no. 3, pp. 681–696, 2012

    work page 2012

  3. [2]

    An exploration-enhanced search algorithm for robot indoor source searching,

    M. Wang, B. Xin, M. Jing, and Y . Qu, “An exploration-enhanced search algorithm for robot indoor source searching,” IEEE Transactions on Robotics, vol. 40, pp. 4160–4178, 2024

    work page 2024

  4. [3]

    Inertial navigation systems for mobile robots,

    B. Barshan and H. F. Durrant-Whyte, “Inertial navigation systems for mobile robots,” IEEE transactions on robotics and automation , vol. 11, no. 3, pp. 328–342, 1995

    1995

  5. [4]

    Measurement and correction of systematic odometry errors in mobile robots,

    J. Borenstein and L. Feng, “Measurement and correction of systematic odometry errors in mobile robots,” IEEE Transactions on robotics and automation, vol. 12, no. 6, pp. 869–880, 1996

    1996

  6. [5]

    Gyrodometry: A new method for combining data from gyros and odometry in mobile robots,

    ——, “Gyrodometry: A new method for combining data from gyros and odometry in mobile robots,” in Proceedings of IEEE International Conference on Robotics and Automation , vol. 1. IEEE, 1996, pp. 423– 428

    work page 1996

  7. [6]

    A dead reckoning localization system for mobile robots using inertial sensors and wheel revolution encoding,

    B.-S. Cho, W.-s. Moon, W.-J. Seo, and K.-R. Baek, “A dead reckoning localization system for mobile robots using inertial sensors and wheel revolution encoding,” Journal of mechanical science and technology , vol. 25, pp. 2907–2917, 2011

    work page 2011

  8. [7]

    An enhanced positioning technique for underground pipeline robot based on inertial sensor/wheel odometer,

    Q. Wang, M. Cai, and Z. Guo, “An enhanced positioning technique for underground pipeline robot based on inertial sensor/wheel odometer,” Measurement, vol. 206, p. 112298, 2023

    work page 2023

Show all 77 references
  1. [8]

    Fusion of optical flow and inertial measurements for robust egomotion estimation. in 2014 ieee,

    M. Bloesch, S. Omari, P. Fankhauser, H. Sommer, C. Gehring, J. Hwangbo, M. A. Hoepflinger, M. Hutter, and R. Siegwart, “Fusion of optical flow and inertial measurements for robust egomotion estimation. in 2014 ieee,” in IEEE/RSJ International Conference on Intelligent Robots a...

    2014

  2. [9]

    Multi- robot relative pose estimation and imu preintegration using passive uwb transceivers,

    M. A. Shalaby, C. C. Cossette, J. Le Ny, and J. R. Forbes, “Multi- robot relative pose estimation and imu preintegration using passive uwb transceivers,” IEEE Transactions on Robotics , vol. 40, pp. 2410–2429, 2024. IEEE TRANSACTIONS ON ROBOTICS, VOL. , NO. , APRIL 2025 14

    2024

  3. [10]

    Edge-assisted multi-robot visual-inertial slam with efficient communication,

    X. Liu, S. Wen, J. Zhao, T. Z. Qiu, and H. Zhang, “Edge-assisted multi-robot visual-inertial slam with efficient communication,” IEEE Transactions on Automation Science and Engineering, vol. 22, pp. 2186– 2198, 2024

    2024

  4. [11]

    Distributed indoor positioning system with inertial mea- surements and map matching,

    A. Perttula, H. Lepp ¨akoski, M. Kirkko-Jaakkola, P. Davidson, J. Collin, and J. Takala, “Distributed indoor positioning system with inertial mea- surements and map matching,” IEEE Transactions on Instrumentation and Measurement, vol. 63, no. 11, pp. 2682–2695, 2014

    2014

  5. [12]

    Cooperative navigation of mavs in gps denied areas,

    R. Sharma and C. Taylor, “Cooperative navigation of mavs in gps denied areas,” in IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems , 2008, pp. 481–486

    2008

  6. [13]

    Vision-based localization for leader–follower formation control,

    G. L. Mariottini, F. Morbidi, D. Prattichizzo, N. Vander Valk, N. Michael, G. Pappas, and K. Daniilidis, “Vision-based localization for leader–follower formation control,” IEEE Transactions on Robotics , vol. 25, no. 6, pp. 1431–1438, 2009

    2009

  7. [14]

    On-board range-based relative localization for micro air vehicles in indoor leader–follower flight,

    S. Van Der Helm, M. Coppola, K. N. McGuire, and G. C. De Croon, “On-board range-based relative localization for micro air vehicles in indoor leader–follower flight,” Autonomous Robots , vol. 44, no. 3, pp. 415–441, 2020

    2020

  8. [15]

    Integrated relative-measurement-based net- work localization and formation maneuver control,

    X. Fang, L. Xie, and X. Li, “Integrated relative-measurement-based net- work localization and formation maneuver control,” IEEE Transactions on Automatic Control , vol. 69, no. 3, pp. 1906–1913, 2023

    1906

  9. [16]

    A kalman filter-based framework for enhanced sensor fusion,

    A. Assa and F. Janabi-Sharifi, “A kalman filter-based framework for enhanced sensor fusion,” IEEE Sensors Journal, vol. 15, no. 6, pp. 3281– 3292, 2015

    2015

  10. [17]

    Distributed decentralized ekf for very large-scale networks with application to satellite mega-constellations navigation,

    L. Pedroso and P. Batista, “Distributed decentralized ekf for very large-scale networks with application to satellite mega-constellations navigation,” Control Engineering Practice , vol. 135, p. 105509, 2023

    2023

  11. [18]

    Groups a path to geometry,

    D. Kenwright and S. Blaylock, “Groups a path to geometry,” University of Cambridge, Computer Laboratory, Tech. Rep., 2013, available at: https://www.cl.cam.ac.uk/

    2013

  12. [19]

    R. P. Paul, Robot manipulators: mathematics, programming, and control: the computer control of robot manipulators . Richard Paul, 1981

    1981

  13. [20]

    Consistent integration and propagation of dis- parate sensor observations,

    H. F. Durrant-Whyte, “Consistent integration and propagation of dis- parate sensor observations,” The International journal of robotics re- search, vol. 6, no. 3, pp. 3–24, 1987

    1987

  14. [21]

    Manipulation and propagation of uncertainty and verification of applicability of actions in assembly tasks,

    S.-F. Su and C. G. Lee, “Manipulation and propagation of uncertainty and verification of applicability of actions in assembly tasks,” IEEE Transactions on Systems, Man, and Cybernetics , vol. 22, no. 6, pp. 1376–1389, 1992

    1992

  15. [22]

    Error propagation on the euclidean group with applications to manipulator kinematics,

    Y . Wang and G. S. Chirikjian, “Error propagation on the euclidean group with applications to manipulator kinematics,” IEEE Transactions on Robotics, vol. 22, no. 4, pp. 591–602, 2006

    2006

  16. [23]

    G. S. Chirikjian, Stochastic models, information theory, and Lie groups, volume 2: Analytic methods and modern applications. Springer Science & Business Media, 2011, vol. 2

    2011

  17. [24]

    Integrating generic sensor fusion algorithms with sound state representations through encap- sulation of manifolds,

    C. Hertzberg, R. Wagner, U. Frese, and L. Schr ¨oder, “Integrating generic sensor fusion algorithms with sound state representations through encap- sulation of manifolds,” Information Fusion , vol. 14, no. 1, pp. 57–77, 2013

    2013

  18. [25]

    The banana distribution is gaussian: A localization study with exponential coordinates,

    A. W. Long, K. C. Wolfe, M. J. Mashner, G. S. Chirikjian et al., “The banana distribution is gaussian: A localization study with exponential coordinates,” Robotics: Science and Systems VIII , vol. 265, p. 1, 2013

    2013

  19. [26]

    Associating uncertainty with three- dimensional poses for use in estimation problems,

    T. D. Barfoot and P. T. Furgale, “Associating uncertainty with three- dimensional poses for use in estimation problems,” IEEE Transactions on Robotics, vol. 30, no. 3, pp. 679–693, 2014

    2014

  20. [27]

    A white-noise-on-jerk motion prior for continuous-time trajectory estimation on se (3),

    T. Y . Tang, D. J. Yoon, and T. D. Barfoot, “A white-noise-on-jerk motion prior for continuous-time trajectory estimation on se (3),” IEEE Robotics and Automation Letters , vol. 4, no. 2, pp. 594–601, 2019

    2019

  21. [28]

    A data- driven motion prior for continuous-time trajectory estimation on se (3),

    J. N. Wong, D. J. Yoon, A. P. Schoellig, and T. D. Barfoot, “A data- driven motion prior for continuous-time trajectory estimation on se (3),” IEEE Robotics and Automation Letters , vol. 5, no. 2, pp. 1429–1436, 2020

    2020

  22. [29]

    Characterizing the uncertainty of jointly distributed poses in the lie algebra,

    J. G. Mangelson, M. Ghaffari, R. Vasudevan, and R. M. Eustice, “Characterizing the uncertainty of jointly distributed poses in the lie algebra,” IEEE Transactions on Robotics, vol. 36, no. 5, pp. 1371–1388, 2020

    2020

  23. [30]

    Exploiting symmetries to design ekfs with consistency properties for navigation and slam,

    M. Brossard, A. Barrau, and S. Bonnabel, “Exploiting symmetries to design ekfs with consistency properties for navigation and slam,” IEEE Sensors Journal, vol. 19, no. 4, pp. 1572–1579, 2018

    2018

  24. [31]

    Consistent ekf-based visual-inertial odometry on matrix lie group,

    S. Heo and C. G. Park, “Consistent ekf-based visual-inertial odometry on matrix lie group,” IEEE Sensors Journal , vol. 18, no. 9, pp. 3780– 3788, 2018

    2018

  25. [32]

    Left-invariant extended kalman filter and attitude estima- tion,

    S. Bonnabel, “Left-invariant extended kalman filter and attitude estima- tion,” in IEEE Conference on Decision and Control , 2007, pp. 1027– 1032

    2007

  26. [33]

    Distributed invariant extended kalman filter for 3-d dynamic state estimation using lie groups,

    J. Xu, P. Zhu, and W. Ren, “Distributed invariant extended kalman filter for 3-d dynamic state estimation using lie groups,” in American Control Conference, 2022, pp. 2367–2372

    2022

  27. [34]

    Lie algebraic unscented kalman filter for pose estimation,

    A. M. Sjøberg and O. Egeland, “Lie algebraic unscented kalman filter for pose estimation,” IEEE Transactions on Automatic Control , vol. 67, no. 8, pp. 4300–4307, 2021

    2021

  28. [35]

    Unscented state estimation for rigid body motion on se (3),

    J. Bohn and A. K. Sanyal, “Unscented state estimation for rigid body motion on se (3),” in IEEE Conference on Decision and Control , 2012, pp. 7498–7503

    2012

  29. [36]

    An engineer’s guide to particle filtering on matrix lie groups,

    G. Marjanovic and V . Solo, “An engineer’s guide to particle filtering on matrix lie groups,” in IEEE International Conference on Acoustics, Speech and Signal Processing , 2016, pp. 3969–3973

    2016

  30. [37]

    On-manifold preintegration for real-time visual–inertial odometry,

    C. Forster, L. Carlone, F. Dellaert, and D. Scaramuzza, “On-manifold preintegration for real-time visual–inertial odometry,”IEEE Transactions on Robotics, vol. 33, no. 1, pp. 1–21, 2016

    2016

  31. [38]

    Computing map trajectories by rep- resenting, propagating and combining pdfs over groups,

    Drummond and Roussopoulos, “Computing map trajectories by rep- resenting, propagating and combining pdfs over groups,” in IEEE International Conference on Computer Vision , 2003, pp. 1275–1282

    2003

  32. [39]

    Bayesian fusion on lie groups,

    K. C. Wolfe, M. Mashner, and G. S. Chirikjian, “Bayesian fusion on lie groups,” Journal of Algebraic Statistics , vol. 2, no. 1, 2011

    2011

  33. [40]

    A novel federated structure of invariant ekf using left-/right-invariant observations,

    J. H. Hwang, J. Cha, and C. G. Park, “A novel federated structure of invariant ekf using left-/right-invariant observations,” IEEE Sensors Journal, vol. 22, no. 21, pp. 20 645–20 654, 2022

    2022

  34. [41]

    Tracking multiple unmanned aerial vehicles on se (3) using a monocular camera,

    M. E. Petersen, J. C. Ellingson, and R. W. Beard, “Tracking multiple unmanned aerial vehicles on se (3) using a monocular camera,” IEEE Transactions on Aerospace and Electronic Systems , vol. 59, no. 4, pp. 3726–3735, 2022

    2022

  35. [42]

    Joint localization based on split covariance intersection on the lie group,

    L. Li and M. Yang, “Joint localization based on split covariance intersection on the lie group,” IEEE Transactions on Robotics , vol. 37, no. 5, pp. 1508–1524, 2021

    2021

  36. [43]

    Distributed invariant extended kalman filter using lie groups: Algorithm and experiments,

    J. Xu, P. Zhu, Y . Zhou, and W. Ren, “Distributed invariant extended kalman filter using lie groups: Algorithm and experiments,” IEEE Transactions on Control Systems Technology , vol. 31, no. 6, pp. 2777– 2789, 2023

    2023

  37. [45]

    Unscented kalman filter on lie groups for visual inertial odometry,

    M. Brossard, S. Bonnabel, and A. Barrau, “Unscented kalman filter on lie groups for visual inertial odometry,” in IEEE/RSJ International Conference on Intelligent Robots and Systems . IEEE, 2018, pp. 649– 655

    2018

  38. [46]

    Analytic imu preintegration that associates uncertainty on matrix lie groups for consistent visual–inertial navigation systems,

    S.-H. Tsao and S.-S. Jan, “Analytic imu preintegration that associates uncertainty on matrix lie groups for consistent visual–inertial navigation systems,” IEEE Robotics and Automation Letters, vol. 8, no. 6, pp. 3820– 3827, 2023

    2023

  39. [47]

    A lie group manifold- based nonlinear estimation algorithm and its application to low-accuracy sins/gnss integrated navigation,

    S. Du, Y . Huang, B. Lin, J. Qian, and Y . Zhang, “A lie group manifold- based nonlinear estimation algorithm and its application to low-accuracy sins/gnss integrated navigation,” IEEE Transactions on Instrumentation and Measurement, vol. 71, pp. 1–27, 2022

    2022

  40. [48]

    On lie group imu and linear velocity preintegration for autonomous navigation considering the earth rotation compensation,

    P. Vial, J. Sol `a, N. Palomeras, and M. Carreras, “On lie group imu and linear velocity preintegration for autonomous navigation considering the earth rotation compensation,” IEEE Transactions on Robotics , vol. 41, pp. 1346–1364, 2025

    2025

  41. [49]

    Lie-theoretic multi-robot localization,

    X. Li and G. S. Chirikjian, “Lie-theoretic multi-robot localization,” in Riemannian Computing in Computer Vision . Springer, 2016, pp. 165– 186

    2016

  42. [50]

    Sharing visual-inertial data for collaborative decentralized simultaneous localization and mapping,

    R. Dubois, A. Eudes, and V . Fr ´emont, “Sharing visual-inertial data for collaborative decentralized simultaneous localization and mapping,” Robotics and Autonomous Systems , vol. 148, p. 103933, 2022

    2022

  43. [51]

    On data sharing strategy for decentralized collaborative visual- inertial simultaneous localization and mapping,

    ——, “On data sharing strategy for decentralized collaborative visual- inertial simultaneous localization and mapping,” in IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems , 2019, pp. 2123– 2130

    2019

  44. [52]

    Distributed collaborative localization of multiple vehicles from relative pose measurements,

    J. Knuth and P. Barooah, “Distributed collaborative localization of multiple vehicles from relative pose measurements,” in 47th Annual Allerton Conference on Communication, Control, and Computing, 2009, pp. 314–321

    2009

  45. [53]

    Multi- robot relative pose estimation in se(2) with observability analysis: A comparison of extended kalman filtering and robust pose graph opti- mization,

    G. Shin, H. Sim, S. Nam, Y . Kim, J. Heo, and K.-K. K. Kim, “Multi- robot relative pose estimation in se(2) with observability analysis: A comparison of extended kalman filtering and robust pose graph opti- mization,” IEEE Transactions on Intelligent Vehicles , pp. 1–23, 2024

    2024

  46. [54]

    Distributed invariant kalman filter for cooperative localization using matrix lie groups,

    Y . Zhou, Y . Liu, P. Zhu, and X. Wang, “Distributed invariant kalman filter for cooperative localization using matrix lie groups,” arXiv preprint arXiv:2405.04000, 2024

    2024 arXiv

  47. [55]

    Distributed invariant kalman filter for object-level multi-robot pose slam,

    H. Li, Q. Zeng, H. Li, Y . Zhang, and J. Wu, “Distributed invariant kalman filter for object-level multi-robot pose slam,” arXiv preprint arXiv:2409.09410, 2024. IEEE TRANSACTIONS ON ROBOTICS, VOL. , NO. , APRIL 2025 15

    2024 arXiv

  48. [56]

    Fiducial markers for pose estimation: Overview, applica- tions and experimental comparison of the artag, apriltag, aruco and stag markers,

    M. Kalaitzakis, B. Cain, S. Carroll, A. Ambrosi, C. Whitehead, and N. Vitzilaios, “Fiducial markers for pose estimation: Overview, applica- tions and experimental comparison of the artag, apriltag, aruco and stag markers,” Journal of Intelligent & Robotic Systems , vol. 101, p...

    2021

  49. [57]

    Absolute humanoid localization and mapping based on imu lie group and fiducial markers,

    M. Fourmy, D. Atchuthan, N. Mansard, J. Sola, and T. Flayols, “Absolute humanoid localization and mapping based on imu lie group and fiducial markers,” in IEEE/RAS 19th International Conference on Humanoid Robots, 2019, pp. 237–243

    2019

  50. [58]

    Visual localization of inspection robot using extended kalman filter and aruco markers,

    J. Zheng, S. Bi, B. Cao, and D. Yang, “Visual localization of inspection robot using extended kalman filter and aruco markers,” in IEEE Inter- national Conference on Robotics and Biomimetics , 2018, pp. 742–747

    2018

  51. [59]

    Indoor localization for an autonomous model car: A marker-based multi-sensor fusion framework,

    X. Li, S. Patel, D. Stronzek-Pfeifer, and C. B ¨uskens, “Indoor localization for an autonomous model car: A marker-based multi-sensor fusion framework,” in IEEE 19th International Conference on Automation Science and Engineering , 2023, pp. 1–8

    2023

  52. [60]

    Joon and W

    A. Joon and W. Kowalczyk, “Leader–follower approach for non- holonomic mobile robots based on extended kalman filter sensor data fusion and extended on-board camera perception controlled with behav- ior tree,” Sensors, vol. 23, no. 21, p. 8886, 2023

    2023

  53. [61]

    Multi-robot real-time cooperative localization based on high-speed feature detection and two-stage filtering,

    Z. Zhou, W. Tang, Z. Wang, L. Wang, and R. Zhang, “Multi-robot real-time cooperative localization based on high-speed feature detection and two-stage filtering,” in IEEE International Conference on Real-time Computing and Robotics , 2021, pp. 690–696

    2021

  54. [62]

    Cooperative localization by fusing pose estimates from static environmental and mobile fiducial features,

    Z. Li, R. Acuna, and V . Willert, “Cooperative localization by fusing pose estimates from static environmental and mobile fiducial features,” in Latin American Robotic Symposium, Brazilian Symposium on Robotics and Workshop on Robotics in Education . IEEE, 2018, pp. 65–70

    2018

  55. [63]

    Probabilistic trajectory estimation based leader following for multi-robot systems,

    M. Shan, Y . Zou, M. Guan, C. Wen, K.-Y . Lim, C.-L. Ng, and P. Tan, “Probabilistic trajectory estimation based leader following for multi-robot systems,” in 14th International Conference on Control, Automation, Robotics and Vision . IEEE, 2016, pp. 1–6

    2016

  56. [64]

    A leader-follower formation control of mobile robots by position-based visual servo method using fisheye camera,

    S. Oh-hara and A. Fujimori, “A leader-follower formation control of mobile robots by position-based visual servo method using fisheye camera,” ROBOMECH Journal, vol. 10, no. 1, p. 30, 2023

    2023

  57. [65]

    Vision-based leader-follower mobile robots for coop- erative object handling,

    D. Jayarathne, H. Vidumal, S. Senevirathne, P. Jayasekara, and T. Lalitharatne, “Vision-based leader-follower mobile robots for coop- erative object handling,” in 19th International Conference on Control, Automation and Systems . IEEE, 2019, pp. 747–752

    2019

  58. [66]

    Non-linear state error based extended kalman filters with applications to navigation,

    A. Barrau, “Non-linear state error based extended kalman filters with applications to navigation,” Ph.D. dissertation, Mines Paristech, 2015

    2015

  59. [67]

    R. M. Murray, Z. Li, and S. S. Sastry, A mathematical introduction to robotic manipulation. CRC press, 2017

    2017

  60. [68]

    T. D. Barfoot, State estimation for robotics . Cambridge University Press, 2017

    2017

  61. [69]

    Continuous- discrete extended kalman filter on matrix lie groups using concentrated gaussian distributions,

    G. Bourmaud, R. M ´egret, M. Arnaudon, and A. Giremus, “Continuous- discrete extended kalman filter on matrix lie groups using concentrated gaussian distributions,” Journal of Mathematical Imaging and Vision , vol. 51, pp. 209–228, 2015

    2015

  62. [70]

    Discrete extended kalman filter on lie groups,

    G. Bourmaud, R. M ´egret, A. Giremus, and Y . Berthoumieu, “Discrete extended kalman filter on lie groups,” in European Signal Processing Conference, 2013, pp. 1–5

    2013

  63. [71]

    T. K. Moon and W. C. Stirling, Mathematical methods and algorithms for signal processing . Prentice hall Upper Saddle River, NJ, 2000, vol. 1

    2000

  64. [72]

    On tests and measures of group divergence,

    P. C. Mahalanobis, “On tests and measures of group divergence,” J. Asiat. Soc. Bengal , vol. 26, pp. 541–588, 1930

    1930

  65. [73]

    Detecting anomalies in un- manned vehicles using the mahalanobis distance,

    R. Lin, E. Khalastchi, and G. A. Kaminka, “Detecting anomalies in un- manned vehicles using the mahalanobis distance,” in IEEE international conference on robotics and automation , 2010, pp. 3038–3044

    2010

  66. [74]

    Phasespace motion capture,

    “Phasespace motion capture,” accessed: 2025-02-05. [Online]. Available: https://www.phasespace.com/

    2025

  67. [75]

    Agilex limo robot,

    “Agilex limo robot,” accessed: 2025-02-05. [Online]. Available: https://global.agilex.ai/products/limo-pro

    2025

  68. [76]

    Camera calibration,

    “Camera calibration,” accessed: 2025-02-05. [Online]. Available: http://wiki.ros.org/camera calibration

    2025

  69. [77]

    The aruco ros library,

    “The aruco ros library,” accessed: 2025-02-05. [Online]. Available: http://wiki.ros.org/aruco ros APPENDIX A PROPERTIES OF THE DEVELOPED LIE GROUPGX We begin by showing GX is a matrix Lie group, i.e., verifying its closure, associativity, and the existence of identity and inve...

    2025

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.