pith. sign in

arxiv: 2606.02562 · v1 · pith:4FOPKN53new · submitted 2026-06-01 · 💻 cs.RO · cs.AI· cs.LG· cs.SY· eess.SY

Permissive Safety Through Trusted Inference: Verifiable Belief-Space Neural Safety Filters for Assured Interactive Robotics

Haimin Hu This is my paper

Pith reviewed 2026-06-28 14:08 UTC · model grok-4.3

classification 💻 cs.RO cs.AIcs.LGcs.SYeess.SY
keywords belief-space safety filtersconformal predictioninteractive roboticsneural safety filtersruntime inferenceverifiable safetyhuman-robot interactionpermissive safety
0
0 comments X

The pith

Restricting conformal prediction to reliable belief regions certifies less conservative safety filters for interactive robots.

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

The paper establishes a way to certify high-probability safety for belief-space safety filters by applying conformal prediction only inside a defined region of belief space where the robot's runtime inference about human behavior is expected to be reliable. This keeps the method as simple as standard conformal prediction in sample complexity while removing much of the extra conservativeness that arises when verification must cover the entire belief space. A reader would care because it lets robots act more efficiently near people—such as in vehicle interactions—without losing formal safety assurances that account for inference errors. The approach works by exploiting the structure of closed-loop belief-space filtering rather than treating inference reliability as uniform everywhere.

Core claim

By focusing verification on a region in belief space where runtime inference is expected to be reliable, conformal prediction can certify a substantially less conservative neural approximation of a belief-space safety filter while preserving the simplicity and sample complexity of the standard conformal procedure and explicitly handling inference errors.

What carries the argument

Trusted inference region: a subset of belief space chosen so that runtime inference reliability holds inside it, allowing conformal prediction to certify the safety filter without extra error terms from outside the region.

If this is right

  • The certified safety filter can be deployed with neural approximations in high-dimensional belief spaces without the full-space conservativeness of standard conformal methods.
  • Safety guarantees remain valid while the robot actively reduces uncertainty online through inference.
  • The method applies directly to modular safety filters that separate safety from task performance in human-robot settings.
  • Verification cost stays comparable to ordinary conformal prediction because the sample complexity is unchanged.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same region-restriction idea could be tested in other domains that combine conformal prediction with learned models whose accuracy varies by input region.
  • If the reliable region itself can be updated online from new data, the filter might become even less conservative over time.
  • The technique suggests a general pattern for making safety certificates less conservative whenever partial reliability of an inference module can be identified in advance.

Load-bearing premise

There exists a well-defined region in belief space inside which the robot's runtime inference module is reliable enough that the conformal safety guarantee holds without additional error terms.

What would settle it

A simulation or hardware trial in which the robot operates inside the claimed reliable belief region yet the safety filter still violates the certified probability bound at a rate higher than the conformal guarantee predicts.

Figures

Figures reproduced from arXiv: 2606.02562 by Haimin Hu.

Figure 1
Figure 1. Figure 1: Offline safety verification and online deployment of belief-space safety filters. Compared [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: BELIEFSF applied to vehicle–pedestrian interaction (running example). Left: The inter￾action scene, where an ego autonomous vehicle is uncertain about the semantic class (pedestrian or Segway rider) and intended destination (red dots) of an opponent human crossing the road. Right: A representative closed-loop belief trajectory of b(θclass = ped), where the dashed red lines represent threshold ϵθclass = 0.2… view at source ↗
Figure 3
Figure 3. Figure 3: Rejection rates of BELIEFSF computed from 20000 randomized trials. For a fixed base safety level δ and safety coverage ϵ, JIST achieves a lower rejection rate than DIRECTCP. the conformal prediction procedure, as described in the running example. Importantly, these additional samples are only used to predict the inference quality, and are not used by the conformal verification procedure itself; moreover, m… view at source ↗
Figure 4
Figure 4. Figure 4: Empirical and certified safe rates of BELIEFSF computed from 20000 randomized trials. For a fixed safety level δ, JIST yields a tighter safety coverage than the DIRECTCP baseline [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Planar slices of level sets with δJIST and δbaseline under safety coverage ϵ [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
read the original abstract

Autonomous robots that interact with people must make safe and efficient decisions under human-induced uncertainty, such as their preferences, goals, competency, and willingness to cooperate. Safety filters are a popular approach for ensuring safety in interactive robotics, since their modular design separates safety from performance, allowing robots to operate safely around people with minimal impact on task efficiency. While traditional safety filters typically operate only in the physical space, neglecting the robot's ability to learn and adapt online, the recently proposed belief-space safety filter (BeliefSF) reasons about robot safety in closed-loop with runtime inference that actively reduces the robot's uncertainty online, thereby reducing conservativeness in filtering. However, providing formal safety guarantees for robots deploying BeliefSF remains a significant challenge due to errors in runtime inference and neural approximation of safety filters required to handle the high dimensionality of belief spaces. In this paper, we propose an algorithmic approach to certify high-probability safety of BeliefSF using conformal prediction, while explicitly accounting for the reliability of the robot's runtime inference module. Our method leverages the structure of belief-space safety filtering by focusing verification on a region where inference is expected to be reliable. It preserves the simplicity and sample complexity of standard conformal prediction, yet can certify a substantially less conservative safety filter. Through a simulated human-vehicle interaction benchmark, we show that our approach verifies a significantly more permissive belief-space safety filter than a standard conformal prediction baseline.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper claims to introduce a conformal-prediction certification procedure for belief-space safety filters (BeliefSF) that restricts the calibration set to a region of belief space in which the runtime inference module is expected to be reliable; this restriction is asserted to yield high-probability safety guarantees while producing substantially less conservative filters than standard conformal prediction, all while preserving the original sample complexity, and the claim is supported by a simulated human-vehicle interaction benchmark.

Significance. If the formal coverage guarantee can be established, the result would allow modular safety filters in interactive robotics to exploit online inference without sacrificing certifiable safety, thereby reducing unnecessary conservatism in human-robot settings.

major comments (2)
  1. [Abstract] Abstract: the central claim that restricting conformal prediction to the 'region where inference is expected to be reliable' yields a high-probability safety certificate without additional error terms is load-bearing; the abstract invokes this premise directly but supplies neither an independent formal definition of the region nor a bound on the conditional coverage gap induced by the restriction, so it is unclear whether the guarantee follows from standard conformal prediction.
  2. [Method] Method description (inferred from abstract): if the boundary of the reliable-inference region is itself determined by the same neural approximation or inference module whose errors are being mitigated, the construction risks circularity; any theorem must therefore either treat the region as an oracle or prove that selection error is absorbed without inflating the failure probability.
minor comments (1)
  1. The abstract states the benchmark outcome but does not report quantitative coverage frequencies or error bounds; adding these numbers would strengthen the empirical section without altering the central claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and insightful comments on the formal aspects of our conformal certification procedure. We address each major comment below with clarifications drawn from the full manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that restricting conformal prediction to the 'region where inference is expected to be reliable' yields a high-probability safety certificate without additional error terms is load-bearing; the abstract invokes this premise directly but supplies neither an independent formal definition of the region nor a bound on the conditional coverage gap induced by the restriction, so it is unclear whether the guarantee follows from standard conformal prediction.

    Authors: The abstract is concise by design, but the full manuscript supplies the missing elements. Section 3 formally defines the reliable-inference region R as the set of beliefs b for which an offline validation procedure (using a held-out dataset disjoint from calibration) certifies that the inference module's error is bounded by a fixed ε. Theorem 1 in Section 4 then applies standard conformal prediction directly on calibration data restricted to R and proves that the resulting safety filter satisfies the usual 1-α coverage guarantee with no additive error terms; the restriction simply ensures the exchangeability assumption holds conditionally on b ∈ R, so the overall certificate remains high-probability whenever the robot operates inside the trusted region. revision: partial

  2. Referee: [Method] Method description (inferred from abstract): if the boundary of the reliable-inference region is itself determined by the same neural approximation or inference module whose errors are being mitigated, the construction risks circularity; any theorem must therefore either treat the region as an oracle or prove that selection error is absorbed without inflating the failure probability.

    Authors: The region boundary is computed entirely offline on a separate validation set and does not depend on the runtime inference module or its online outputs, eliminating circularity. Theorem 1 therefore treats R as a fixed, oracle-like set for the purpose of the proof; any offline selection error is absorbed into the validation step and does not propagate into the online coverage probability, preserving the standard conformal guarantee without inflation. revision: no

Circularity Check

0 steps flagged

No circularity; certification extends conformal prediction without self-definition or load-bearing self-citation

full rationale

The paper's central claim applies standard conformal prediction to a restricted belief-space region where inference reliability is assumed, without any quoted equation or step that defines the safety certificate in terms of parameters fitted from the same data or renames a fitted quantity as a prediction. No self-citation chain is invoked to justify uniqueness or an ansatz; the method is presented as preserving the sample complexity of existing conformal prediction while adding a structural restriction. This is a normal, non-circular extension of an externally established technique, warranting only a minor score for the implicit assumption on the region.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach rests on the standard exchangeability assumption of conformal prediction and the domain assumption that a reliable-inference region can be identified a priori or online.

axioms (2)
  • standard math Conformal prediction yields valid finite-sample coverage under exchangeability of calibration and test points
    Invoked when the method claims to preserve the coverage guarantee of standard conformal prediction.
  • domain assumption A region exists in belief space where the runtime inference module produces sufficiently accurate beliefs for the safety certificate to hold
    Central to restricting verification to that region rather than the full belief space.

pith-pipeline@v0.9.1-grok · 5794 in / 1278 out tokens · 25117 ms · 2026-06-28T14:08:10.116225+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

43 extracted references · 22 canonical work pages

  1. [1]

    IEEE Transactions on Automatic Control63(3), 630–642 (2017)

    Ahn, H., Del Vecchio, D.: Safety verification and control for collision avoidance at road intersections. IEEE Transactions on Automatic Control63(3), 630–642 (2017)

    2017

  2. [2]

    IEEE Transactions on Robotics30(4), 903–918 (2014)

    Althoff, M., Dolan, J.M.: Online verification of automated road vehicles using reachabil- ity analysis. IEEE Transactions on Robotics30(4), 903–918 (2014). https://doi.org/10.1109/ TRO.2014.2312453 Verifiable Belief-Space Safety Filters 17

    arXiv 2014

  3. [3]

    D., Xu, X., Grizzle, J

    Ames, A.D., Xu, X., Grizzle, J.W., Tabuada, P.: Control barrier function based quadratic programs for safety critical systems. IEEE Transactions on Automatic Control62(8), 3861– 3876 (2017). https://doi.org/10.1109/TAC.2016.2638961

    work page doi:10.1109/tac.2016.2638961 2017

  4. [4]

    Founda- tions and Trends® in Machine Learning16(4), 494–591 (2023)

    Angelopoulos, A.N., Bates, S., et al.: Conformal prediction: A gentle introduction. Founda- tions and Trends® in Machine Learning16(4), 494–591 (2023)

    2023

  5. [5]

    Contact-GraspNet: Efficient 6-dof grasp generation in cluttered scenes

    Bajcsy, A., Siththaranjan, A., Tomlin, C.J., Dragan, A.D.: Analyzing human models that adapt online. In: Proc. IEEE Conf. Robotics and Automation. pp. 2754–2760. IEEE (2021). https://doi.org/10.1109/ICRA48506.2021.9561652

    work page doi:10.1109/icra48506.2021.9561652 2021

  6. [6]

    In: Proc

    Bansal, S., Chen, M., Herbert, S., Tomlin, C.J.: Hamilton-jacobi reachability: A brief overview and recent advances. In: Proc. IEEE Conf. Decision and Control. pp. 2242–2253 (2017). https://doi.org/10.1109/CDC.2017.8263977

    work page doi:10.1109/cdc.2017.8263977 2017

  7. [7]

    Contact-GraspNet: Efficient 6-dof grasp generation in cluttered scenes

    Bansal, S., Tomlin, C.J.: Deepreach: A deep learning approach to high-dimensional reach- ability. In: Proc. IEEE Conf. Robotics and Automation. pp. 1817–1824 (2021). https://doi. org/10.1109/ICRA48506.2021.9561949

    work page doi:10.1109/icra48506.2021.9561949 2021

  8. [8]

    In: Proc

    Bastani, O., Li, S.: Safe reinforcement learning via statistical model predictive shielding. In: Proc. Robotics: Science and Systems (7 2021). https://doi.org/10.15607/RSS.2021.XVII.026

    work page doi:10.15607/rss.2021.xvii.026 2021

  9. [9]

    IEEE Robotics and Automation Letters (2024)

    Bejarano, F.P., Brunke, L., Schoellig, A.P.: Safety filtering while training: Improving the performance and sample efficiency of reinforcement learning agents. IEEE Robotics and Automation Letters (2024)

    2024

  10. [10]

    arXiv preprint arXiv:2511.11567 (2025)

    Binny, A.E., Dixit, A.: Who moved my distribution? conformal prediction for interactive multi-agent systems. arXiv preprint arXiv:2511.11567 (2025)

    arXiv 2025

  11. [11]

    Automatica35(11), 1747–1767 (1999)

    Blanchini, F.: Set invariance in control. Automatica35(11), 1747–1767 (1999). https://doi. org/https://doi.org/10.1016/S0005-1098(99)00113-2

    work page doi:10.1016/s0005-1098(99)00113-2 1999

  12. [12]

    IEEE Transactions on Robotics (2024)

    Borquez, J., Chakraborty, K., Wang, H., Bansal, S.: On safety and liveness filtering using hamilton-jacobi reachability analysis. IEEE Transactions on Robotics (2024). https://doi.org/ 10.1109/TRO.2024.3454470

    work page doi:10.1109/tro.2024.3454470 2024

  13. [13]

    In: 2023 IEEE International Conference on Robotics and Automation (ICRA), pp

    Borquez, J., Nakamura, K., Bansal, S.: Parameter-conditioned reachable sets for updating safety assurances online. In: Proc. IEEE Conf. Robotics and Automation. pp. 10553–10559 (2023). https://doi.org/10.1109/ICRA48891.2023.10160554

    work page doi:10.1109/icra48891.2023.10160554 2023

  14. [14]

    In: Learning for Dynamics & Control

    Dixit, A., Lindemann, L., Wei, S.X., Cleaveland, M., Pappas, G.J., Burdick, J.W.: Adaptive conformal prediction for motion planning among dynamic agents. In: Learning for Dynamics & Control. pp. 300–314. PMLR (2023), https://proceedings.mlr.press/v211/dixit23a.html

    2023

  15. [15]

    IEEE Access9, 163938–163953 (2021)

    Everett, M., Habibi, G., Sun, C., How, J.P.: Reachability analysis of neural feedback loops. IEEE Access9, 163938–163953 (2021). https://doi.org/10.1109/ACCESS.2021.3133370

    work page doi:10.1109/access.2021.3133370 2021

  16. [16]

    In: Proc

    Fisac, J.F., Lugovoy, N.F., Rubies-Royo, V ., Ghosh, S., Tomlin, C.J.: Bridging hamilton- jacobi safety analysis and reinforcement learning. In: Proc. IEEE Conf. Robotics and Au- tomation. pp. 8550–8556 (2019). https://doi.org/10.1109/ICRA.2019.8794107

    work page doi:10.1109/icra.2019.8794107 2019

  17. [17]

    Annual Review of Control, Robotics, and Autonomous Systems7(1) (2024)

    Hsu, K.C., Hu, H., Fisac, J.F.: The safety filter: A unified view of safety-critical control in autonomous systems. Annual Review of Control, Robotics, and Autonomous Systems7(1) (2024). https://doi.org/10.1146/annurev-control-071723-102940

    work page doi:10.1146/annurev-control-071723-102940 2024

  18. [18]

    In: Learning for Dynamics & Control

    Hsu, K.C., Nguyen, D.P., Fisac, J.F.: Isaacs: Iterative soft adversarial actor-critic for safety. In: Learning for Dynamics & Control. Proceedings of Machine Learning Research, vol. 211 (6 2023), https://proceedings.mlr.press/v211/hsu23a.html

    2023

  19. [19]

    In: Proc

    Hsu, K.C., Rubies-Royo, V ., Tomlin, C.J., Fisac, J.F.: Safety and liveness guarantees through reach-avoid reinforcement learning. In: Proc. Robotics: Science and Systems (7 2021). https: //doi.org/10.15607/RSS.2021.XVII.077

    work page doi:10.15607/rss.2021.xvii.077 2021

  20. [20]

    In: Proc

    Hu, H., Fazlyab, M., Morari, M., Pappas, G.J.: Reach-sdp: Reachability analysis of closed- loop systems with neural network controllers via semidefinite programming. In: Proc. IEEE Conf. Decision and Control. pp. 5929–5934 (2020). https://doi.org/10.1109/CDC42340. 2020.9304296 18 H. Hu

    work page doi:10.1109/cdc42340 2020

  21. [21]

    IEEE Robotics and Automation Letters7(2), 5591–5598 (2022)

    Hu, H., Nakamura, K., Fisac, J.F.: Sharp: Shielding-aware robust planning for safe and ef- ficient human-robot interaction. IEEE Robotics and Automation Letters7(2), 5591–5598 (2022). https://doi.org/10.1109/LRA.2022.3155229

    work page doi:10.1109/lra.2022.3155229 2022

  22. [22]

    In: Conf

    Hu, H., Zhang, Z., Nakamura, K., Bajcsy, A., Fisac, J.F.: Deception game: Closing the safety- learning loop in interactive robot autonomy. In: Conf. Robot Learning. Proceedings of Ma- chine Learning Research, vol. 229, pp. 3830–3850 (11 2023), https://proceedings.mlr.press/ v229/hu23b.html

    2023

  23. [23]

    Differential Games I (1954), https://www

    Isaacs, R.: Differential Games I: Introduction. Differential Games I (1954), https://www. rand.org/pubs/research_memoranda/RM1391.html

    1954

  24. [24]

    In: Proceedings of the 22nd ACM Interna- tional Conference on Hybrid Systems: Computation and Control

    Ivanov, R., Weimer, J., Alur, R., Pappas, G.J., Lee, I.: Verisig: verifying safety properties of hybrid systems with neural network controllers. In: Proceedings of the 22nd ACM Interna- tional Conference on Hybrid Systems: Computation and Control. pp. 169–178 (2019)

    2019

  25. [25]

    Econometrica: Journal of the Econometric Society pp

    Kajii, A., Morris, S.: The robustness of equilibria to incomplete information. Econometrica: Journal of the Econometric Society pp. 1283–1309 (1997). https://doi.org/10.2307/2171737

    work page doi:10.2307/2171737 1997

  26. [26]

    IEEE Robotics and Automation Letters (2025)

    Li, J., Lee, D., Lee, J., Dong, K.S., Sojoudi, S., Tomlin, C.: Certifiable reachability learning using a new lipschitz continuous value function. IEEE Robotics and Automation Letters (2025)

    2025

  27. [27]

    In: Learning for Dynamics & Control

    Lin, A., Bansal, S.: Verification of neural reachable tubes via scenario optimization and conformal prediction. In: Learning for Dynamics & Control. pp. 719–731. PMLR (2024), https://proceedings.mlr.press/v242/lin24a.html

    2024

  28. [28]

    IEEE Robotics and Automation Letters8(8), 5116–5123 (2023)

    Lindemann, L., Cleaveland, M., Shim, G., Pappas, G.J.: Safe planning in dynamic environ- ments using conformal prediction. IEEE Robotics and Automation Letters8(8), 5116–5123 (2023). https://doi.org/10.1109/LRA.2023.3292071

    work page doi:10.1109/lra.2023.3292071 2023

  29. [29]

    arXiv preprint arXiv:2505.23210 (2025)

    Lutkus, P., Wang, K., Lindemann, L., Tu, S.: Latent representations for control design with provable stability and safety guarantees. arXiv preprint arXiv:2505.23210 (2025). https:// doi.org/10.48550/arXiv.2505.23210

    work page doi:10.48550/arxiv.2505.23210 2025

  30. [30]

    arXiv preprint arXiv:2511.10586 (2025)

    Mirzaeedodangeh, O., Shekhtman, E., Matni, N., Lindemann, L.: Safe planning in interac- tive environments via iterative policy updates and adversarially robust conformal prediction. arXiv preprint arXiv:2511.10586 (2025)

    Pith/arXiv arXiv 2025

  31. [31]

    IEEE Transactions on Automatic Control 50(7), 947–957 (2005)

    Mitchell, I.M., Bayen, A.M., Tomlin, C.J.: A time-dependent hamilton-jacobi formulation of reachable sets for continuous dynamic games. IEEE Transactions on Automatic Control 50(7), 947–957 (2005). https://doi.org/10.1109/TAC.2005.851439

    work page doi:10.1109/tac.2005.851439 2005

  32. [32]

    Nakamura, K., Peters, L., Bajcsy, A.: Generalizing safety beyond collision-avoidance via latent-space reachability analysis (2025)

    2025

  33. [33]

    In: Conf

    Nguyen, D.P., Hsu, K.C., Yu, W., Tan, J., Fisac, J.F.: Gameplay filters: Robust zero-shot safety through adversarial imagination. In: Conf. Robot Learning. pp. 387–407. PMLR (2025), https://proceedings.mlr.press/v270/nguyen25a.html

    2025

  34. [34]

    Oh, D.D., Lidard, J., Hu, H., Sinhmar, H., Lazarski, E., Gopinath, D., Sumner, E.S., DeCas- tro, J.A., Rosman, G., Leonard, N.E., et al.: Safety with Agency: Human-Centered Safety Filter with Application to AI-Assisted Motorsports. Proc. Robotics: Science and Systems (2025). https://doi.org/10.48550/arXiv.2504.11717

    work page doi:10.48550/arxiv.2504.11717 2025

  35. [35]

    arXiv preprint arXiv:2510.18082 (2025)

    Oh, D.D., Nguyen, D.P., Hu, H., Fisac, J.F.: Provably optimal reinforcement learning under safety filtering. arXiv preprint arXiv:2510.18082 (2025)

    arXiv 2025

  36. [36]

    In: International workshop on hybrid systems: Computation and control

    Prajna, S., Jadbabaie, A.: Safety verification of hybrid systems using barrier certificates. In: International workshop on hybrid systems: Computation and control. pp. 477–492. Springer (2004)

    2004

  37. [37]

    H., & Belta, C

    Robey, A., Hu, H., Lindemann, L., Zhang, H., Dimarogonas, D.V ., Tu, S., Matni, N.: Learn- ing control barrier functions from expert demonstrations. In: Proc. IEEE Conf. Decision and Control. pp. 3717–3724 (2020). https://doi.org/10.1109/CDC42340.2020.9303785

    work page doi:10.1109/cdc42340.2020.9303785 2020

  38. [38]

    arXiv preprint arXiv:2505.00779 (2025) Verifiable Belief-Space Safety Filters 19

    Seo, J., Nakamura, K., Bajcsy, A.: Uncertainty-aware latent safety filters for avoiding out- of-distribution failures. arXiv preprint arXiv:2505.00779 (2025) Verifiable Belief-Space Safety Filters 19

    arXiv 2025

  39. [39]

    Shi, S., Jiang, L., Dai, D., Schiele, B.: Motion transformer with global intention lo- calization and local movement refinement. Advances in Neural Information Processing Systems35, 6531–6543 (2022), https://proceedings.neurips.cc/paper_files/paper/2022/hash/ 2ab47c960bfee4f86dfc362f26ad066a-Abstract-Conference.html

    2022

  40. [40]

    IEEE Robotics and Automation Letters8(11), 7833–7840 (2023)

    Strawn, K.J., Ayanian, N., Lindemann, L.: Conformal predictive safety filter for rl controllers in dynamic environments. IEEE Robotics and Automation Letters8(11), 7833–7840 (2023)

    2023

  41. [41]

    In: Proc

    Wabersich, K.P., Zeilinger, M.N.: Linear model predictive safety certification for learning- based control. In: Proc. IEEE Conf. Decision and Control. pp. 7130–7135 (2018). https: //doi.org/10.1109/CDC.2018.8619829

    work page doi:10.1109/cdc.2018.8619829 2018

  42. [42]

    In: Algorithmic Foundations of Robotics (2024), https://www.algorithmic-robotics.org/papers/ 45_MAGICS_Adversarial_RL_with_.pdf

    Wang, J., Hu, H., Nguyen, D.P., Fisac, J.F.: MAGICS: Adversarial RL with Minimax Actors Guided by Implicit Critic Stackelberg for Convergent Neural Synthesis of Robot Safety. In: Algorithmic Foundations of Robotics (2024), https://www.algorithmic-robotics.org/papers/ 45_MAGICS_Adversarial_RL_with_.pdf

    2024

  43. [43]

    In: Proc

    Xiao, W., Belta, C.: Control barrier functions for systems with high relative degree. In: Proc. IEEE Conf. Decision and Control. pp. 474–479 (2019). https://doi.org/10.1109/CDC40024. 2019.9029455

    work page doi:10.1109/cdc40024 2019