Fixed time convergence guarantees for Higher Order Control Barrier Functions

Janani S K; Shishir Kolathaya

REVIEW 3 major objections 6 minor 26 references

Fixed time convergence guarantees for Higher Order Control Barrier Functions

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper proposes a second-order control barrier function design that guarantees convergence to the safe set exactly at a user-specified time T, with no parameters to tune.

desk verdict The continuous-time repeated-root CBF construction is correct and worth knowing, but the paper's own Table II shows the fixed-time guarantee fails under zero-order hold, so the main practical claim needs major revision. read the letter →

arxiv 2507.13888 v1 pith:2XSZI7RP submitted 2025-07-18 eess.SY cs.ROcs.SY

Janani S K , Shishir Kolathaya This is my paper

classification eess.SYcs.ROcs.SY

keywords controlbarrierfunctionshigher-orderCBFsfixed-timeconvergencefinite-timesafetyrelativedegreetwoquadraticprogrammingrobotnavigationrepeatedrootscharacteristicpolynomial

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper proposes a way to make higher-order control barrier functions (HOCBFs) reach a safe set on a fixed schedule rather than only asymptotically. For a twice-differentiable safety function h with relative degree two, the authors enforce a repeated-root differential constraint (d/dt + cT)^2 h >= 0, with c computed from the initial conditions and the desired arrival time T. The theory says h(t) hits zero exactly at t = T and stays positive afterwards, so a controller can be told in advance when safety will be restored. The authors demonstrate the construction on point-mass, unicycle, and bicycle models and report faster computation and fewer user-chosen parameters than a prior finite-time HOCBF method.

What carries the argument

The central object is the repeated-root differential operator (d/dt + cT)^2 applied to the barrier function h. All roots of the characteristic polynomial coincide at -cT, so the homogeneous solution is (a0 + a1 t)$e^{{-cTt}}$; the parameter c is selected so the linear factor a0 + a1 t has a single zero at t = T. Enforcing the operator to be nonnegative turns the safety condition into a linear constraint on the control input, making the resulting quadratic program easy to solve in real time.

What would settle it

A direct check is to simulate the unicycle or bicycle model with dt = 0.001 and T = 5 s over the paper's 100 random initial conditions; if, as Table II reports, only 1/100 and 5/100 of the runs reach the safe set by T, that is concrete evidence against the continuous-time theorem's applicability to the sampled controller. One could also search for a zero-order-hold counterexample where h(t) dips below the comparison trajectory (a0 + a1 t)$e^{{-cTt}}$ between samples and never reaches zero by time T.

Watch

Extended reading notes

Core claim

The central claim is Theorem 1: for a $C^{2}$ barrier function of relative degree two with h(0) < 0 and any desired time T > 0, choose c = (-h(0) - hdot(0)T)/(h(0)$T^{2}$). If the controller enforces (d/dt + cT)^2 h >= 0, then the closed loop reaches h(t) = 0 at t = T and has h(t) > 0 for all t > T. The construction uses repeated roots in the characteristic polynomial of the homogeneous equation, so the equality-case trajectory is (a0 + a1 t)$e^{{-cTt}}$; choosing c makes the linear factor vanish exactly at T. The paper argues this gives fixed-time convergence to the safe set in a user-specified horizon, with a linear constraint in the CBF-QP, no extra tuning parameters, and lower computation than the recursive finite-time HOCBF baseline in [12].

Load-bearing premise

The continuous-time guarantee assumes the inequality (d/dt + cT)^2 h >= 0 holds at every instant, while the controller enforces it only at discrete sampling instants with zero-order hold, and the paper does not prove that discretization preserves the bound.

Editorial extensions

If this is right

For second-order systems, a robot or vehicle can be guaranteed to enter the safe set at a prescribed wall-clock time T, not merely eventually.
The safety constraint is linear in the control input, so the real-time CBF-QP is inexpensive; the paper reports roughly half the computation time of the existing finite-time HOCBF baseline.
No user-selected gains or exponents are needed: c is fixed directly by h(0), hdot(0), and T, removing a parameter-tuning loop.
As the discretization step shrinks, the actual convergence time approaches T, indicating that the continuous-time guarantee is the limit of the sampled implementation.
In the point-mass simulations the proposed controller reaches the safe set in 100% of the random trials tested, while the comparison baseline reached only 1%.

Reading between the lines

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

The repeated-root construction should extend naturally to relative degree n by enforcing (d/dt + cT)^n h >= 0 with an (n-1)-degree polynomial factor having a single root at T; the paper lists nth-order generalization as future work.
Because the worst-case arrival time is fixed by T, the same constraint could be composed with temporal-logic or scheduling tasks where each safety subtask must finish by a deadline.
The paper's Table II shows that the unicycle and bicycle models converge poorly at dt = 0.001 (1/100 and 5/100 of random initial conditions by T = 5 s), suggesting the continuous-time theorem does not automatically transfer to zero-order-hold sample-and-hold control; a sampled-data analysis would be needed to close that gap.
Adding a slack variable to the inequality could compensate for discretization lag or control saturation, which the paper itself identifies as a direction for future work.

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.

Desk Editor's Note

read the letter

Here is my read. The continuous-time core is a legitimate small extension: the repeated-root constraint (d/dt + cT)^2 h >= 0 for relative-degree-two CBFs is not in the cited literature, and it gives a clean, linear, parameter-free (except T) alternative to the recursive finite-time conditions in [12]. The derivation of c from the initial conditions is correct for the equality case, and the inequality case is repairable with a standard comparison-principle argument (e.g., z = e^{lambda t}(h - h_ref), z'' >= 0). Credit where due: for the point-mass model with small dt it converges essentially on time, the computation is cheaper, and the comparison to the baseline is broadly fair.

The soft spots, in order of severity:

1. The discrete-time gap is load-bearing. Theorem 1 is a continuous-time statement; the QP enforces the constraint only at sample instants under zero-order hold, with no inter-sample bound. Table II is the empirical signature: at dt = 0.001 and T = 5 s, only 1/100 unicycle and 5/100 bicycle initial conditions reach the safe set by 5 s, while the theorem predicts 100/100. By 7.5 s almost all converge. That is roughly 50% overshoot of the horizon, not the 'little amount' claimed in Section VI. The paper's own data therefore contradicts the abstract's claim that the method 'reliably enforces convergence within the desired time.'

2. The inequality-case proof is asserted, not derived. As written, the theorem says h(t) -> 0 at t = T, but under the inequality h may cross zero earlier; what is actually guaranteed is h(t) >= h_ref(t) and hence h(T) >= 0. This is fixable but the statement is inaccurate.

3. Calling this 'fixed-time' is mildly misleading. The gain c depends on the initial state, so this is prescribed-time for a known initial condition, not fixed-time in the usual initial-condition-independent sense. Minor wording issue, but it matters for how the contribution is positioned.

4. The comparison with [12] lacks tuning detail. The authors say they tried various parameters, but the reader cannot tell whether [12] got a fair shake. The computation-time comparison is suggestive rather than conclusive.

The citation pattern is fine: [22] and [12] are the right anchors, and there is no self-citation inflation. The paper is written by people who understand the ODE mechanics, but the gap between continuous-time guarantee and sampled implementation is the whole ballgame for a method whose selling point is exact timing.

Who is this for: researchers in CBF-based safety who want a simple prescribed-time linear safety filter for second-order robotics. It deserves a serious referee, but only with the expectation of major revision: either an inter-sample analysis, an explicit safety margin on T, or an honest rescoping to approximate convergence. As is, Table II should block acceptance.

Referee Report

3 major / 6 minor

Summary. The paper proposes a fixed-time higher-order CBF design for relative-degree-two systems. The method chooses c from the initial conditions so that the homogeneous reference solution of (D + cT)^2 h = 0 has a root at t = T, and then enforces the inequality (D + cT)^2 h >= 0 in a QP. The authors claim this yields convergence to the safe set within the user-specified time T, with a closed-form and parameter-light alternative to existing finite-time HOCBFs. The continuous-time equality-case derivation is correct, but the inequality-case proof is omitted, the statement of the theorem is imprecise for the inequality case, and the discrete-time zero-order-hold implementation is shown by the paper's own Table II to violate the claimed timing for the nonholonomic models.

Significance. The continuous-time construction is simple and computationally attractive: c is given in closed form, the CBF constraint is linear, and no parameter search beyond T is needed. This is a useful incremental addition to the fixed-time HOCBF literature, and the paper deserves credit for reporting the random-initial-condition convergence counts in Table II. However, the advertised headline claim that the method 'reliably enforces convergence within the desired time' is not supported for the discrete-time implementation: at dt = 0.001 and T = 5 s, only 1/100 unicycle and 5/100 bicycle initial conditions reach the safe set by 5 s, with recovery only by 7.5 s. The central contribution is therefore a continuous-time design with an unresolved discretization gap rather than a demonstrated fixed-time guarantee for the implemented controller.

major comments (3)

[Section IV, Theorem 1] The proof of Theorem 1 verifies only the equality case (D + cT)^2 h = 0, constructing c so that the homogeneous solution h_ref(t) = (h(0) + (hdot(0) + cT h(0))t) e^{-cT t} vanishes at t = T. For the inequality constraint actually used in the QP, the theorem asserts h(t) -> 0 at t = T, but no comparison argument is provided. The claim as stated is also inaccurate: from (D + cT)^2 h >= 0 one obtains h(t) >= h_ref(t), which gives h(T) >= 0 and h(t) > 0 for t > T, but h need not approach 0 at T; it may enter the safe set strictly earlier. A repair is available: for w = h - h_ref, the transformed variable z = e^{cT t} w satisfies z'' >= 0 with z(0) = z'(0) = 0, so z >= 0 and the comparison holds. This step should be included, and the theorem statement should be corrected to 'h(T) >= 0' (i.e., reachability within time T).
[Section V, Table II and Section VI] The continuous-time guarantee requires (D + cT)^2 h >= 0 at every instant, but the QP in Section V imposes this constraint only at sample times under zero-order hold, with no inter-sample estimate. The paper's own data contradict the advertised fixed-time guarantee: at dt = 0.001 and T = 5 s, only 1/100 unicycle and 5/100 bicycle initial conditions reach the safe set by 5 s, while the theorem predicts 100/100; convergence is recovered only by 7.5 s, an overshoot of about 50% of T. Section VI attributes this to discretization but calls it 'a little amount,' which is not supported by Table II. This is a load-bearing gap: either add a formal discrete-time or robustness analysis, or clearly scope the claims to the continuous-time setting and provide a validated compensation strategy (for example, a back-off in the nominal T) for the sample-and-hold implementation.
[Section IV, comparison with [12]] The comparison with the existing method in [12] is presented as a major advantage of the proposed approach, but the reported baseline performance is very sensitive to parameter tuning and the paper does not describe the search procedure for the baseline parameters in enough detail. For example, Table III reports 1% convergence for the existing method on the point-mass model, while the text says the method was run for 15 s for a required convergence time of 5 s; the reader cannot determine whether this is a fundamental limitation of [12] or a consequence of a particular tuning. The comparison should be made more systematic, with the same random initial conditions, the same time horizon, and a documented parameter-selection rule.

minor comments (6)

[Abstract] The abstract ends with 'within a user-specified finite'; the word 'time' is missing.
[Section II-E.1] The point-mass model description begins with 'A unicycle model has state variables...'; this should be corrected to 'point-mass'.
[Section IV, Equation (13)] Equation (13) has typesetting errors: '|h|p' should be '|h|^p' and the expression for alpha appears to be missing a division operator.
[Section III] The problem formulation says the barrier function is 'the negative Euclidean distance to a goal', but the simulations use h = 4 - ||s - s_g||^2, which is a squared-distance barrier; the formulation should match the implementation.
[Section IV-A] The claim that the proposed method has 'no such parameter that we would have to fix' is imprecise because T is a user-specified parameter; the intended statement is that there are no tuning parameters beyond the desired convergence time.
[References] Reference [6] appears in the bibliography but is not cited in the text; please check the citation list.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: c is synthesized so the reference trajectory crosses zero at the user-chosen T, and the guarantee follows from a comparison inequality rather than from any fitted prediction.

full rationale

The derivation chain is self-contained. In Theorem 1, c is not a parameter fitted to observed outcomes; it is computed from the initial conditions and the user-specified horizon T by requiring the equality-case solution h_ref(t)=(h(0)+(hdot(0)+cT h(0))t)e^{-cTt} to satisfy h_ref(T)=0. This is standard control synthesis, not a circular prediction, because T is an input specification and the conclusion that h(T)>=0 (hence crossing by time T) is obtained by a comparison h>=h_ref that is independent of the choice of c. The paper asserts rather than proves that inequality, but the comparison is valid (e.g., w=e^{cTt}(h-h_ref) has w''>=0 with w(0)=w'(0)=0). The discretization gap between the continuous-time theorem and the zero-order-hold QP is real and documented in Table II, and the unproven inequality-case comparison is an omitted proof; both are correctness/completeness issues, not circularity. No load-bearing self-citation appears: the central construction builds on [22] (Garg and Panagou), and the authors' own prior works are cited only as background. Therefore no circular step is present.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central guarantee depends on five assumptions: the comparison principle for the linear inequality, relative degree two, unsafe initial condition, small-angle bicycle approximation, and a benign discretization assumption. The last is the most fragile: Table II shows the discrete-time implementation misses the deadline for most unicycle and bicycle runs, so the continuous-time theorem does not transfer to the simulation by itself. The only user-chosen free parameter is the convergence time T; c is derived from initial conditions.

free parameters (1)

T = user-specified (e.g., 10 s in simulations)
Desired convergence time, chosen by the user in advance; not fitted to data. The method requires T > 0 and all guarantees depend on it.

assumptions (5)

standard math Comparison principle for linear differential inequalities of the form (d/dt + a)^2 h >= 0 with equal initial conditions implies h(t) >= h_ref(t).
The proof of Theorem 1 uses this to conclude h(t) >= (a0 + a1 t) e^{-cT t}; the principle is not stated explicitly but is standard.
domain assumption The system is control-affine and the barrier h has relative degree two, so the control appears in h-double-dot.
Assumed throughout Section IV and used to write the QP constraint in terms of u.
domain assumption Initial condition is unsafe, h(0) < 0, and h-dot(0) is known.
Theorem 1 explicitly assumes h(0) < 0; the c formula divides by h(0), so h(0) = 0 is not handled.
domain assumption The small-angle approximation for the bicycle model (cos beta ~= 1, sin beta ~= beta) is valid.
Remark 1 in Section II-E introduces this to obtain the affine dynamics (11), and the simplified model (12) is used in simulations.
domain assumption Discretization with zero-order hold preserves the continuous-time comparison bound closely enough that the guarantee remains valid.
The paper attributes simulation delays to discretization (Section VI) but provides no formal bound; Table II shows large violations for unicycle and bicycle at dt = 0.001, so this assumption is load-bearing.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Fixed time convergence guarantees for Higher Order Control Barrier Functions." pith.science (2026). https://pith.science/paper/2XSZI7RP

@misc{pith2026250713888,
  author       = {Pith},
  title        = {Pith review of: Fixed time convergence guarantees for Higher Order Control Barrier Functions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2XSZI7RP}},
  note         = {Machine review of arXiv:2507.13888}
}

read the original abstract

We present a novel method for designing higher-order Control Barrier Functions (CBFs) that guarantee convergence to a safe set within a user-specified finite. Traditional Higher Order CBFs (HOCBFs) ensure asymptotic safety but lack mechanisms for fixed-time convergence, which is critical in time-sensitive and safety-critical applications such as autonomous navigation. In contrast, our approach imposes a structured differential constraint using repeated roots in the characteristic polynomial, enabling closed-form polynomial solutions with exact convergence at a prescribed time. We derive conditions on the barrier function and its derivatives that ensure forward invariance and fixed-time reachability, and we provide an explicit formulation for second-order systems. Our method is evaluated on three robotic systems - a point-mass model, a unicycle, and a bicycle model and benchmarked against existing HOCBF approaches. Results demonstrate that our formulation reliably enforces convergence within the desired time, even when traditional methods fail. This work provides a tractable and robust framework for real-time control with provable finite-time safety guarantees.

Figures

Figures reproduced from arXiv: 2507.13888 by the authors.

**Figure 1.** Comparison of the proposed and existing methods across different models.The proposed method is more closer [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

26 extracted references · 12 canonical work pages

[12]

High Order Control Lyapunov-Barrier Functions for Temporal Logic Specifications

W. Xiao, C. A. Belta, and C. G. Cassandras, “High order control lyapunov-barrier functions for temporal logic specifications,” 2021. [Online]. Available: https://arxiv.org/abs/2102.06787

work page Pith review arXiv 2021
[22]

Robust control barrier and control lyapunov functions with fixed-time convergence guarantees,

——, “Robust control barrier and control lyapunov functions with fixed-time convergence guarantees,” in 2021 American Control Con- ference (ACC), 2021, pp. 2292–2297

work page 2021
[1]

Constrained policy op- timization,

J. Achiam, D. Held, A. Tamar, and P. Abbeel, “Constrained policy op- timization,” in International conference on machine learning. PMLR, 2017, pp. 22–31

2017
[2]

Model-based safe deep reinforcement learning via a constrained proximal policy optimization algorithm,

A. K. Jayant and S. Bhatnagar, “Model-based safe deep reinforcement learning via a constrained proximal policy optimization algorithm,” in Advances in Neural Information Processing Systems , vol. 35, 2022, pp. 24 432–24 445

2022
[3]

Safe model-based reinforcement learning with stability guarantees,

F. Berkenkamp, M. Turchetta, A. Schoellig, and A. Krause, “Safe model-based reinforcement learning with stability guarantees,” in Advances in Neural Information Processing Systems , vol. 30, 2017

2017
[4]

Hamilton-Jacobi reachability: A brief overview and recent advances,

S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin, “Hamilton-Jacobi reachability: A brief overview and recent advances,” in 2017 IEEE 56th Annual Conference on Decision and Control (CDC) , 2017

2017
[5]

A physics- informed machine learning framework for safe and optimal control of autonomous systems,

M. Tayal, A. Singh, S. Kolathaya, and S. Bansal, “A physics- informed machine learning framework for safe and optimal control of autonomous systems,” in Forty-second International Conference on Machine Learning , 2025. [Online]. Available: https://openreview.net/forum?id=SrfwiloGQF

2025
[6]

Semi-supervised safe visuomotor policy synthesis using barrier certificates,

M. Tayal, A. Singh, P. Jagtap, and S. Kolathaya, “Semi-supervised safe visuomotor policy synthesis using barrier certificates,” arXiv preprint arXiv:2409.12616, 2024

arXiv 2024

Show all 26 references

[7]

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

2017
[8]

Control barrier functions: Theory and applications,

A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, and P. Tabuada, “Control barrier functions: Theory and applications,” in 18th European control conference (ECC) . IEEE, 2019, pp. 3420– 3431

2019
[9]

Safety-critical control of a planar quadrotor,

G. Wu and K. Sreenath, “Safety-critical control of a planar quadrotor,” in 2016 American Control Conference (ACC) , 2016, pp. 2252–2258

2016
[10]

Control barrier functions in dynamic uavs for kinematic obstacle avoidance: A col- lision cone approach,

M. Tayal, R. Singh, J. Keshavan, and S. Kolathaya, “Control barrier functions in dynamic uavs for kinematic obstacle avoidance: A col- lision cone approach,” in 2024 American Control Conference (ACC) . IEEE, 2024, pp. 3722–3727

2024
[11]

Safety-critical control for dy- namical bipedal walking with precise footstep placement,

Q. Nguyen and K. Sreenath, “Safety-critical control for dy- namical bipedal walking with precise footstep placement,” IFAC- PapersOnLine, vol. 48, no. 27, pp. 147–154, 2015

2015
[13]

Exponential control barrier functions for enforcing high relative-degree safety-critical constraints,

Q. Nguyen and K. Sreenath, “Exponential control barrier functions for enforcing high relative-degree safety-critical constraints,” in 2016 American Control Conference (ACC) , 2016, pp. 322–328

2016
[14]

Control barrier functions for systems with high relative degree,

W. Xiao and C. Belta, “Control barrier functions for systems with high relative degree,” 2019. [Online]. Available: https: //arxiv.org/abs/1903.04706

2019 arXiv
[15]

Sufficient conditions for uniform asymptotic stability and input-to-state stability using high- order control barrier functions,

M. Marley, R. Skjetne, and A. R. Teel, “Sufficient conditions for uniform asymptotic stability and input-to-state stability using high- order control barrier functions,” IEEE Transactions on Automatic Control, vol. 69, no. 4, pp. 2352–2366, 2024

2024
[16]

High-order control barrier function-based safety control of constrained robotic systems: An augmented dynamics approach,

H. Wang, J. Peng, F. Zhang, and Y . Wang, “High-order control barrier function-based safety control of constrained robotic systems: An augmented dynamics approach,” IEEE/CAA Journal of Automatica Sinica, vol. 11, no. 12, pp. 2487–2496, 2024

2024
[17]

Hocbf- based quadratic programs with application to safety-critical planning and control for overtake maneuver,

H. Ding, W. Zhuang, J. Hu, M. Zhao, W. Wang, and G. Yin, “Hocbf- based quadratic programs with application to safety-critical planning and control for overtake maneuver,” in 2024 8th CAA International Conference on Vehicular Control and Intelligence (CVCI) , 2024, pp. 1–6

2024
[18]

A control barrier function based approach for safe and efficient navigation of unicycle mobile robots,

A. Rahmanian, A. A. Nasab, and M. H. Asemani, “A control barrier function based approach for safe and efficient navigation of unicycle mobile robots,” in 2024 10th International Conference on Control, Instrumentation and Automation (ICCIA) , 2024, pp. 1–6

2024
[19]

Finite-time stability for time-varying nonlinear dynamical systems,

W. M. Haddad, S. G. Nersesov, and L. Du, “Finite-time stability for time-varying nonlinear dynamical systems,” in 2008 American Control Conference, 2008, pp. 4135–4139

2008
[20]

Control of multi-agent systems with finite time control barrier certificates and temporal logic,

M. Srinivasan, S. Coogan, and M. Egerstedt, “Control of multi-agent systems with finite time control barrier certificates and temporal logic,” in 2018 IEEE Conference on Decision and Control (CDC) , 2018, pp. 1991–1996

2018
[21]

Control-lyapunov and control-barrier functions based quadratic program for spatio-temporal specifications,

K. Garg and D. Panagou, “Control-lyapunov and control-barrier functions based quadratic program for spatio-temporal specifications,” in 2019 IEEE 58th Conference on Decision and Control (CDC) . IEEE, Dec. 2019, p. 1422–1429. [Online]. Available: http://dx.doi. org/10.1109/CDC4...

2019
[23]

Implicit fixed-time convergence iss safe control barrier functions for signal temporal logic tasks,

M. Li and Z. Sun, “Implicit fixed-time convergence iss safe control barrier functions for signal temporal logic tasks,” in 2022 IEEE 17th International Conference on Control and Automation (ICCA) , 2022, pp. 722–727

2022
[24]

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

2019
[25]

Learning a formally verified control barrier function in stochastic environment,

M. Tayal, H. Zhang, P. Jagtap, A. Clark, and S. Kolathaya, “Learning a formally verified control barrier function in stochastic environment,” in 2024 IEEE 63rd Conference on Decision and Control (CDC), 2024, pp. 4098–4104

2024
[26]

Cp-ncbf: A con- formal prediction-based approach to synthesize verified neural control barrier functions,

M. Tayal, A. Singh, P. Jagtap, and S. Kolathaya, “Cp-ncbf: A con- formal prediction-based approach to synthesize verified neural control barrier functions,” arXiv preprint arXiv:2503.17395 , 2025

2025 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [12]

High Order Control Lyapunov-Barrier Functions for Temporal Logic Specifications

W. Xiao, C. A. Belta, and C. G. Cassandras, “High order control lyapunov-barrier functions for temporal logic specifications,” 2021. [Online]. Available: https://arxiv.org/abs/2102.06787

work page Pith review arXiv 2021

[2] [22]

Robust control barrier and control lyapunov functions with fixed-time convergence guarantees,

——, “Robust control barrier and control lyapunov functions with fixed-time convergence guarantees,” in 2021 American Control Con- ference (ACC), 2021, pp. 2292–2297

work page 2021

[3] [1]

Constrained policy op- timization,

J. Achiam, D. Held, A. Tamar, and P. Abbeel, “Constrained policy op- timization,” in International conference on machine learning. PMLR, 2017, pp. 22–31

2017

[4] [2]

Model-based safe deep reinforcement learning via a constrained proximal policy optimization algorithm,

A. K. Jayant and S. Bhatnagar, “Model-based safe deep reinforcement learning via a constrained proximal policy optimization algorithm,” in Advances in Neural Information Processing Systems , vol. 35, 2022, pp. 24 432–24 445

2022

[5] [3]

Safe model-based reinforcement learning with stability guarantees,

F. Berkenkamp, M. Turchetta, A. Schoellig, and A. Krause, “Safe model-based reinforcement learning with stability guarantees,” in Advances in Neural Information Processing Systems , vol. 30, 2017

2017

[6] [4]

Hamilton-Jacobi reachability: A brief overview and recent advances,

S. Bansal, M. Chen, S. Herbert, and C. J. Tomlin, “Hamilton-Jacobi reachability: A brief overview and recent advances,” in 2017 IEEE 56th Annual Conference on Decision and Control (CDC) , 2017

2017

[7] [5]

A physics- informed machine learning framework for safe and optimal control of autonomous systems,

M. Tayal, A. Singh, S. Kolathaya, and S. Bansal, “A physics- informed machine learning framework for safe and optimal control of autonomous systems,” in Forty-second International Conference on Machine Learning , 2025. [Online]. Available: https://openreview.net/forum?id=SrfwiloGQF

2025

[8] [6]

Semi-supervised safe visuomotor policy synthesis using barrier certificates,

M. Tayal, A. Singh, P. Jagtap, and S. Kolathaya, “Semi-supervised safe visuomotor policy synthesis using barrier certificates,” arXiv preprint arXiv:2409.12616, 2024

arXiv 2024