Pith. sign in

REVIEW 4 major objections 4 minor 38 references

The Invariant Zonotopic Set-Membership Filter for State Estimation on Groups

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper claims that set-membership state estimation can be moved onto Lie groups by propagating a zonotope along the invariant estimation error, with the true state guaranteed to stay inside the propagated set.

desk verdict The paper's novel combination of zonotopic set-membership filtering with invariant filtering on Lie groups is worth reading, but the central containment theorem is not proven because it drops all higher-order terms without bounding or inflating the set. read the letter →

arxiv 2506.08530 v1 pith:3OUR6B32 submitted 2025-06-10 eess.SY cs.SY

Tao Li , Yi Li , Lulin Zhang , Jiuxiang Dong This is my paper
classification eess.SYcs.SY
keywords set-membershipfilteringinvariantLiegroupszonotopestateestimationgroupaffinesystemsunknownbutboundednoiseobservergaintuning
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

State estimation usually assumes noise has known statistics; this paper handles the opposite case, where noise is only known to be bounded. It moves the problem onto a Lie group, where the system's symmetry makes the estimation error evolve almost linearly, and represents uncertainty as a zonotope wrapped around the current state estimate. The paper claims that if the true state starts inside such a zonotope, the proposed recursion keeps it inside the updated zonotope at every future step, and that this invariant version outperforms the ordinary Euclidean zonotopic set-membership filter. The practical payoff would be guaranteed error enclosures for systems like vehicles, robots, or aircraft whose noise models are unknown, with better convergence and smaller uncertainty intervals than the traditional filter.

What carries the argument

The central object is the invariant zonotope on a Lie group, $X=(X \, \dotplus_G \, H)_\varrho$, whose error part is $H=\exp(\Xi)=\exp([\langle 0,H\rangle]^\wedge)$: a set of group elements obtained by composing a center $X$ with exponentials of every vector in a Euclidean zonotope pulled back to the Lie algebra. The argument runs through the left or right invariant estimation error, which the exponential map converts into a vector-space error $\varepsilon$; the group affine property $f(X_1,u)^{-1}f(X_2,u)=g(X_1^{-1}X_2,u)$ with $g(\exp([\varepsilon]^\wedge),u)=\exp([A\varepsilon]^\wedge)+O(\|\varepsilon\|^2)$ makes that error evolve linearly to first order. The generator update in Theorem 3.1, $H_{k+1}=[(A_k-L_kC_k)R_s(H_k),\,D_w H_w,\,-L_k D_v H_v]$, is the algorithm that carries the set forward, with $R_s$ keeping the zonotope order bounded.

What would settle it

Using the SE(2) vehicle setup of Section IV, take a single step with an initial error whose norm is not small and noise at its bounds, compute the exact set $\{f(X_0,u_0)\exp([D_w w_0]^\wedge): w_0\in W_0\}$, and compare it with the zonotope generated by equation (10d). If any exact reachable point falls outside that zonotope, the enclosure claim of Theorem 3.1 fails at that step; repeating the test with decreasing noise bounds would locate where the linearization error breaks containment.

Watch

Extended reading notes

Core claim

The central claim is Theorem 3.1: for the group affine discrete system (5) with unknown but bounded process noise $w_k$ and measurement noise $v_k$ described by zonotopes, if the state satisfies $X_k \in (\hat{X}_k \, \dotplus_G \, H_k)_\varrho$ with $H_k=\exp([\langle 0,H_k\rangle]^\wedge)$, then the update $\hat{X}_{k+1}=(f(\hat{X}_k,u_k),\,\exp([L_k z_k]^\wedge),\,\bullet)_\varrho$ and $H_{k+1}=[(A_k-L_kC_k)R_s(H_k),\,D_w H_w,\,-L_k D_v H_v]$ produces a zonotope that still contains $X_{k+1}$. The proof's mechanism is that the invariant error $\eta=\hat{X}^{-1}X$ (left case) or $X\hat{X}^{-1}$ (right case) lifts to a vector-space error $\varepsilon$ through the exponential map, and the group affine property linearizes its one-step evolution to equation (11). This yields the Invariant Zonotopic Set-Membership Filter (InZSMF), and the paper demonstrates on an $SE(2)$ vehicle example that the resulting center estimates, uncertainty intervals, and convergence speed improve on the traditional Euclidean ZSMF, especially under large initial errors.

Load-bearing premise

The load-bearing premise is that in the proof of Theorem 3.1, products of exponentials can be replaced by first-order linear approximations and the discarded higher-order terms are small enough not to break the enclosure; the paper does not bound or inflate for those terms.

Editorial extensions

If this is right

  • If Theorem 3.1 holds, InZSMF is a genuine set-membership filter: the true state remains inside the propagated zonotope at every step, so the reported interval bounds are valid enclosures.
  • The left-invariant error dynamics are independent of the state trajectory in the noise-free case, so the filter does not need to be re-linearized along the actual path; this is why large initial errors degrade convergence less than in Euclidean ZSMF.
  • The observer gain can be tuned by pole placement, which shapes overshoot and transient behavior, or by minimizing the F-radius of the generator matrix, which shrinks the zonotope; the F-radius version is faster per step.
  • On the simulated SE(2) vehicle with poor initial estimates, the paper reports heading RMSE improvements up to about 53% and position RMSE improvements up to about 50% over ordinary ZSMF, with average interval areas also shrinking.
  • The method applies to any system whose discrete dynamics can be written as a group affine system on a matrix Lie group, so the class includes rigid-body pose estimation problems beyond the 2D vehicle example.

Reading between the lines

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

  • A rigorous version of Theorem 3.1 would likely require an explicit bound on the neglected second-order terms $O(\|\varepsilon_k\|^2)$ from the linearization and the Baker-Campbell-Hausdorff expansion, with the zonotope inflated by that bound; without it, the enclosure guarantee formally holds only when those higher-order terms are negligible.
  • The same invariant-error construction could be transplanted to other set representations such as ellipsoids, intervals, or polytopes, since the key step is only the linear lift of the error to the Lie algebra.
  • The reported advantage at large initial errors suggests a testable hybrid strategy: start with the F-radius-optimal gain for fast convergence and switch to pole-placement gain once the error set is small, combining Algorithm 2's speed with Algorithm 1's controlled overshoot.
  • A controlled comparison on the same system with exactly the same noise zonotopes, using a Euclidean ZSMF applied to group coordinates, could separate how much of the gain comes from the symmetry structure versus the nonlinear coordinate change.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes an Invariant Zonotopic Set-Membership Filter (InZSMF) for discrete-time group-affine systems on Lie groups with unknown-but-bounded noise. The state uncertainty is represented by a group zonotope, and the filter propagates a zonotope of the invariant estimation error using a linearized error recursion. Two observer-gain tuning methods are proposed: pole placement and F-radius minimization. The central theoretical claim is Theorem 3.1, which states that if the true state lies in the group zonotope at time k, then the one-step propagated group zonotope contains the true state at time k+1. The paper also reports a 2D vehicle simulation comparing InZSMF with a conventional ZSMF and claims faster convergence, smaller RMSE, and smaller average interval area.

Significance. If valid, the paper would extend invariant filtering theory to zonotopic set-membership filtering on Lie groups, a natural direction given the success of the invariant extended Kalman filter. The two gain-tuning algorithms are concrete and implementable, and the simulation study is detailed, with runtime measurements and a clear experimental setup. The empirical comparison is not circular: noise bounds and pole locations are engineering inputs, not fitted parameters, and the paper does not force the superiority conclusion through the derivation. However, the value of the paper rests entirely on the guaranteed-enclosure property, and that property is not established by the current proof.

major comments (4)
  1. [Section III-A (proof of Theorem 3.1), Eq. (11) and (10d)] The proof of Theorem 3.1 is an approximate derivation, not a containment proof. It invokes Lemma 2.1 to replace g(exp([epsilon]^), u) by exp([A epsilon]^) + O(||epsilon||^2), invokes the BCH formula in Lemma 2.2 to replace products of exponentials by the exponential of a sum, and invokes Lemma 3.1 to linearize the innovation as C epsilon + O(||epsilon||^2). Each step introduces an uncontrolled remainder, and no bound on these remainders or set inflation is added to the generator update (10d). The conclusion X_{k+1} in (Xhat_{k+1} L_G H_{k+1}) therefore does not follow from the displayed chain of approximate equalities. The argument can at most motivate the approximate linear error recursion (11); it does not prove the set-membership guarantee that is the central contribution of the paper.
  2. [Section III-A, Eq. (10b)-(10d) and Definition 2.7] Even if the linearized vector error epsilon_{k+1} were shown to lie in <0,H_{k+1}>, the group set H_{k+1} = exp([<0,H_{k+1}>]^) would not necessarily contain the true product set. The invariant error eta_{k+1} is a product of exponentials, exp([-L z]) g(eta_k) exp([D_w w]), and the exponential map does not distribute over Minkowski sums on the Lie algebra because group multiplication is not vector addition. The BCH commutator terms that are dropped can be comparable to or larger than the zonotope generators, and the proof contains no argument that the actual product set is enclosed by exp([<0,H_{k+1}>]^). The proof must establish enclosure of the true group error set, not merely enclosure of an approximate logarithmic image.
  3. [Section III-B, Theorem 3.2] The stability theorem addresses only the linearized error equation (11). Its proof concludes boundedness of eta_k from boundedness of the linearized epsilon_k, but the epsilon_k in Theorem 3.2 is the state of the approximate linear recursion, not the true invariant error whose evolution in Theorem 3.1 involves dropped higher-order terms. Hence the actual nonlinear group error is not shown to be bounded or stable. In addition, the pole-placement construction requires A_k^varrho - L_k C_k^varrho to equal a fixed Schur matrix for all k; in general A_k and C_k may be time- or state-dependent, and the paper does not justify the existence of such a uniform L_k for the class of systems in (5).
  4. [Section IV, Experiment 1 and Tables I-II] The simulations show that the specific noise realizations used in the experiments remained inside the computed sets, but they do not test the worst-case containment property claimed in Theorem 3.1. A set-membership guarantee is a deterministic, worst-case property, and empirical containment on a few trajectories cannot substitute for the missing proof. Furthermore, the comparison is between a ZSMF based on an Euler discretization and an InZSMF based on a group discretization that itself neglects BCH terms in Appendix B; part of the reported performance difference may therefore be due to the different discrete models rather than to the invariant structure alone.
minor comments (4)
  1. [Section II, Definitions 2.6-2.7 and Eqs. (10a)-(10d)] The notation for the left- and right-invariant Minkowski sums is not applied consistently in the algorithm equations. The proof of Theorem 3.1 details only the left-invariant case and states that the right-invariant case is similar; please spell out the right-invariant generator update, including the order and signs of the D_w and D_v terms.
  2. [Section IV, AAR definitions] The definitions of AAR(theta) and AAR(x) use the same symbols theta_k and x_k for the true state and for the upper and lower interval bounds, which makes the formulas ambiguous. Please introduce distinct notation for the bounds and clarify whether AAR is an interval width, an area, or a ratio.
  3. [Tables I and II] The percentage improvement rows are not fully self-explanatory. The text defines the improvement as -(M1 - M2)/M1, but the tables do not specify which column corresponds to M1 and which to M2; please add column headers or a table note.
  4. [Throughout] There are numerous typographical and copyediting issues, including 'the law of union' for associativity, the repeated 'REFERENCES' header, and inconsistent use of subscripts and superscripts. A careful copyedit is needed before any resubmission.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the InZSMF derivation is self-contained and externally benchmarked; self-citations are not load-bearing.

full rationale

The paper's central construction is a propagated zonotope on a Lie group, not a fitted quantity renamed as a prediction. In Theorem 3.1, the generator update (10d) is the standard set-membership recursion: the linearized error set is mapped through (A_k - L_k C_k), then Minkowski-summed with the noise sets D_w W_k and -L_k D_v V_k. This is a definition of the proposed filter recursion, not an input secretly reused as the claimed containment result; the theorem's asserted containment is an additional mathematical claim about the true state, and the proof attempts to justify it from the group affine structure. The gain tuning methods are also not fitted to the performance metrics: the pole configuration method fixes A_k - L_k C_k = A-bar with a chosen spectral radius, and the F-radius method minimizes trace(P_{k+1}) from the model data A_k, C_k, D_w, D_v, H_w, H_v. No RMSE or AAR value from Section IV enters the derivation. The simulation comparison against the traditional ZSMF is an external, independent benchmark; the reported improvements are empirical observations, not consequences of circular definitions. The paper does cite prior work by its own authors (references [12] and [13]) in the introduction's list of set-membership applications, but these citations are not load-bearing for the invariant filtering construction. The main load-bearing external citation is Barrau and Bonnabel's invariant filtering framework [19], [21], which is not self-citation. The most serious concern in the paper is a soundness gap, not circularity: the proof of Theorem 3.1 repeatedly replaces exact exponential products by first-order approximations and drops O-terms without inflating the zonotope, so the claimed guaranteed containment is not established by the displayed argument. That is a correctness or rigor issue, not a case of the derivation reducing to its inputs by construction. The paper's own conclusion even acknowledges a scope limitation: the method requires the original system to be transformable into a group affine system. Overall, no circular step matching the enumerated patterns is present.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The paper introduces a new mathematical construction (zonotopes on a Lie group), but it is a definition, not a physical entity. The free parameters are design choices for the simulation, not fitted values. The main burden is the unquantified first-order linearization in the set propagation.

free parameters (2)
  • observer poles = [0.95, 0.98, 0.98]
    Chosen by hand for the pole configuration gain in Experiment 2; affects convergence and overshoot, but is a design choice, not fitted to data.
  • generator order reduction threshold = 30
    The paper states in Section IV that order reduction is performed when the generator matrix order exceeds 30; this threshold is user-selected and affects the precision-speed tradeoff.
assumptions (5)
  • standard math Group affine property of f and existence of A^rho_k (Lemma 2.1, from [21])
    Invoked in Theorem 3.1 to linearize the error evolution; this is a known property of invariant filtering.
  • domain assumption System (5) is a valid discrete-time group affine model with right-multiplicative process noise
    Remark 3 claims many two-frames systems can be transformed to this form; the simulation uses this for a 2D vehicle.
  • ad hoc to paper First-order expansion of the innovation (Lemma 3.1) is valid and higher-order terms are negligible for set containment
    The proof of Theorem 3.1 drops O(epsilon squared) terms without bounding them; this is a load-bearing assumption specific to the containment claim.
  • domain assumption Observability of the linearized error system to allow pole placement
    Theorem 3.2 assumes an observable system; the simulation implicitly assumes this for the vehicle model.
  • ad hoc to paper Baker-Campbell-Hausdorff approximation exp(A)exp(B) approximately exp(A+B) is accurate enough to ignore in discretization
    In Appendix B, X_k exp(delta U_k) exp(delta W_k) is used as an approximation for exp(delta(U+W)); the neglected commutator terms are not quantified in the set.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Invariant Zonotopic Set-Membership Filter for State Estimation on Groups." pith.science (2026). https://pith.science/paper/3OUR6B32

@misc{pith2026250608530,
  author       = {Pith},
  title        = {Pith review of: The Invariant Zonotopic Set-Membership Filter for State Estimation on Groups},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3OUR6B32}},
  note         = {Machine review of arXiv:2506.08530}
}
read the original abstract

The invariant filtering theory based on the group theory has been successful in statistical filtering methods. However, there exists a class of state estimation problems with unknown statistical properties of noise disturbances, and it is worth discussing whether the invariant observer still has performance advantages. In this paper, considering the problem of state estimation with unknown but bounded noise disturbances, an Invariant Zonotopic Set-Membership Filter (InZSMF) method on groups is innovatively proposed, which extends the invariant filtering theory to the field of non-statistical filtering represented by set-membership filtering. Firstly, the InZSMF method transforms the state space from the traditional Euclidean vector space to the Lie group space to construct group affine discrete systems with unknown but bounded noise uncertainty defined by the zonotope on groups. Secondly, the nonlinear observer on the group is defined and the corresponding linearized estimation error is derived. Then, two observer gain tuning algorithms under the InZSMF method are proposed, respectively, the pole configuration method and the F-radius optimization method. Finally, through simulation experiments, it is shown that the InZSMF state estimation method is generally superior to the traditional Zonotopic Set-Membership Filter (ZSMF) state estimation method. Especially, when the initial estimations are imprecise, the convergence speed of state estimation, the accuracy of set-membership center estimation, and the average interval area of zonotopic estimation of the InZSMF method are significantly better than those of the ZSMF method.

Figures

Figures reproduced from arXiv: 2506.08530 by the authors.

Figure 1
Figure 1. Illustration of the zonotope on the Lie group. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The comparison of the ZSMF and InZSMF methods. The left column () [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. The comparison of the ZSMF and InZSMF methods under two [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

38 extracted references · 36 canonical work pages

  1. [1]

    Set membership state and parameter estimation for systems described by nonlinear differential equations,

    T. Ra ¨ıssi, N. Ramdani, and Y . Candau, “Set membership state and parameter estimation for systems described by nonlinear differential equations,”Automatica, vol. 40, no. 10, pp. 1771–1777, 2004

    work page 2004

  2. [2]

    Zonotopes and kalman observers: Gain optimality under distinct uncertainty paradigms and robust convergence,

    C. Combastel, “Zonotopes and kalman observers: Gain optimality under distinct uncertainty paradigms and robust convergence,”Automatica, vol. 55, pp. 265–273, 2015

    2015

  3. [3]

    Stability of linear set-membership filters with respect to initial conditions: An observation-information perspective,

    Y . Cong, X. Wang, and X. Zhou, “Stability of linear set-membership filters with respect to initial conditions: An observation-information perspective,”Automatica, vol. 172, p. 111993, 2025

    work page 2025

  4. [4]

    A new approach to linear filtering and prediction problems,

    R. E. Kalman, “A new approach to linear filtering and prediction problems,”Transactions of the ASME—Journal of Basic Engineering, vol. 82, no. 1, pp. 35–45, 1960

    work page 1960

  5. [5]

    Rethinking the mathematical frame- work and optimality of set-membership filtering,

    Y . Cong, X. Wang, and X. Zhou, “Rethinking the mathematical frame- work and optimality of set-membership filtering,”IEEE Transactions on Automatic Control, vol. 67, no. 5, pp. 2544–2551, 2022

    work page 2022

  6. [6]

    Single-beacon localization for mobile robot: A set membership filtering approach,

    X. Qin, Y . Cong, J. Lai, J. Yang, and X. Wang, “Single-beacon localization for mobile robot: A set membership filtering approach,” IEEE Robotics and Automation Letters, vol. 9, no. 11, pp. 9899–9906, 2024

    work page 2024

  7. [7]

    A zonotope-based interval estimation method for systems with nonlinear equality constraints,

    Y . Ma, Z. Wang, M. Kinnaert, and Y . Shen, “A zonotope-based interval estimation method for systems with nonlinear equality constraints,” IEEE Transactions on Automatic Control, vol. 70, no. 2, pp. 1358–1365, 2025

    work page 2025

  8. [8]

    Interval estimation methods for discrete-time linear time-invariant systems,

    W. Tang, Z. Wang, Y . Wang, T. Ra¨ıssi, and Y . Shen, “Interval estimation methods for discrete-time linear time-invariant systems,”IEEE Transac- tions on Automatic Control, vol. 64, no. 11, pp. 4717–4724, 2019

    work page 2019

Show all 38 references
  1. [9]

    Observer gain optimization for minimization of minimal robust positively invariant set and invariant set-based fault detection,

    F. Xu, S. Olaru, and M. M. Seron, “Observer gain optimization for minimization of minimal robust positively invariant set and invariant set-based fault detection,”Automatica, vol. 160, p. 111413, 2024

    2024

  2. [10]

    Fault detection for satellite reaction wheel based on zonotopic set-membership estimation,

    Z. Wang, C. Xu, Z. Liu, and Y . Shen, “Fault detection for satellite reaction wheel based on zonotopic set-membership estimation,”IEEE Transactions on Aerospace and Electronic Systems, vol. 61, no. 2, pp. 1577–1588, 2025

    2025

  3. [11]

    Interval sliding mode observer based incipient sensor fault detection with application to a traction device in china railway high-speed,

    K. Zhang, B. Jiang, X. Yan, and J. Shen, “Interval sliding mode observer based incipient sensor fault detection with application to a traction device in china railway high-speed,”IEEE Transactions on Vehicular Technology, vol. 68, no. 3, pp. 2585–2597, 2019

    2019

  4. [12]

    Fault detection unknown input observer for local nonlinear fuzzy autonomous ground vehicles system based on a joint peak-to-peak analysis and zonotopic analysis threshold,

    Y . Li and J. Dong, “Fault detection unknown input observer for local nonlinear fuzzy autonomous ground vehicles system based on a joint peak-to-peak analysis and zonotopic analysis threshold,”IEEE Transac- tions on Vehicular Technology, pp. 1–11, 2025

    2025

  5. [13]

    State and fault interval estimation for discrete-time takagi–sugeno fuzzy systems via intermediate observer base on zonotopic analysis,

    L. Zhang, Z. Li, Y . Li, and J. Dong, “State and fault interval estimation for discrete-time takagi–sugeno fuzzy systems via intermediate observer base on zonotopic analysis,”IEEE Transactions on Fuzzy Systems, vol. 32, no. 11, pp. 6588–6593, 2024

    2024

  6. [14]

    Iterative inter- val estimation-based fault detection for discrete time t–s fuzzy systems,

    M. Shen, T. Zhang, Z.-G. Wu, Q.-G. Wang, and S. Zhu, “Iterative inter- val estimation-based fault detection for discrete time t–s fuzzy systems,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 53, no. 11, pp. 6966–6974, 2023

    2023

  7. [15]

    Dynamic event-triggered fault detection via zonotopic residual evaluation and its application to vehicle lateral dynamics,

    X. Wang, Z. Fei, H. Yan, and Y . Xu, “Dynamic event-triggered fault detection via zonotopic residual evaluation and its application to vehicle lateral dynamics,”IEEE Transactions on Industrial Informatics, vol. 16, no. 11, pp. 6952–6961, 2020

    2020

  8. [16]

    A note on event- triggered functional interval observer design for uncertain systems,

    J. Huang, Z. Yin, Y . Zhang, T. Ra ¨ıssi, and J. Fan, “A note on event- triggered functional interval observer design for uncertain systems,” IEEE Transactions on Instrumentation and Measurement, vol. 74, pp. 1–10, 2025

    2025

  9. [17]

    Secure estimation, attack isolation, and reconstruction based on zonotopic unknown input observer,

    H. Liu, Y . Li, Q.-L. Han, T. Ra ¨ıssi, and T. Chai, “Secure estimation, attack isolation, and reconstruction based on zonotopic unknown input observer,”IEEE Transactions on Automatic Control, vol. 68, no. 12, pp. 7312–7325, 2023

    2023

  10. [18]

    On invariant asymptotic observers,

    N. Aghannan and P. Rouchon, “On invariant asymptotic observers,” in Proceedings of the 41st IEEE Conference on Decision and Control, 2002., vol. 2, 2002, pp. 1479–1484 vol.2

    2002

  11. [19]

    The invariant extended kalman filter as a stable observer,

    A. Barrau and S. Bonnabel, “The invariant extended kalman filter as a stable observer,”IEEE Transactions on Automatic Control, vol. 62, no. 4, pp. 1797–1812, 2017

    2017

  12. [20]

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

    S. Bonnabel, “Left-invariant extended kalman filter and attitude estima- tion,” in2007 46th IEEE Conference on Decision and Control, 2007, pp. 1027–1032. 11

    2007

  13. [21]

    Invariant kalman filtering,

    A. Barrau and S. Bonnabel, “Invariant kalman filtering,”Annual Review of Control, Robotics, and Autonomous Systems, vol. 1, no. 1, pp. 237– 257, 2018

    2018

  14. [22]

    Symmetry-preserving ob- servers,

    S. Bonnabel, P. Martin, and P. Rouchon, “Symmetry-preserving ob- servers,”IEEE Transactions on Automatic Control, vol. 53, no. 11, pp. 2514–2526, 2008

    2008

  15. [23]

    A new class of symmetry- preserving observers,

    D. R. Lima, R. Ushirobira, and D. Efimov, “A new class of symmetry- preserving observers,” in2024 IEEE 63rd Conference on Decision and Control (CDC), 2024, pp. 7009–7013

    2024

  16. [24]

    The geometry of navigation problems,

    A. Barrau and S. Bonnabel, “The geometry of navigation problems,” IEEE Transactions on Automatic Control, vol. 68, no. 2, pp. 689–704, 2023

    2023

  17. [25]

    A nonlinear filter for pose estimation based on fast unscented transform on lie groups,

    Y . Jin, W. Zhang, J. Tang, H. Sun, and L. Shi, “A nonlinear filter for pose estimation based on fast unscented transform on lie groups,”IEEE Robotics and Automation Letters, vol. 9, no. 11, pp. 10 431–10 438, 2024

    2024

  18. [26]

    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

  19. [27]

    Improved kalman-particle kernel filter on lie groups applied to angles-only uav navigation,

    C. Chahbazian, K. Dahia, N. Merlinge, B. Winter-Bonnet, K. Honore, and C. Musso, “Improved kalman-particle kernel filter on lie groups applied to angles-only uav navigation,” in2022 International Conference on Robotics and Automation (ICRA), 2022, pp. 1689–1694

    2022

  20. [28]

    Unscented kalman filtering on manifolds for auv navigation - experimental results,

    S. T. Krauss and D. J. Stilwell, “Unscented kalman filtering on manifolds for auv navigation - experimental results,” inOCEANS 2022, Hampton Roads, 2022, pp. 1–6

    2022

  21. [29]

    Convergence and consistency analysis for a 3-d invariant-ekf slam,

    T. Zhang, K. Wu, J. Song, S. Huang, and G. Dissanayake, “Convergence and consistency analysis for a 3-d invariant-ekf slam,”IEEE Robotics and Automation Letters, vol. 2, no. 2, pp. 733–740, 2017

    2017

  22. [30]

    Invariant extended kalman filtering for underwater navigation,

    E. R. Potokar, K. Norman, and J. G. Mangelson, “Invariant extended kalman filtering for underwater navigation,”IEEE Robotics and Automa- tion Letters, vol. 6, no. 3, pp. 5792–5799, 2021

    2021

  23. [31]

    Contact- aided invariant extended kalman filtering for robot state estimation,

    R. Hartley, M. Ghaffari, R. M. Eustice, and J. W. Grizzle, “Contact- aided invariant extended kalman filtering for robot state estimation,”The International Journal of Robotics Research, vol. 39, no. 4, pp. 402–430, 2020

    2020

  24. [32]

    Laplace particle filter on lie groups applied to angles-only navigation,

    C. Chahbazian, N. Merlinge, K. Dahia, B. Winter-Bonnet, J. Marini, and C. Musso, “Laplace particle filter on lie groups applied to angles-only navigation,” in2021 IEEE 24th International Conference on Information Fusion (FUSION), 2021, pp. 1–8

    2021

  25. [33]

    Unscented kalman filtering on lie groups,

    M. Brossard, S. Bonnabel, and J.-P. Condomines, “Unscented kalman filtering on lie groups,” in2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017, pp. 2485–2491

    2017

  26. [34]

    Lie symmetries applied to interval integration,

    J. Damers, L. Jaulin, and S. Rohou, “Lie symmetries applied to interval integration,”Automatica, vol. 144, p. 110502, 2022

    2022

  27. [35]

    Set inversion and box contraction on lie groups using interval analysis,

    N. Merlinge, “Set inversion and box contraction on lie groups using interval analysis,”Automatica, vol. 165, p. 111688, 2024

    2024

  28. [36]

    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

  29. [37]

    Linear observed systems on groups,

    A. Barrau and S. Bonnabel, “Linear observed systems on groups,” Systems Control Letters, vol. 129, pp. 36–42, 2019

    2019

  30. [38]

    Set-membership estimation based on ellipsoid bundles for discrete-time lpv descriptor systems,

    W. Tang, Q. Zhang, Z. Wang, and Y . Shen, “Set-membership estimation based on ellipsoid bundles for discrete-time lpv descriptor systems,” Automatica, vol. 145, p. 110580, 2022

    2022

Pith tools

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