Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

End-to-End Learning of Safe Optimal Feedback Control in High Dimensions with Control Barrier Function Layers

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read This paper claims that CBF safety filters can be learned end-to-end at 1,200 state dimensions by replacing exact QP differentiation with operator splitting and Jacobian-free gradients.

desk verdict The paper's scaling result is the real contribution and the DYS contraction analysis is a genuine advance, but the convergence theorem rests on an unverified coupling assumption and is proved only for gradient flow, not the Adam used in practice. read the letter →

arxiv 2607.20674 v1 pith:4HJ32QRR submitted 2026-07-22 cs.LG cs.SYeess.SYmath.OC

Xingjian Li , Kelvin Kan , Deepanshu Verma , Krishna Kumar , Stanley Osher , Samy Wu Fung This is my paper · ORCID
classification cs.LGcs.SYeess.SYmath.OC
keywords controlbarrierfunctionssafetyfiltersend-to-endlearningJacobian-freebackpropagationDavis-YinsplittingnonsmoothoptimizationClarkestationaritymulti-agent
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 tries to show that hard safety constraints can be folded into neural feedback-policy training at scales previously out of reach. Prior end-to-end methods that differentiate through CBF-based quadratic-program safety filters were limited to about 16 state dimensions, while the authors report state dimensions up to 1,200 (100 quadcopters) and control dimensions up to 400. Their recipe combines an operator-splitting solver (Davis-Yin splitting) with an approximate gradient (Jacobian-Free Backpropagation), which avoids both unrolling many solver iterations and solving the large linear systems of implicit differentiation. The paper also provides a nonsmooth-analysis proof that the resulting biased gradient update is a descent direction and converges to a Clarke stationary point under gradient flow. If correct, this makes safety-critical end-to-end control learning practical in a regime where exact-gradient methods run out of memory or blow up.

What carries the argument

The central object is the Davis-Yin splitting fixed-point operator T_theta for the lifted CBF-QP projection (control u plus slack s), whose fixed point gives the safety-filtered control. DYS decomposes the projection into simple closed-form projections and a gradient step, making each iteration cheap. The second mechanism is Jacobian-Free Backpropagation: instead of computing (I - dT/dy)^{-1} dT/dtheta, it uses only dT/dtheta, cutting per-step gradient cost from O((m+c)^3 + (m+c)^2 p) to O((m+c)p). Clarke generalized Jacobians handle the nonsmoothness of the projection, letting the proof show every generalized Jacobian of T_theta is a contraction and that the JFB update is a descent directio

What would settle it

Compute the norm of the gradient-mismatch matrix Xi_theta defined in Assumption 6(iii) for a trained policy at a fixed point where safety constraints are active; if it exceeds the threshold rho_theta, the pointwise alignment lemma and the descent theorem are violated.

Watch

Extended reading notes

Core claim

The authors claim that end-to-end training of neural feedback controllers with embedded control-barrier-function safety filters, previously limited to roughly 16 state dimensions, can be pushed to at least 1,200 state and 400 control dimensions without giving up the hard safety guarantee h(z(t)) >= 0. The route is to solve the CBF quadratic program by Davis-Yin splitting, a three-operator fixed-point iteration whose steps are cheap projections, and to differentiate through the fixed point with Jacobian-Free Backpropagation, which replaces the exact fixed-point derivative with a single partial derivative. They further claim convergence of the resulting trajectory-level training scheme: under

Load-bearing premise

The load-bearing premise is that the slack variables introduced to convert the barrier inequality into an equality do not distort the approximate gradient too much; the authors state this condition is generally intractable to verify analytically and support it only with numerical experiments, so the descent and convergence theorems stand or fall on that alignment.

Editorial extensions

If this is right

  • The reported ceiling of end-to-end CBF-QP policy learning moves from about 16 state dimensions to 1,200 state and 400 control dimensions, demonstrated on single-integrator, double-integrator, and quadcopter multi-agent tasks.
  • Safety remains hard by construction: the embedded filter enforces h(z(t)) >= 0 at every rollout, and the reported minimum barrier values stay positive.
  • Training memory is cut by up to roughly 10x compared with automatic-differentiation unrolling and implicit-differentiation baselines, and per-epoch time is lower whenever the baseline runs.
  • The convergence proof gives a theoretical foundation for training through nonsmooth safety layers, establishing convergence to a Clarke stationary point under the JFB gradient flow.
  • Because the forward solver and the differentiation rule are decoupled, any algorithm that computes the CBF-QP solution can be used in the forward pass while retaining the JFB backward pass and its guarantees.

Reading between the lines

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

  • Inference: The convergence guarantee is for continuous-time gradient flow, not for the discrete Adam updates actually used in the experiments; a discrete-step version is a natural next step and is not supplied by the paper.
  • Inference: If the slack-variable alignment condition (Assumption 6(iii)) could be checked cheaply online, the method could be extended to adaptive step sizes or to detect when JFB's biased gradient stops being a safe descent direction.
  • Inference: The reported scalability still relies on fixed-point iterations per time step; for very large agent counts or tight horizons, the number of iterations, rather than memory, may become the limiting factor.
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

3 major / 5 minor

Summary. The paper proposes a scalable end-to-end training framework for neural feedback controllers with embedded CBF-QP safety filters. The method combines Davis–Yin splitting (DYS) for the forward safety projection with Jacobian-Free Backpropagation (JFB) for the backward pass, and is demonstrated on multi-agent single-integrator, double-integrator, and quadcopter problems with state dimensions up to 1200 and control dimensions up to 400. The theoretical contribution claims convergence of the training update to a Clarke stationary point under nonsmoothness induced by the safety layer, with the main result stated as Theorem 12. Numerical comparisons against AD-unrolled DYS and CVXPY Layers report lower memory usage, competitive costs, and stability in regimes where the baselines diverge.

Significance. If the claims hold, the paper addresses a real bottleneck: prior end-to-end training through differentiable CBF-QP layers has been restricted to small state dimensions, while this work demonstrates scaling to 1200-dimensional problems. The contraction analysis for every Clarke selection of the DYS operator (Corollary 9 and Theorem 19) is a solid technical contribution, and the memory/time comparisons in Tables 2 and Figure 5 support the scalability story. However, the convergence theory is significantly weaker than the abstract implies: the main theorem concerns continuous-time gradient flow under assumptions that are partly unverified, and the algorithm actually executed is discrete Adam. The paper is therefore valuable as an empirical scaling contribution with a partial theory, but the theoretical claims need to be reframed or substantially strengthened.

major comments (3)
  1. [Assumption 6(iii), Eqs. (75)–(76), Lemma 22 and Remark] Assumption 6(iii) is the load-bearing hypothesis of the descent argument. The bound on ||Ξ_θ|| in Eq. (76) is used in Lemma 22 to prove the pointwise alignment ⟨v,w⟩ ≥ 0, and Lemmas 10, Theorem 11, and Theorem 12 all inherit it. The appendix remark after Lemma 22 states that this condition is 'generally intractable to verify analytically' and cites Figure 4a as empirical evidence. However, Figure 4a plots JFB/AD gradient alignment during training; it does not measure Ξ_θ as defined in Eq. (75), nor does it compare it to the threshold ρ_θ in Eq. (76). Because the assumption is essentially the statement that the gradient mismatch is below the value needed for the proof, the convergence guarantee is conditional on an unverified condition that is close in content to the desired conclusion. The manuscript should either provide a direct numerical certification of (75)–(76) along actual trainin
  2. [Theorem 12 vs. Algorithm 1 and Section 'Implementation Details'] Theorem 12 proves convergence for the continuous-time gradient flow (Eq. (28)), but Algorithm 1 and all experiments use discrete Adam updates (line 23 of Algorithm 1; 'We use the Adam optimizer' in the Implementation Details). No argument is given to connect the gradient-flow result to the discrete stochastic optimizer actually used. The abstract and the contribution list state that the training update converges, which is misleading. The paper does note the gradient-flow limitation in a remark after Theorem 12, but the main text should be corrected to state clearly that the convergence guarantee does not apply to Algorithm 1, or the theory should be extended to a discrete-time setting.
  3. [Assumptions 7 and 8; Figure 4b] The descent theorem also depends on Assumptions 7 and 8, which are stated as regularity/variance bounds but are not independently verified. Assumption 8.2 in particular bounds deviations in terms of inf_{φ∈∂Cθ E_x[J_x]} ||φ|| and ||E_x[d^JFB_x]||, and the threshold δ_{v,θ} itself depends on ρ_θ − ||Ξ_θ|| from the unverified Assumption 6(iii). Figure 4b shows that time averages of gradients remain bounded, but this is far weaker than the specific inequalities required in Assumptions 7 and 8. The claim that Figure 4 'supports' these assumptions should be moderated; at present the proof is conditional on several assumptions that are not checked quantitatively.
minor comments (5)
  1. [Figure 4a caption] The caption says the plot serves as a 'direct verification of Theorem 11.' Since the plot only measures gradient alignment on one training run and does not verify Assumptions 6–8, please rephrase to 'consistent with Theorem 11' or 'suggests descent direction is satisfied in practice.'
  2. [Appendix, 'Additional Numerical Results'] The sentence 'We also tested multiple learning rates ranging from0.01and1→, es10 −5' is garbled. Please correct the typographical error and state the exact learning-rate range tested.
  3. [Tables 1 and 4] Table 1 reports point estimates for the baselines without standard errors or confidence intervals. Table 4 provides multi-seed statistics only for the proposed DYS-JFB method and only for three tasks. Adding repeated-seed results for the baselines where they converge would make the 'competitive performance' claim more supportable.
  4. [Algorithm 1, line 12] The 'stop-gradient' annotation is ambiguous. Please clarify that it applies to the DYS fixed-point iterations used to compute the forward solution, while the final DYS evaluation in line 16 enables gradient flow.
  5. [Discussion, 'Key advantages'] The claim that DYS 'requires no problem-dependent step-size tuning' should be phrased more carefully: the contraction result holds for any ζ∈(0,1), but the experiments fix ζ=0.5. This is a convenience, not a complete absence of hyperparameters.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular reduction found; the convergence proof is conditional on an explicit unverified assumption (6(iii)), which is a verification gap rather than circularity.

full rationale

The paper's derivation chain is self-contained up to its stated assumptions. The DYS contraction result (Corollary 9) is proved from Lemmas 14–19 using linear algebra and projection properties, without presupposing descent or convergence. The descent analysis derives an explicit lower bound for pointwise alignment, and Assumption 6(iii) is then invoked as a sufficient condition: Lemma 22 obtains ⟨v,w⟩ ≥ 0 from the bound ∥Ξθ∥2 ≤ ρθ, and the paper admits this condition is 'generally intractable to verify analytically for a given problem' and points to Figure 4a. That figure actually measures JFB/AD alignment rather than Ξθ directly, so the support for Assumption 6(iii) is weaker than claimed; however, this is an unverified hypothesis, not a fitted parameter renamed as a prediction and not a definitional equivalence with the desired conclusion. Assumptions 7 and 8 are similarly explicit variance/regularity bounds; their constants reference earlier quantities (e.g., ρθ and ∥Ξθ∥2), which is self-consistency of assumptions, not circularity. The paper cites prior JFB/DYS work by overlapping authors, but these citations supply the fixed-point differentiation machinery and standard conditioning assumptions rather than an external uniqueness theorem forcing the method. The central scalability claim is independently supported by comparative runtime/memory tables, trajectory rollouts, and barrier-value checks, so it is not forced by construction. Overall, no load-bearing circular step is exhibited; score 2 reflects the self-citation density and the conditional, hard-to-verify nature of Assumption 6(iii), not circular reasoning.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

No new physical entities are introduced. The main uncharged imports are strong, partly unverifiable regularity/coupling assumptions that make the convergence theorem go through; these function as ad hoc axioms rather than fitted parameters.

free parameters (3)
  • Terminal cost weight schedule (ω and continuation strategy) = not reported
    The authors state 'we omit the weighting coefficients used in practice' and use a continuation strategy to stabilize training; this choice affects all numerical results and is not specified enough to reproduce.
  • DYS step size ζ = 0.5
    Fixed for all experiments; the theory only requires ζ∈(0,1), so it is not load-bearing for the proof, but it affects contraction speed and practical convergence.
  • Obstacle radii and CBF/HOCBF class-K functions = per-problem values, e.g., 0.3, 0.35, 0.5, 0.7
    These define the safety filter and benchmark difficulty; they are chosen per problem and partly shape the reported performance.
assumptions (6)
  • domain assumption A(z) has full row rank for all z (Assumption 3)
    Used in Lemmas 17-19 to prove DYS contraction. Not verified for the multi-agent obstacle configurations; redundant constraints or c>m would violate it.
  • domain assumption Compact domains for x, z and θ (Assumption 4)
    Needed to obtain a uniform contraction constant γ<1 in Corollary 9. Control state spaces with target-reaching objectives are typically unbounded unless artificially restricted.
  • domain assumption Tθ is C^1 in (θ,z) and Lipschitz in (y,z), with integrability and smoothness of trajectory costs (Assumption 5)
    The projection P_C1 is nonsmooth in y, but C^1 in θ,z may fail when active constraint sets change discontinuously with z; the authors state this 'generally holds' but do not analyze the CBF case in detail.
  • ad hoc to paper Weak-coupling bound on slack-induced gradient mismatch (Assumption 6(iii))
    The condition ||Ξθ||≤ρθ is exactly what makes the JFB update a descent direction (Lemma 22). The authors admit it is 'generally intractable to verify analytically' and only cite empirical Figure 4a.
  • ad hoc to paper Variance and deviation bounds (Assumptions 7 and 8)
    These are complex, partly self-referential bounds needed for Lemmas 10 and Theorem 11. They are not verified analytically; Assumption 8 involves δ_{v,θ} defined through the same quantities being bounded.
  • domain assumption Forward invariance theorem for CBFs applies to the discretized piecewise-constant control used in experiments
    The hard-safety claim is evaluated at RK4 sample points, but the CBF forward-invariance theorem assumes continuous-time locally Lipschitz u(t); safety between sample points is not guaranteed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of End-to-End Learning of Safe Optimal Feedback Control in High Dimensions with Control Barrier Function Layers." pith.science (2026). https://pith.science/paper/4HJ32QRR

@misc{pith2026260720674,
  author       = {Pith},
  title        = {Pith review of: End-to-End Learning of Safe Optimal Feedback Control in High Dimensions with Control Barrier Function Layers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4HJ32QRR}},
  note         = {Machine review of arXiv:2607.20674}
}
read the original abstract

We consider the problem of learning high-dimensional semi-global feedback controllers under hard safety constraints enforced by control barrier functions (CBFs). Incorporating CBFs into end-to-end policy training requires embedding a quadratic-program-based safety filter as an optimization layer, but computational and differentiation bottlenecks have largely restricted prior approaches to low-dimensional systems, typically with at most 16 state dimensions. We address this limitation by combining operator splitting with the recently developed Jacobian-Free Backpropagation (JFB) method to enable scalable end-to-end training while preserving hard safety guarantees through the CBF safety filter. We justify this training methodology theoretically using nonsmooth analysis techniques and demonstrate its effectiveness on high-dimensional multi-agent nonlinear control problems with state and control dimensions up to 1200 and 400, respectively.

Figures

Figures reproduced from arXiv: 2607.20674 by the authors.

Figure 1
Figure 1. High-dimensional control examples: trajectories of 50 single integrators (left) and 100 quadcopters (right) generated by trained semi-global feedback controllers. (QP) solved at each time step, which minimally modifies a nominal control input to ensure safety. For neural feedback policies, this naturally leads to a QP-based safety filter embedded as an optimization layer within the training loop. This end-to-end for… view at source ↗
Figure 2
Figure 2. Numerical results for six control scenarios. In each case, the feedback controller is trained in environments with [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Evaluation of barrier functions and projection activations on trained models for each example. Here safety is ensured [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Numerical verification of the theoretical results and key assumptions. [PITH_FULL_IMAGE:figures/full_fig_p029_4.png]
Figure 5
Figure 5. Figure 5: Per epoch training time results compared across all tested method under three different dynamics models. In all [PITH_FULL_IMAGE:figures/full_fig_p032_5.png]
Figure 6
Figure 6. Figure 6: Controlled trajectories using the learned feedback controller over different initial conditions and obstacles shapes, our [PITH_FULL_IMAGE:figures/full_fig_p034_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Truncated Differentiation Through Primal-Dual Solvers for Inverse Potential Mean-Field Games

    math.OC 2026-07 conditional novelty 6.0 of 10

    Truncated backpropagation through the final r PDHG iterations (JFB-r) gives a surrogate-gradient view, converges to the implicit gradient under fixed active-set/exact-detach conditions, and reduces memory/runtime in i...

Reference graph

Works this paper leans on

97 extracted references · 1 canonical work pages · cited by 1 Pith paper

  1. [1]

    International conference on machine learning , pages=

    Optnet: Differentiable optimization as a layer in neural networks , author=. International conference on machine learning , pages=. 2017 , organization=

    2017

  2. [2]

    Greenwade

    George D. Greenwade. The C omprehensive T ex A rchive N etwork ( CTAN ). TUGBoat. 1993

    1993

  3. [3]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Jfb: Jacobian-free backpropagation for implicit networks , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  4. [4]

    SIAM Journal on Mathematics of Data Science , volume=

    Implicit deep learning , author=. SIAM Journal on Mathematics of Data Science , volume=. 2021 , publisher=

    2021

  5. [5]

    2019 IEEE 58th conference on decision and control (CDC) , pages=

    Control barrier functions for systems with high relative degree , author=. 2019 IEEE 58th conference on decision and control (CDC) , pages=. 2019 , organization=

    2019

  6. [6]

    IEEE Transactions on Automatic Control , volume=

    Control barrier function based quadratic programs for safety critical systems , author=. IEEE Transactions on Automatic Control , volume=. 2016 , publisher=

    2016

  7. [7]

    IEEE Transactions on Robotics , volume=

    Safe control with learned certificates: A survey of neural lyapunov, barrier, and contraction methods for robotics and control , author=. IEEE Transactions on Robotics , volume=. 2023 , publisher=

    2023

  8. [8]

    Conference on Robot Learning , pages=

    Safe control under input limits with neural control barrier functions , author=. Conference on Robot Learning , pages=. 2023 , organization=

    2023

Show all 97 references
  1. [9]

    Conference on robot learning , pages=

    Neural graph control barrier functions guided distributed collision-avoidance multi-agent control , author=. Conference on robot learning , pages=. 2023 , organization=

    2023

  2. [10]

    2019 18th European control conference (ECC) , pages=

    Control barrier functions: Theory and applications , author=. 2019 18th European control conference (ECC) , pages=. 2019 , organization=

    2019

  3. [11]

    2006 , publisher=

    Numerical optimization , author=. 2006 , publisher=

    2006

  4. [12]

    Transactions on Machine Learning Research , year =

    Differentiating Through Integer Linear Programs with Quadratic Regularization and Davis-Yin Splitting , author=. Transactions on Machine Learning Research , year =

  5. [13]

    1990 , publisher=

    Optimization and nonsmooth analysis , author=. 1990 , publisher=

    1990

  6. [14]

    arXiv preprint arXiv:2509.18404 , year=

    Zero-Shot Transferable Solution Method for Parametric Optimal Control Problems , author=. arXiv preprint arXiv:2509.18404 , year=

  7. [15]

    Set-valued and variational analysis , volume=

    A three-operator splitting scheme and its optimization applications , author=. Set-valued and variational analysis , volume=. 2017 , publisher=

    2017

  8. [16]

    Evans, Lawrence C , year =

  9. [17]

    2010 , publisher=

    Partial Differential Equations , author=. 2010 , publisher=

    2010

  10. [18]

    arXiv preprint arXiv:2602.00921 , year=

    On the Convergence of Jacobian-Free Backpropagation for Optimal Control Problems with Implicit Hamiltonians , author=. arXiv preprint arXiv:2602.00921 , year=

    arXiv

  11. [19]

    1988 , publisher=

    Real Analysis , author=. 1988 , publisher=

    1988

  12. [20]

    2000 , publisher=

    Introduction to real analysis , author=. 2000 , publisher=

    2000

  13. [21]

    Journal of Process Control , volume=

    Differentiable predictive control: Deep learning alternative to explicit model predictive control for unknown nonlinear systems , author=. Journal of Process Control , volume=. 2022 , publisher=

    2022

  14. [22]

    IEEE Transactions on Systems, Man, and Cybernetics: Systems , volume=

    Learning constrained parametric differentiable predictive control policies with guarantees , author=. IEEE Transactions on Systems, Man, and Cybernetics: Systems , volume=. 2024 , publisher=

    2024

  15. [23]

    2022 IEEE 61st Conference on Decision and Control (CDC) , pages=

    Differentiable predictive control with safety guarantees: A control barrier function approach , author=. 2022 IEEE 61st Conference on Decision and Control (CDC) , pages=. 2022 , organization=

    2022

  16. [24]

    2024 IEEE International Conference on Robotics and Automation (ICRA) , pages=

    How to train your neural control barrier function: Learning safety filters for complex input-constrained systems , author=. 2024 IEEE International Conference on Robotics and Automation (ICRA) , pages=. 2024 , organization=

    2024

  17. [25]

    Advances in neural information processing systems , volume=

    Differentiable mpc for end-to-end planning and control , author=. Advances in neural information processing systems , volume=

  18. [26]

    arXiv preprint arXiv:2308.08426 , year=

    Differentiable robust model predictive control , author=. arXiv preprint arXiv:2308.08426 , year=

    arXiv

  19. [27]

    Advances in Neural Information Processing Systems , volume=

    Pontryagin differentiable programming: An end-to-end learning and control framework , author=. Advances in Neural Information Processing Systems , volume=

  20. [28]

    Advances in Neural Information Processing Systems , volume=

    Safe pontryagin differentiable programming , author=. Advances in Neural Information Processing Systems , volume=

  21. [29]

    Advances in neural information processing systems , volume=

    Differentiable convex optimization layers , author=. Advances in neural information processing systems , volume=

  22. [30]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=

    Deep declarative networks , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=. 2021 , publisher=

    2021

  23. [31]

    arXiv preprint arXiv:2410.10807 , year=

    Hardnet: Hard-constrained neural networks with universal approximation guarantees , author=. arXiv preprint arXiv:2410.10807 , year=

  24. [32]

    Advances in neural information processing systems , volume=

    Deep equilibrium models , author=. Advances in neural information processing systems , volume=

  25. [33]

    Proceedings of the Twelfth ACM International Conference on Future Energy Systems , pages=

    Enforcing policy feasibility constraints through differentiable projection for energy optimization , author=. Proceedings of the Twelfth ACM International Conference on Future Energy Systems , pages=

  26. [34]

    International conference on artificial intelligence and statistics , pages=

    Optimizing millions of hyperparameters by implicit differentiation , author=. International conference on artificial intelligence and statistics , pages=. 2020 , organization=

    2020

  27. [35]

    2022 , publisher=

    Large-scale convex optimization: algorithms & analyses via monotone operators , author=. 2022 , publisher=

    2022

  28. [36]

    arXiv preprint arXiv:2104.12225 , year=

    DC3: A learning method for optimization with hard constraints , author=. arXiv preprint arXiv:2104.12225 , year=

    arXiv

  29. [37]

    arXiv preprint arXiv:2506.07003 , year=

    End-to-End Probabilistic Framework for Learning with Hard Constraints , author=. arXiv preprint arXiv:2506.07003 , year=

  30. [38]

    arXiv preprint arXiv:2207.08675 , year=

    Learning differentiable solvers for systems with hard constraints , author=. arXiv preprint arXiv:2207.08675 , year=

    arXiv

  31. [39]

    arXiv preprint arXiv:2410.23667 , year=

    Projected neural differential equations for learning constrained dynamics , author=. arXiv preprint arXiv:2410.23667 , year=

    arXiv

  32. [40]

    Automatica , volume=

    A predictive safety filter for learning-based control of constrained nonlinear dynamical systems , author=. Automatica , volume=. 2021 , publisher=

    2021

  33. [41]

    arXiv preprint arXiv:2004.07584 , year=

    Reinforcement learning for safety-critical control under model uncertainty, using control lyapunov functions and control barrier functions , author=. arXiv preprint arXiv:2004.07584 , year=

    2004 arXiv

  34. [42]

    2025 American Control Conference (ACC) , pages=

    Kernel expansions for high-dimensional mean-field control with non-local interactions , author=. 2025 American Control Conference (ACC) , pages=. 2025 , organization=

    2025

  35. [43]

    SIAM Journal on Scientific Computing , volume=

    A neural network approach for stochastic optimal control , author=. SIAM Journal on Scientific Computing , volume=. 2024 , publisher=

    2024

  36. [44]

    Journal of Computational Physics , volume=

    Random features for high-dimensional nonlocal mean-field games , author=. Journal of Computational Physics , volume=. 2022 , publisher=

    2022

  37. [45]

    2022 , author=

    Learning to optimize: Where deep learning meets optimization and inverse problems , journal=. 2022 , author=

    2022

  38. [46]

    IEEE Control Systems Letters , year=

    Mean-field control barrier functions for stochastic multi-agent systems , author=. IEEE Control Systems Letters , year=

  39. [47]

    2025 American Control Conference (ACC) , pages=

    Mean-field control barrier functions: A framework for real-time swarm control , author=. 2025 American Control Conference (ACC) , pages=. 2025 , organization=

    2025

  40. [48]

    Communications on Applied Mathematics and Computation , pages=

    Recent advances in numerical solutions for Hamilton-Jacobi PDEs , author=. Communications on Applied Mathematics and Computation , pages=. 2026 , publisher=

    2026

  41. [49]

    Scientific reports , volume=

    Taming hyperparameter tuning in continuous normalizing flows using the JKO scheme , author=. Scientific reports , volume=. 2023 , publisher=

    2023

  42. [50]

    Proceedings of the National Academy of Sciences , volume=

    Alternating the population and control neural networks to solve high-dimensional stochastic mean-field games , author=. Proceedings of the National Academy of Sciences , volume=. 2021 , publisher=

    2021

  43. [51]

    Proceedings of the National Academy of Sciences , volume=

    A machine learning framework for solving high-dimensional mean field game and mean field control problems , author=. Proceedings of the National Academy of Sciences , volume=. 2020 , publisher=

    2020

  44. [52]

    Neurocomputing , pages=

    A generalization bound for a family of implicit networks , author=. Neurocomputing , pages=. 2026 , publisher=

    2026

  45. [53]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    On logical extrapolation for mazes with recurrent and implicit networks , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  46. [54]

    Fixed Point Theory and Algorithms for Sciences and Engineering , volume=

    Feasibility-based fixed point networks , author=. Fixed Point Theory and Algorithms for Sciences and Engineering , volume=. 2021 , publisher=

    2021

  47. [55]

    Scientific Reports , volume=

    Explainable AI via learning to optimize , author=. Scientific Reports , volume=. 2023 , publisher=

    2023

  48. [56]

    Learning for dynamics and control , pages=

    Learning for safety-critical control with control barrier functions , author=. Learning for dynamics and control , pages=. 2020 , organization=

    2020

  49. [57]

    arXiv preprint arXiv:1904.09043 , year=

    Differentiating through a cone program , author=. arXiv preprint arXiv:1904.09043 , year=

    1904 arXiv

  50. [58]

    Journal of Machine Learning Research , volume=

    CVXPY: A Python-embedded modeling language for convex optimization , author=. Journal of Machine Learning Research , volume=

  51. [59]

    IFAC-PapersOnLine , volume=

    Safeguarding autonomous systems: emerging approaches, assumptions and metrics-a systematic literature review , author=. IFAC-PapersOnLine , volume=. 2022 , publisher=

    2022

  52. [60]

    Annual Review of Control, Robotics, and Autonomous Systems , volume=

    The safety filter: A unified view of safety-critical control in autonomous systems , author=. Annual Review of Control, Robotics, and Autonomous Systems , volume=. 2023 , publisher=

    2023

  53. [61]

    Annual Review of Control, Robotics, and Autonomous Systems , volume=

    Planning and decision-making for autonomous vehicles , author=. Annual Review of Control, Robotics, and Autonomous Systems , volume=. 2018 , publisher=

    2018

  54. [62]

    Annual Review of Control, Robotics, and Autonomous Systems , volume=

    Safe learning in robotics: From learning-based control to safe reinforcement learning , author=. Annual Review of Control, Robotics, and Autonomous Systems , volume=. 2022 , publisher=

    2022

  55. [63]

    Proceedings of the IEEE , volume=

    A game theoretic approach to controller design for hybrid systems , author=. Proceedings of the IEEE , volume=. 2000 , publisher=

    2000

  56. [64]

    IEEE Control Systems Magazine , volume=

    Runtime assurance for safety-critical systems: An introduction to safety filtering approaches for complex control systems , author=. IEEE Control Systems Magazine , volume=. 2023 , publisher=

    2023

  57. [65]

    1996 , publisher=

    Robust and optimal control , author=. 1996 , publisher=

    1996

  58. [66]

    The international journal of advanced manufacturing technology , volume=

    Review on model predictive control: An engineering perspective , author=. The international journal of advanced manufacturing technology , volume=. 2021 , publisher=

    2021

  59. [67]

    Robotics and Autonomous Systems , volume=

    Safety-critical advanced robots: A survey , author=. Robotics and Autonomous Systems , volume=. 2017 , publisher=

    2017

  60. [68]

    IEEE Robotics and Automation Letters , volume=

    A predictive safety filter for learning-based racing control , author=. IEEE Robotics and Automation Letters , volume=. 2021 , publisher=

    2021

  61. [69]

    IEEE Transactions on Robotics , volume=

    Barriernet: Differentiable control barrier functions for learning of safe robot control , author=. IEEE Transactions on Robotics , volume=. 2023 , publisher=

    2023

  62. [70]

    2020 59th IEEE Conference on Decision and Control (CDC) , pages=

    Learning control barrier functions from expert demonstrations , author=. 2020 59th IEEE Conference on Decision and Control (CDC) , pages=. 2020 , organization=

    2020

  63. [71]

    IEEE Transactions on Control Systems Technology , volume=

    A neural network approach for high-dimensional optimal control applied to multiagent path finding , author=. IEEE Transactions on Control Systems Technology , volume=. 2022 , publisher=

    2022

  64. [72]

    2021 European Control Conference (ECC) , pages=

    A neural network approach applied to multi-agent optimal control , author=. 2021 European Control Conference (ECC) , pages=. 2021 , organization=

    2021

  65. [73]

    2016 , publisher=

    Numerical methods for ordinary differential equations , author=. 2016 , publisher=

    2016

  66. [74]

    Journal of Process Control , volume=

    A dynamic penalty approach to state constraint handling in deep reinforcement learning , author=. Journal of Process Control , volume=. 2022 , publisher=

    2022

  67. [75]

    arXiv preprint arXiv:1805.11074 , year=

    Reward constrained policy optimization , author=. arXiv preprint arXiv:1805.11074 , year=

    arXiv

  68. [76]

    Journal of Computational Physics , volume=

    Optimal control of PDEs using physics-informed neural networks , author=. Journal of Computational Physics , volume=. 2023 , publisher=

    2023

  69. [77]

    Advances in neural information processing systems , volume=

    Pytorch: An imperative style, high-performance deep learning library , author=. Advances in neural information processing systems , volume=

  70. [78]

    arXiv preprint arXiv:1412.6980 , year=

    Adam: A method for stochastic optimization , author=. arXiv preprint arXiv:1412.6980 , year=

    arXiv

  71. [79]

    Electronics , volume=

    Optimizable control barrier functions to improve feasibility and add behavior diversity while ensuring safety , author=. Electronics , volume=. 2022 , publisher=

    2022

  72. [80]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Deep residual learning for image recognition , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  73. [81]

    Advances in neural information processing systems , volume=

    Multiscale deep equilibrium models , author=. Advances in neural information processing systems , volume=

  74. [82]

    arXiv preprint arXiv:2106.14342 , year=

    Stabilizing equilibrium models by jacobian regularization , author=. arXiv preprint arXiv:2106.14342 , year=

    arXiv

  75. [83]

    Advances in Neural Information Processing Systems , volume=

    Robust implicit networks via non-Euclidean contractions , author=. Advances in Neural Information Processing Systems , volume=

  76. [84]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Lyapunov-stable deep equilibrium models , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  77. [85]

    Advances in neural information processing systems , volume=

    Monotone operator equilibrium networks , author=. Advances in neural information processing systems , volume=

  78. [86]

    2002 , publisher=

    The implicit function theorem: history, theory, and applications , author=. 2002 , publisher=

    2002

  79. [87]

    Automatica , volume =

    Sufficient Conditions for Feasibility of Optimal Control Problems Using Control Barrier Functions , author =. Automatica , volume =. 2022 , publisher =

    2022

  80. [88]

    IEEE Conference on Decision and Control (CDC) , year =

    Auxiliary-Variable Adaptive Control Barrier Functions for Safety Critical Systems , author =. IEEE Conference on Decision and Control (CDC) , year =

  81. [89]

    American Control Conference (ACC) , pages =

    Safety-Critical Model Predictive Control with Discrete-Time Control Barrier Function , author =. American Control Conference (ACC) , pages =

  82. [90]

    IEEE Transactions on Automatic Control , volume=

    High-order control barrier functions , author=. IEEE Transactions on Automatic Control , volume=. 2021 , publisher=

    2021

  83. [91]

    SIAM review , volume=

    Optimization methods for large-scale machine learning , author=. SIAM review , volume=. 2018 , publisher=

    2018

  84. [92]

    Advances in Neural Information Processing Systems , volume=

    Optimal control for transformer architectures: Enhancing generalization, robustness and efficiency , author=. Advances in Neural Information Processing Systems , volume=

  85. [93]

    arXiv preprint arXiv:2510.09904 , year=

    Stability of transformers under layer normalization , author=. arXiv preprint arXiv:2510.09904 , year=

  86. [94]

    2024 , issn =

    Neural network approaches for parameterized optimal control , journal =. 2024 , issn =. doi:10.3934/fods.2024042 , author =

    2024 doi

  87. [95]

    Osher and Samy Wu Fung , title =

    Eric Gelphman and Deepanshu Verma and Nicole Tianjiao Yang and Stanley J. Osher and Samy Wu Fung , title =. 2026 American Control Conference (ACC) , year =

    2026

  88. [96]

    Foundations and Trends

    Distributed optimization and statistical learning via the alternating direction method of multipliers , author=. Foundations and Trends. 2011 , publisher=

    2011

  89. [97]

    Journal of mathematical imaging and vision , volume=

    A first-order primal-dual algorithm for convex problems with applications to imaging , author=. Journal of mathematical imaging and vision , volume=. 2011 , publisher=

    2011

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.