Pith. sign in

REVIEW 4 major objections 3 minor 39 references

Leveraging Equivariances and Symmetries in the Control Barrier Function Synthesis

T0 review · 4 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read The paper proves that equivariant dynamics and symmetric constraints induce symmetries in reachability-based control barrier functions, so values on one slice determine the whole safe set.

desk verdict Solid symmetry-transfer theorem for reachability-based CBFs, plus a promising but under-verified extension to non-symmetric constraints. read the letter →

arxiv 2509.04399 v1 pith:AFM5NGQ7 submitted 2025-09-04 eess.SY cs.ROcs.SY

classification eess.SYcs.ROcs.SY
keywords controlbarrierfunctionsequivariancesymmetryreachabilityanalysisDiniderivativesafesetinvariancekinematicbicyclemodelcomputationalreduction
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

Control Barrier Functions (CBFs) are a standard tool for safe control, but constructing one usually means solving a numerical optimal-control problem at every state. This paper proves that when the system dynamics are equivariant under a transformation, and the constraint set is symmetric under the same transformation, the reachability-based CBF inherits that symmetry: its value at a transformed state equals its value at the original state. Consequently, the CBF needs to be computed only on a slice of state space, with the rest inferred by applying the transformation. The paper then shows that equivariances alone, without symmetric constraints, allow a partially known CBF to be dragged around a new constraint boundary, and the pointwise maximum of the dragged copies is itself a valid CBF. If correct, the results turn symmetry into a computational lever, and numerical experiments report up to a 99% reduction in synthesis time.

What carries the argument

The central object is the reachability-based CBF HT of (5): a max-min over input trajectories of the running minimum of the constraint function h minus a penalty γt. Around it sits the parameterized diffeomorphism D(x;p) encoding constraint symmetry (h(x)=h(D(x;p))) and dynamics equivariance (f(D(x;p), D_u(u;p))=J_D(x;p)f(x,u)). The proof maps any input trajectory to its D-image; the max-min value is invariant, so HT inherits the symmetry. The second construction replaces HT with an arbitrary (possibly partial) CBF b and defines B(x)=max_{σ∈S_P(x)} b(D(x;σ)); condition (31) — b locally maximal in σ on the slice — is what turns the Dini inequality for b into the Dini inequality for B, while A

What would settle it

Take the ellipse-dragging example (Example 3): sample a grid of points x on M and small parameter perturbations Δσ, and evaluate the local-max inequality b(D(x;σ)) ≥ b(D(x;σ+Δσ)). A single violation is a counterexample to condition (31); if the Dini inequality for B fails at that state, Theorem 9's conclusion fails as well.

Watch

Extended reading notes

Core claim

The paper's first main result (Theorem 3) states that if the constraint function h is symmetric under a diffeomorphism D (h(x)=h(D(x;p))), the dynamics are strongly equivariant under D (f(D(x;p), D_u(u;p)) = J_D(x;p) f(x,u)), and the reachability target set F is invariant under D, then the reachability-based control barrier function HT defined by (5) satisfies HT(x)=HT(D(x;p)) for all x and p. Hence HT is fully determined by its values on a slice M via the identity (16). The second result relaxes the requirement of symmetric constraints: given any CBF b, whether fully known (Theorem 8) or known only on a slice with a local-maximality condition (Theorem 9), the function B(x)=max_σ b(D(x;σ)) i

Load-bearing premise

The load-bearing premise is condition (31): the partially known safety function b must not increase when the symmetry transformation that slides it along the constraint is perturbed slightly; if this local-maximality fails, the max-over-sigma trace need not be a valid barrier function.

Editorial extensions

If this is right

  • Safe-set certificates for equivariant systems can be built from a slice: only points on a symmetry-reduced set M need the expensive reachability computation; the rest follow from (16) or (34).
  • The numerical study shows total computation time falling to about 1.0%–12% of the direct method in the tested systems, with the fraction of explicitly computed grid points as low as 0.59% of the domain.
  • A single partially known CBF, computed for a conservative approximation of one constraint, can be dragged to synthesize CBFs for multiple new constraints, including non-symmetric convex corners and non-convex obstacles.
  • The equivariance-based construction of Theorems 8 and 9 works for any CBF in the Dini sense, not only the reachability-based HT, so the savings compound on top of whatever synthesis method produced b.
  • Because the zero super-level set of the constructed B is a subset of the target constraint set (C ⊆ H), the resulting controller preserves the original safety specification even when the method is applied conservatively.

Reading between the lines

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

  • The identical-structure argument in Theorem 3 suggests that any value function defined by a max-min optimal-control program whose data are equivariant and symmetric will inherit the same symmetry, not just the specific HT in (5); one could test this by replacing the running cost h minus γt with any function invariant under the same D.
  • Condition (31) is the real bottleneck for practice; an automated verification procedure — sampling x on M and σ and checking the local-max inequality — is an obvious next step, and a failure might be fixable by shrinking M or choosing a more conservative approximation h-tilde.
  • The reported savings should grow with state-space dimension: the slice M has dimension roughly that of the symmetry orbits, so in attitude-rich systems with larger symmetry groups (e.g., 3D rotations) the ratio of explicit to inferred points could be far smaller than the 1–3% seen here.
  • The trace construction resembles a geometric sweep: sliding the slice along the boundary of a constraint is analogous to a Minkowski-sum-like dilation of a known safe set; making this analogy precise could give constructive recipes for choosing D, M, and b for arbitrary obstacle geometries.
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 / 3 minor

Summary. The paper studies how equivariances in the dynamics and symmetries in state constraints can be exploited in Control Barrier Function (CBF) synthesis. The first part considers the reachability-based CBF H_T defined by the optimal control problem in Eq. (5). Under strong equivariance of the dynamics, symmetry of the constraint h, and invariance of the known safe set F, Theorem 3 asserts H_T(x)=H_T(D(x;p)), so H_T can be computed on a slice M and inferred elsewhere via Eq. (16). A local version is given in Theorem 4, and several case studies (pendulum, bicycle kinematics, linear systems) illustrate the slice construction. The second part relaxes the need for symmetric constraints. Lemma 7 states that dragging a CBF b through an equivariance diffeomorphism yields another CBF. Theorem 8 forms B(x)=max_σ b(D(x;σ)) and proves it is a CBF when b is fully known. Theorem 9 is the main extension: using only a partially known CBF on a set M and conditions (31)-(33), the function B(x)=max_{σ∈S_P(x)} b(D(x;σ)) in Eq. (34) is claimed to be a CBF. Examples 3-4 apply this to a kinematic bicycle model with an elliptical obstacle and with a corner, and Section VI reports computational gains up to about 99% versus direct pointwise evaluation, with code released on GitHub.

Significance. If the technical gaps identified below are resolved, the paper makes a useful contribution: it identifies structural conditions under which expensive reachability-based CBF synthesis can be reduced to a symmetry slice, and it proposes a method for recycling a partially known CBF to handle non-symmetric constraints. The proofs of Theorem 3 and Theorem 8 are largely correct in structure, and the numerical study is substantial and reproducible. The idea of using equivariance to 'drag' a known CBF along a constraint boundary is appealing and goes beyond the existing symmetry-constrained CBF literature. However, the main new result for partial knowledge, Theorem 9, rests on assumptions that are not rigorously verified in the examples, and the Lipschitz-regularity step is asserted too quickly. These issues are fixable but are central to the paper's claimed scope.

major comments (4)
  1. [Section III-A, Theorem 3] The statement H_T(x)=H_T(D(x;p)) for all x∈D is only meaningful if D(x;p)∈D. The theorem does not assume D(D;p)⊆D; D(F;p)=F is not enough to imply this. The proof of (11)-(14) also uses D^{-1}(F;p)=F and D_u^{-1}(U;p)=U, which do follow from the stated invariance and strong equivariance, but the missing domain condition on D is needed. Please add D(D;p)⊆D as an assumption, or restrict the conclusion to x∈D∩D^{-1}(D;p).
  2. [Section V-A, Lemma 7] The proof asserts 'Since b is a CBF, it holds db(x; f(x,u)) ≥ -α(b(x))' for the arbitrary u used in the chain. Definition 1 only guarantees the inequality for some control in U, not for every u. To conclude sup_u dbσ(x; f(x,u)) ≥ -α(bσ(x)), one needs the realizing control to lie in the image of D_u, i.e. D_u(U;σ)=U, which is strong equivariance. If Definition 3's 'isomorphism' is meant to make D_u a bijection U→U, then D_u(U)=U automatically and the weak/strong distinction collapses; if it only means D_u(U)⊆U, Lemma 7 is not justified. Please clarify Definition 3 and state the exact assumption needed by Lemma 7 and therefore by Theorems 8-9.
  3. [Section V-B, Theorem 9, Eq. (31)] Condition (31) is the load-bearing premise in the proof of the Dini inequality: it is exactly the step b_{σ̃}(φ_{Δt}) ≥ b_{σ0}(φ_{Δt}) in Eqs. (36)-(40). In Examples 3 and 4 it is claimed to 'follow directly from the geometric construction' or to be 'verified analogously', but no derivation is given for the specific b=H_T from (5) and D in (44). A single point where (31) fails breaks the chain. The paper even admits in Section V-C that verifying (31) 'may be non-trivial in some cases.' Please provide a rigorous verification for Examples 3 and 4, or a general sufficient condition that can be checked. In addition, the proof applies (31) with ε(σ0) at points where the condition is stated with ε(σ̃); this needs a uniformity/lower-semicontinuity argument or a strengthened assumption.
  4. [Section V-B / Example 3, Assumption 4 and Lemma 10] The text states 'As D is smooth in both of its arguments, the regularity assumption, Assumption 4, holds.' Smoothness of D does not imply Assumption A4.2, the Lipschitz single-valued localization of S_P; S_P is set-valued exactly on overlapping regions such as the dark blue region in Fig. 16. A4.2 is a transversality/univalence condition that must be verified. Since local Lipschitz continuity of B is part of the CBF definition, the assertion that B is locally Lipschitz continuous is currently unsupported. Equation (43) is also asserted without proof in Example 3. Please either prove these conditions for the constructed D and M, or state a more explicit verifiable sufficient condition.
minor comments (3)
  1. [Definition 3 and Definition 5] The codomain of D_u is written as U→U with the additional condition D_u(U;p)⊆U, which is redundant if the codomain is U. This ambiguity should be fixed, as it interacts with the weak/strong equivariance distinction used later.
  2. [Theorem 8 and Theorem 9] Definition 1 requires the zero super-level set C to be compact. The theorems state formulas for C (Eqs. (29) and (35)) but do not prove compactness or add it as a standing assumption. This should be stated explicitly, even if it is automatic in the examples.
  3. [Throughout] There are several typos and minor notational slips: in the proof of Theorem 3 the displayed dynamics constraint reads 'xpxq=f(xpsq,upsqq' instead of 'ẋ(s)=f(x(s),u(s))'; Corollary 11 has 'satisfing'; the local equivariance definitions and Figure 10 could be clearer about the composition order of D_I, D_II, D_III. These do not affect the results but should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the symmetry-transfer results are proved by change-of-variables and explicit assumptions; the unverified condition (31) is a correctness gap, not a circular step.

full rationale

The derivation chain is self-contained rather than circular. Theorem 3 proves HT(D(x;p)) = HT(x) by a direct change of variables in the reachability optimization (5), using only the assumed symmetries of h, strong equivariance of f, and D-invariance of F; no fitted parameter or conclusion is embedded in the definition of HT. The domain reduction (16) is an exact corollary, not an extrapolation from fitted values. Theorem 8 and Lemma 7 establish that dragging a known CBF through an equivariant diffeomorphism preserves the CBF property, and the max construction is proven via the Dini inequality with the super-level set computed explicitly. Theorem 9 is conditional on explicit sufficient conditions (31)-(33), (43), and Assumption 4; these are premises, not restatements of the conclusion. The paper itself flags that verifying (31) "may be non-trivial in some cases" (Section V-C), and Examples 3-4 give only geometric sketches rather than full proofs of (31). That is an unsupported step and a correctness risk, but it is not circular: no equation in the proof reduces to (31) by construction, and the theorem is honestly stated as conditional. The self-citations to [26] supply the prior reachability-based CBF definition, Theorem 2, and the numerical baseline; [26] is an independent earlier preprint, and this paper extends it rather than reducing its central claims to it. The numerical comparison in Table I, with mean deviation 0.0031 against direct computation, provides an external check that the equivariance-inferred values match independent direct values. Overall, no fitted input is relabeled as a prediction and no load-bearing step is equivalent to its own input.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central theorems carry no fitted constants: H_T, B, and C are defined by optimization and set operations. The design degrees of freedom are gamma, T, delta (reachability CBF tuning), and the sets M, eps_M, eps(sigma), delta (Theorem 9 tuning). The paper's premises are mostly inherited from the authors' prior framework [26] (Assumptions 1-2, Lipschitz regularity of H_T), plus new technical conditions (31)-(33) and Assumption 4 that are specific to this paper. The main honesty concern is that (31) and (43)/A4.2 are asserted to hold in the examples without a formal check. No invented physical entities are posited.

free parameters (4)
  • gamma (discount rate in H_T, Eq. (5)) = chosen in [0, delta/T] (design range)
    User-chosen design parameter of the reachability CBF; constrained by Theorem 2, not fitted to data.
  • T (finite horizon, Eq. (4)) = sup over D\F of minimal time to reach F
    Defined by the reachability problem data under Assumption 2; a computation parameter rather than a fitted constant.
  • M and eps_M (partial-knowledge set, Assumption 3) = chosen per example, e.g., x in [0, a+DeltaA], y=0, psi in [0,2pi)
    The subset on which the initial CBF b is computed; its choice controls coverage of the new constraint and the verification burden of (31); DeltaA and grid step are hand-chosen.
  • eps(sigma) and delta in conditions (31)-(33) = chosen small per example
    Neighborhood constants that must satisfy (31)-(33); their verification is geometric in the examples.
assumptions (6)
  • domain assumption Assumption 1: there exists a forward control invariant V subset H with margin delta > 0, and a known closed subset F subset V.
    Inherited from [26]; required for H_T to be a CBF in Theorem 2 and for the symmetry transfer in Theorem 3.
  • domain assumption Assumption 2: controllability on closure of F^c or T-reachability of F from every x0 in D\F.
    Guarantees that T in (4) is finite; inherited from [26].
  • domain assumption Forward completeness of (2) and local Lipschitz continuity of H_T (Theorem 2).
    Stated at the end of Section II ('its forward completeness is assumed') and assumed in Theorem 2 ('If H_T is locally Lipschitz continuous').
  • domain assumption Assumption 3: b is known and locally Lipschitz on M + B_epsM.
    The premise defining 'partial knowledge' in Theorem 9; Section V-B.
  • ad hoc to paper Conditions (31)-(33): b locally maximal in sigma on M_sigma, and the delta/eps neighborhood conditions.
    Load-bearing premises specific to this paper's Theorem 9. Condition (31) is the key inequality that makes the trace construction satisfy the CBF Dini condition; its verification in Examples 3-4 is geometric rather than formal, and the authors note it 'may be non-trivial'.
  • ad hoc to paper Assumption 4: D locally Lipschitz in sigma and S_P admits a Lipschitz single-valued localization.
    Sufficient conditions for B in (34) to be locally Lipschitz (Lemma 10); strong and hard to check in practice, and claimed in Example 3 from smoothness of D alone.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Leveraging Equivariances and Symmetries in the Control Barrier Function Synthesis." pith.science (2026). https://pith.science/paper/AFM5NGQ7

@misc{pith2026250904399,
  author       = {Pith},
  title        = {Pith review of: Leveraging Equivariances and Symmetries in the Control Barrier Function Synthesis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AFM5NGQ7}},
  note         = {Machine review of arXiv:2509.04399}
}
read the original abstract

The synthesis of Control Barrier Functions (CBFs) often involves demanding computations or a meticulous construction. However, structural properties of the system dynamics and constraints have the potential to mitigate these challenges. In this paper, we explore how equivariances in the dynamics, loosely speaking a form of symmetry, can be leveraged in the CBF synthesis. Although CBFs are generally not inherently symmetric, we show how equivariances in the dynamics and symmetries in the constraints induce symmetries in CBFs derived through reachability analysis. This insight allows us to infer their CBF values across the entire domain from their values on a subset, leading to significant computational savings. Interestingly, equivariances can be even leveraged to the CBF synthesis for non-symmetric constraints. Specifically, we show how a partially known CBF can be leveraged together with equivariances to construct a CBF for various new constraints. Throughout the paper, we provide examples illustrating the theoretical findings. Furthermore, a numerical study investigates the computational gains from invoking equivariances into the CBF synthesis.

Figures

Figures reproduced from arXiv: 2509.04399 by the authors.

Figure 1
Figure 1. Illustration of Assumption 1, [26]. The next assumption ensures the existence of a finite time T – sup xPDzF min τě0 τ (4a) s.t. x9ptq “ fpxptq,uptqq (a.e.), (4b) xp0q “ x0, uptq P U, xpτ q P F, (4c) that allows to determine the viability of any state in D by analyzing the system’s behavior over a finite time horizon. Assumption 2. Let either of the following statements hold: A2.1 Dynamics (2) are controllable on th… view at source ↗
Figure 2
Figure 2. Transformation of a point x to set M for an exemplary constraint h defined as the distance to the indicated obstacle. Constraint function h is symmetric with respect to a lateral shift denoted by D. Summarizing (10)-(14), we obtain from (9) the equivalent optimization problem HT pDpx0; pqq – max up¨qPUr0,T s min tPr0,Ts hpxptqq ´ γt s.t. xp0q “ x0, xpxq “ fpxpsq,upsqq (a.e.), upsq P U @s P r0, Ts, xpθq P F for some … view at source ↗
Figure 3
Figure 3. Local symmetry with L being the dotted and D ´1 h pL; pq the gray region. F x0 L L0 hpxq ă 0 φp¨; x0, uq [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (11 more)
Figure 8
Figure 8. Figure 8: Translational equivariance and symmetry: illustration [PITH_FULL_IMAGE:figures/full_fig_p006_8.png]
Figure 9
Figure 9. Figure 9: Rotational equivariance and symmetries: (a) illustra [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]
Figure 10
Figure 10. Figure 10: Equivariance and symmetry with respect to the ori [PITH_FULL_IMAGE:figures/full_fig_p007_10.png]
Figure 13
Figure 13. Figure 13: Conceptual CBF synthesis method via equivariances [PITH_FULL_IMAGE:figures/full_fig_p008_13.png]
Figure 12
Figure 12. Figure 12: Exemplary linear system equivariant with respect to [PITH_FULL_IMAGE:figures/full_fig_p008_12.png]
Figure 14
Figure 14. Figure 14: Illustration of conditions (31)–(33) for [PITH_FULL_IMAGE:figures/full_fig_p010_14.png]
Figure 16
Figure 16. Figure 16: Then (43) prevents a discontinuity on the boundary [PITH_FULL_IMAGE:figures/full_fig_p011_16.png]
Figure 16
Figure 16. Figure 16: The dark blue re￾gion marks those states cov￾ered by Mσ for multiple σ. ppσq ppσ`∆σq npσq npσ`∆σq bσ bσ`∆σ xpos [PITH_FULL_IMAGE:figures/full_fig_p012_16.png]
Figure 18
Figure 18. Figure 18: Construction for a corner. Local Lipschitz continuity of B by Lemma 10, even though M overlaps as it is shifted along the constraint boundary (see dark blue region in [PITH_FULL_IMAGE:figures/full_fig_p013_18.png]
Figure 19
Figure 19. Figure 19: CBFs at orientation ψ “ 0 for various convex obstacles, computed for the bicycle model (vmin ą 0, ζmax “ 20π{180). The red line in (a) marks the explicitly computed values from which all other CBFs are derived. (a) Circle (b) Non-convex obstacle [PITH_FULL_IMAGE:figu…
Figure 20
Figure 20. Figure 20: CBFs at orientation ψ “ 0 for non-convex obstacles, constructed based on the explicitly computed values marked by the red line in (a). VII. CONCLUSION This paper investigated how equivariant dynamics and sym￾metric constraints can be leveraged for CBF synthesis. First…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 32 canonical work pages

  1. [26]

    Predictive synthesis of control barrier functions and its application to time-varying constraints,

    A. Wiltz and D. V . Dimarogonas, “Predictive synthesis of control barrier functions and its application to time-varying constraints,” 2025. [Online]. Available: https://arxiv.org/abs/2504.15830

  2. [1]

    A ‘universal’ construction of artstein’s theorem on nonlinear stabilization,

    E. D. Sontag, “A ‘universal’ construction of artstein’s theorem on nonlinear stabilization,” Systems & control letters , vol. 13, no. 2, pp. 117–123, 1989

  3. [2]

    Constructive safety using control barrier functions,

    P. Wieland and F. Allgöwer, “Constructive safety using control barrier functions,” IFAC Proceedings Volumes, vol. 40, no. 12, pp. 462–467, 2007, 7th IFAC Symposium on Nonlinear Control Systems

  4. [3]

    Control barrier function based quadratic programs for safety critical systems,

    A. D. Ames, X. Xu, J. W. Grizzle, and P. Tabuada, “Control barrier function based quadratic programs for safety critical systems,” IEEE Transactions on Automatic Control, vol. 62, no. 8, pp. 3861–3876, 2017

  5. [4]

    Control barrier functions for signal temporal logic tasks,

    L. Lindemann and D. V . Dimarogonas, “Control barrier functions for signal temporal logic tasks,” IEEE Control Systems Letters, vol. 3, no. 1, pp. 96–101, 2019

  6. [5]

    Handling disjunctions in signal temporal logic based control through nonsmooth barrier functions,

    A. Wiltz and D. V . Dimarogonas, “Handling disjunctions in signal temporal logic based control through nonsmooth barrier functions,” in 2022 IEEE 61st Conference on Decision and Control (CDC) , 2022, pp. 3237–3242

  7. [6]

    Mutualistic interactions in heterogeneous multi-agent systems,

    A. A. Nguyen, F. Jabbari, and M. Egerstedt, “Mutualistic interactions in heterogeneous multi-agent systems,” in 2023 62nd IEEE Conference on Decision and Control (CDC) , 2023, pp. 411–418

  8. [7]

    Decentralized ve- hicle coordination and lane switching without switching of controllers,

    A. Frauenfelder, A. Wiltz, and D. V . Dimarogonas, “Decentralized ve- hicle coordination and lane switching without switching of controllers,” IFAC-PapersOnLine, vol. 56, no. 2, pp. 3334–3339, 2023

Show all 39 references
  1. [8]

    Collision avoidance and geofencing for fixed-wing aircraft with control barrier functions,

    T. G. Molnar, S. K. Kannan, J. Cunningham, K. Dunlap, K. L. Hobbs, and A. D. Ames, “Collision avoidance and geofencing for fixed-wing aircraft with control barrier functions,” IEEE Transactions on Control Systems Technology, pp. 1–16, 2025

  2. [9]

    Contact-aware safety in soft robots using high-order control barrier and lyapunov functions,

    K. Wong, M. Stölzle, W. Xiao, C. D. Santina, D. Rus, and G. Zardini, “Contact-aware safety in soft robots using high-order control barrier and lyapunov functions,” 2025. [Online]. Available: https://arxiv.org/abs/2505.03841

  3. [10]

    J. M. Selig, Geometric fundamentals of robotics . Springer Science & Business Media, 2007

  4. [11]

    Morphological sym- metries in robotics,

    D. O. Apraez, G. Turrisi, V . Kostic, M. Martin, A. Agudo, F. Moreno- Noguer, M. Pontil, C. Semini, and C. Mastalli, “Morphological sym- metries in robotics,” The International Journal of Robotics Research , 2025

  5. [12]

    Observer design for nonlinear systems with equivariance,

    R. Mahony, P. van Goor, and T. Hamel, “Observer design for nonlinear systems with equivariance,” Annu. Rev. Control Robot. Auton. Syst. , vol. 5, no. 1, pp. 221–252, May 2022

  6. [13]

    Equivariant dynamical systems,

    M. Field, “Equivariant dynamical systems,” Bulletin of the American Mathematical Society, vol. 76, no. 6, pp. 1314–1318, 1970

  7. [14]

    Equivariant dynamical systems,

    ——, “Equivariant dynamical systems,” Transactions of the American Mathematical Society, vol. 259, no. 1, pp. 185–205, 1980

  8. [15]

    Symmetry and symmetry-breaking bifurcations in fluid dynamics,

    J. D. Crawford and E. Knobloch, “Symmetry and symmetry-breaking bifurcations in fluid dynamics,” Annual Review of Fluid Mechanics , vol. 23, no. V olume 23, 1991, pp. 341–387, 1991

  9. [16]

    Chossat and R

    P. Chossat and R. Lauterbach, Methods in Equivariant Bifurcations and Dynamical Systems. WORLD SCIENTIFIC, 2000

  10. [17]

    Symmetry invariance for adapting biological systems,

    O. Shoval, U. Alon, and E. Sontag, “Symmetry invariance for adapting biological systems,” SIAM Journal on Applied Dynamical Systems , vol. 10, no. 3, pp. 857–886, 2011. 15

  11. [18]

    Dynamic compensation, parameter identifiability, and equivariances,

    E. D. Sontag, “Dynamic compensation, parameter identifiability, and equivariances,” PLOS Computational Biology , vol. 13, no. 4, pp. 1–17, 2017

  12. [19]

    Understanding image representations by measuring their equivariance and equivalence,

    K. Lenc and A. Vedaldi, “Understanding image representations by measuring their equivariance and equivalence,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2015

  13. [20]

    Spherical cnns,

    T. S. Cohen, M. Geiger, J. Koehler, and M. Welling, “Spherical cnns,”

  14. [21]

    Generalizing convolutional neural networks for equivariance to lie groups on arbitrary continuous data,

    M. Finzi, S. Stanton, P. Izmailov, and A. G. Wilson, “Generalizing convolutional neural networks for equivariance to lie groups on arbitrary continuous data,” in Proceedings of the 37th International Conference on Machine Learning , 2020, pp. 3165–3176

  15. [22]

    Equivariant systems theory and observer design,

    R. Mahony, T. Hamel, and J. Trumpf, “Equivariant systems theory and observer design,” 2020. [Online]. Available: https: //arxiv.org/abs/2006.08276

  16. [23]

    Equivariant filter (eqf),

    P. van Goor, T. Hamel, and R. Mahony, “Equivariant filter (eqf),” IEEE Transactions on Automatic Control, vol. 68, no. 6, pp. 3501–3512, 2023

  17. [24]

    Koopman operator methods for global phase space exploration of equivariant dynamical systems,

    S. Sinha, S. P. Nandanoori, and E. Yeung, “Koopman operator methods for global phase space exploration of equivariant dynamical systems,” IFAC-PapersOnLine, vol. 53, no. 2, pp. 1150–1155, 2020

  18. [25]

    Deep equivariant multi-agent control barrier functions,

    N. Bousias, L. Lindemann, and G. Pappas, “Deep equivariant multi-agent control barrier functions,” 2025. [Online]. Available: https://arxiv.org/abs/2506.07755

  19. [27]

    Nonlinear controllability and observability,

    R. Hermann and A. Krener, “Nonlinear controllability and observability,” IEEE Transactions on Automatic Control , vol. 22, no. 5, pp. 728–740, 1977

  20. [28]

    Asymptotic stability and feedback stabilization,

    R. W. Brockett, “Asymptotic stability and feedback stabilization,” Dif- ferential geometric control theory , vol. 27, no. 1, pp. 181–191, 1983

  21. [29]

    Lyapunov functions and discontinuous stabilizing feedback,

    F. Clarke, “Lyapunov functions and discontinuous stabilizing feedback,” Annual Reviews in Control , vol. 35, no. 1, pp. 13–33, 2011

  22. [30]

    A nonsmooth approach to controller synthesis for boolean specifications,

    P. Glotfelter, J. Cortes, and M. Egerstedt, “A nonsmooth approach to controller synthesis for boolean specifications,” IEEE Transactions on Automatic Control, pp. 5160–5174, 2020

  23. [31]

    Hybrid control barrier functions for continuous-time systems,

    M. Marley, R. Skjetne, and A. R. Teel, “Hybrid control barrier functions for continuous-time systems,” IEEE Transactions on Automatic Control, vol. 69, no. 10, pp. 6605–6619, 2024

  24. [32]

    Control barrier functions for disjunctions of signal temporal logic tasks,

    M. Charitidou and D. V . Dimarogonas, “Control barrier functions for disjunctions of signal temporal logic tasks,” in 2023 European Control Conference (ECC), 2023, pp. 1–6

  25. [33]

    On reachability and minimum cost optimal control,

    J. Lygeros, “On reachability and minimum cost optimal control,” Auto- matica, vol. 40, no. 6, pp. 917–927, 2004

  26. [34]

    Robust control barrier–value functions for safety-critical control,

    J. J. Choi, D. Lee, K. Sreenath, C. J. Tomlin, and S. L. Herbert, “Robust control barrier–value functions for safety-critical control,” in 2021 60th IEEE Conference on Decision and Control (CDC), 2021, pp. 6814–6821

  27. [35]

    From time-invariant to uniformly time-varying control barrier functions: A constructive approach,

    A. Wiltz and D. V . Dimarogonas, “From time-invariant to uniformly time-varying control barrier functions: A constructive approach,” in2024 IEEE 63rd Conference on Decision and Control (CDC) , 2024

  28. [36]

    Trajectory planning for a four-wheel-steering vehicle,

    D. Wang and F. Qi, “Trajectory planning for a four-wheel-steering vehicle,” in Proceedings 2001 ICRA. IEEE International Conference on Robotics and Automation , vol. 4, 2001, pp. 3320–3325

  29. [37]

    R. T. Rockafellar, Variational analysis, 1st ed., ser. Grundlehren der mathematischen Wissenschaften. Springer Berlin, Heidelberg, 1998

  30. [38]

    CasADi – A software framework for nonlinear optimization and opti- mal control,

    J. A. E. Andersson, J. Gillis, G. Horn, J. B. Rawlings, and M. Diehl, “CasADi – A software framework for nonlinear optimization and opti- mal control,” Mathematical Programming Computation , vol. 11, no. 1, pp. 1–36, 2019

  31. [2018]

    Available: https://arxiv.org/abs/1801.10130

    [Online]. Available: https://arxiv.org/abs/1801.10130

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.