Pith. sign in

REVIEW 5 major objections 5 minor 28 references

Designing Control Barrier Function via Probabilistic Enumeration for Safe Reinforcement Learning Navigation

T0 review · 5 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a control barrier function built from probabilistically enumerated unsafe regions corrects unsafe actions of arbitrary reinforcement-learning navigation policies and achieves zero collisions.

desk verdict Promising combination of probabilistic enumeration and CBF filtering, but the central CBF derivative is wrong and the zero-collision guarantee as stated does not hold. read the letter →

arxiv 2504.21643 v1 pith:QMJMTRYT submitted 2025-04-30 cs.AI cs.RO

Luca Marzari , Francesco Trotti , Enrico Marchesini , Alessandro Farinelli This is my paper · ORCID
classification cs.AIcs.RO
keywords controlbarrierfunctionssafereinforcementlearningneuralnetworkverificationprobabilisticenumerationmaplessnavigationquadraticprogrammingsafetyfiltercollisionavoidanceautonomous
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

Trained neural-network navigation policies can produce actions that collide even when their average training reward looks good. The paper's claim is that probabilistic verification can find the bad regions ahead of time: enumeration marks parts of the policy's input space where the network outputs collision-prone or stuck actions, and those parts are removed to form a safe set. A control barrier function is then built on that set, and a quadratic-programming filter corrects any policy action that would leave it, without retraining or changing the policy. In simulated ground-robot and aquatic-drone tasks, and in a real ground-robot deployment, the filtered policies keep every tested trajectory collision-free while matching or improving goal-reaching success. If the claim holds, the framework is a plug-in safety layer for arbitrary learned navigation policies.

What carries the argument

The central object is the control barrier function $h(x)=\|p-p_{\mathrm{obs}}\|^2-d_{\mathrm{safe}}^2$, with $d_{\mathrm{safe}}=\max(\sigma,\|p-p_{\mathrm{area}}\|)$, where $p$ is the agent position, $p_{\mathrm{obs}}$ the nearest sensed obstacle, $p_{\mathrm{area}}$ the centroid of an enumerated unsafe region, and $\sigma$ a sensor-precision margin. The function is coupled to the safety-filter QP whose constraint is $L_f h(x)+L_g h(x)r+\alpha(h(x))\ge 0$. The machinery converts an offline verification result--which input-space regions produce dangerous or stuck actions--into an online action correction that is independent of how the policy was trained.

What would settle it

Compute $\dot h$ for a state with $\|p-p_{\mathrm{area}}\|>\sigma$ by differentiating $h(x)=\|p-p_{\mathrm{obs}}\|^2-\max(\sigma,\|p-p_{\mathrm{area}}\|)^2$ directly, and compare it with the paper's Eq. (8)-(9) constraint; wherever the two differ, there exists a nominal action that the QP would admit even though it drives $h$ below zero, so the claimed forward invariance would fail.

Watch

Extended reading notes

Core claim

The paper establishes a hierarchical recipe: offline, an interval-based probabilistic enumeration tool identifies regions of the policy's input space that lead to unsafe or stuck behavior; removing a finite union of such closed sets from a compact state-space subset yields a valid safe set for a control barrier function. Online, a quadratic program minimizes the magnitude of a corrective action $r^{cbf}=[v_1^{cbf},\omega_3^{cbf}]^T$ subject to the barrier constraint $\nabla h(x)f(x)+\nabla h(x)g(x)r+\alpha(h(x))\ge 0$, where $r$ is the sum of the policy's reference and the correction. When the policy's action already satisfies the constraint, the QP contributes nothing; otherwise it modulates the action to keep the agent inside the safe set, and a nonlinear model predictive controller tracks the corrected reference. The experiments report that all three baseline policies (unconstrained PPO, penalty-based PPO, and Lagrangian PPOLag) reach zero collisions under the filter, with success rates at or above the unfiltered baselines.

Load-bearing premise

The whole safety guarantee rests on the assumption that the barrier built from the enumerated unsafe regions is geometrically faithful and that its derivative is correct as written; the paper never specifies the mapping from enumerated regions to the centroid used in the barrier, and the derivative omits one term.

Editorial extensions

If this is right

  • Under the filter, every tested DRL baseline (PPO, PPO_penalty, PPOLag) achieves zero collisions in simulation, and the real-robot trial moves from about 25% collisions to 0%.
  • Filtered policies match or improve success: in the aquatic task, the penalty-based and Lagrangian baselines improve success by roughly 7 and 10 percentage points respectively.
  • Because enumeration targets neighborhoods of unsafe state-action pairs rather than the deployment map, the same safety layer transfers to unseen environments without retraining.
  • The QP correction also lifts agents out of stuck or local-minimum states that the base policy cannot resolve on its own.
  • The framework is orthogonal to the DRL training process, so it can be layered on top of already-trained policies as a safety filter.

Reading between the lines

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

  • A natural stress test is to replace the centroid $p_{\mathrm{area}}$ with the full enumerated unsafe set and recompute the barrier; the centroid simplification is the point where the geometric fidelity of the safe set could most plausibly break.
  • The same enumeration-to-CBF recipe should extend to other safety specifications, such as altitude keep-out zones or speed limits, whenever the enumerated unsafe set can be summarized by a differentiable distance function.
  • The zero-violation statement is measured over 100 simulated and 10 real trajectories; treating it as a deployment-wide guarantee would need a statistical or formal argument the paper does not make.
  • An immediate extension would rerun the enumeration periodically or track moving unsafe regions, letting the filter respond to dynamic environments while keeping the same QP structure.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. This paper proposes a hierarchical safety layer for DRL-based mapless navigation. After training a policy, probabilistic enumeration [17] is applied to manually defined neighborhoods of unsafe (state, action) pairs to identify unsafe regions; these regions are removed from compact subsets of the state space to form a safe set C. The paper then defines a candidate control barrier function h(x) = ||p - p_obs||^2 - d_safe^2 with d_safe = max(sigma, ||p - p_area||), and uses a QP to adjust the policy's velocity reference so that the CBF condition in Eq. (8) holds. An NMPC tracks the filtered reference. The framework is evaluated with PPO, PPO_penalty, and PPOLag in two Unity environments and on a real Turtlebot3, with the CBF-augmented configurations reported as achieving 0% collisions.

Significance. If the formal claim were sound, this would be a practically significant contribution: a policy-agnostic and environment-agnostic safety filter that converts collision-prone DRL navigation policies into collision-free ones, supported by both simulation and real-robot experiments. The idea of linking probabilistic enumeration of unsafe input regions to CBF-based control is an interesting bridge between formal verification and safe control. The empirical component is substantial, including a high-fidelity aquatic vehicle model and a physical deployment. However, the paper does not provide machine-checked proofs, a correctness argument for the central construction, or released code. Because the core derivation contains a load-bearing mathematical error, the formal safety guarantee is not established; at best the experiments demonstrate an empirical reduction in collisions for the tested scenarios.

major comments (5)
  1. [Sec. 4.2.1, Eq. (9)] Equation (9) is not the gradient of the function h defined in Eq. (7). Because d_safe = max(sigma, ||p - p_area||) depends on p, the correct position-space derivative in the active regime is 2(p - p_obs) - 2(p - p_area), whereas Eq. (9) omits the second term. The QP constraint in Eq. (8) is therefore the derivative condition for a different function, so the filter can certify commands that decrease the true h. This breaks the forward-invariance argument on which the zero-collision claim rests.
  2. [Sec. 4.1, Prop. 1] Proposition 1 is asserted without proof, and its supporting compactness claim is incorrect: if C_i is a finite union of closed sets within a compact set S_hat_i, then S_hat_i \ C_i is relatively open and need not be compact. The subsequent claim in Sec. 4.2.1 that a continuously differentiable h exists for the resulting set C is not demonstrated; Eq. (7) is introduced without showing that C equals {x : h(x) >= 0}.
  3. [Sec. 4.1 / Sec. 4.2.1, Eq. (7)] The paper never specifies how an enumerated unsafe region in the policy input space is converted to the position-space centroid p_area used in Eq. (7). The enumeration input includes beam scans, odometry, and target information, while h is a function of position only. Furthermore, for d_safe = ||p - p_area|| the zero-superlevel set of h is a half-plane, and for d_safe = sigma it is the exterior of a ball around p_obs; neither corresponds to a compact set minus a finite union of closed sets. Even with a corrected gradient, Eq. (8) would enforce invariance of a different set than the enumeration-derived C.
  4. [Sec. 4.2.1-4.2.2] The CBF condition in Eq. (8) is imposed on the filtered reference r, but the plant is driven by the NMPC input u*, which tracks r only approximately. Forward invariance requires the condition to hold along the closed loop with the actual input; no tracking-error bound, robustness margin, or input-to-state safety argument is supplied. This creates another gap between the QP certificate and the reported zero-collision results.
  5. [Sec. 5.1 and Sec. 6] Because the 0% collision figures in Fig. 7 and Fig. 8 are mediated by the flawed QP filter, these results do not substantiate the conclusion in Sec. 6 that the framework provides robust safety guarantees with zero violations. The experiments may support a weaker empirical claim of reduced collisions, but not the formal safety guarantee asserted in the abstract.
minor comments (5)
  1. [Sec. 4.2.1] The notation d_safe = max(sigma, ||p - p_area||2) should use a proper subscript for the Euclidean norm; as printed, the 2 is ambiguous.
  2. [Sec. 4.2.2] The word 'deefined' should be 'defined'.
  3. [Sec. 1 and Sec. 5] 'Lagragian' should be 'Lagrangian', and 'employ' should be 'employs'.
  4. [Sec. 4.1] The same symbol C_i is used for the safe subset and for the unsafe region (C_i = S_hat_i \ C_i), which is confusing and should be renamed.
  5. [Sec. 5.1, Fig. 7] Reporting 100% success and 0% collisions with zero spread for the Turtlebot3 condition is not informative; per-seed or per-trajectory statistics would be more helpful.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the enumeration-to-CBF pipeline is an external-tool-based design with an independent QP safety filter; the identified mathematical gaps are correctness risks, not self-referential reductions.

full rationale

The derivation chain is not circular. The unsafe regions come from the independently published probabilistic enumeration tool of reference [17], which is cited as an external component with its own provable guarantees; the present paper does not redefine that tool in terms of the zero-collision results it later reports, so the self-citations are load-bearing but not circular. The CBF parameters p_area and sigma are design inputs chosen from the enumerated regions and sensor precision, while the reported success and collision rates are empirical measurements obtained after deploying the resulting QP filter; no parameter is fitted to those reported outcomes, so the 'prediction' pattern does not apply. The main mathematical defects in the paper are correctness gaps rather than circularity: Eq. (9) omits the position dependence of d_safe, and Proposition 1 does not actually establish that the union of enumerated safe subsets equals the zero-superlevel set of h in Eq. (7). These are unsupported implications or derivative errors, not cases where an output quantity is identical to an input by construction. The paper's own limitation in Sec. 5.1, stating that the method does not endow the agent with navigation skills, further clarifies the scope of the claims and does not conceal a circular step. Because the central safetyset construction is an explicit design choice and the evaluation is measured against external baselines and a real robot, the appropriate finding is no significant circularity.

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

The framework introduces no new physical entities; the 'unsafe regions' derived from enumeration are data-driven constructs rather than invented objects. The central load-bearing elements are the free parameters sigma, p_area, and the manually defined input neighborhoods, plus the three listed assumptions about the enumeration tool, the existence of a smooth h, and the kinematic-to-dynamics transfer.

free parameters (3)
  • sigma (safe threshold) = not specified
    Manually defined based on sensor precision; it enters d_safe=max(sigma,||p-p_area||) and controls the size of the safety region around each enumerated unsafe area. No sensitivity analysis is reported.
  • p_area (centroid of enumerated unsafe region) = data-derived from enumeration
    Each enumerated unsafe input region is mapped to a position-space centroid that shapes the CBF through d_safe. The mapping is not derived and is central to the safety set definition.
  • Manually defined safety property neighborhoods = chosen per unsafe state
    For each unsafe (state, action) pair, the neighborhood S_i of the input space is manually defined before enumeration. The resulting safe set C and the CBF parameters depend on these hand-picked neighborhoods.
assumptions (3)
  • domain assumption Probabilistic enumeration [17] reliably identifies all relevant unsafe input regions with the stated probabilistic guarantees.
    The safe set computation in Sec. 4.1 inherits its correctness from the external enumeration tool; the tool's assumptions, error rates, and implementation details are not restated or re-validated in this paper.
  • ad hoc to paper A set obtained as a compact set minus a finite union of closed sets admits a continuously differentiable control barrier function.
    Proposition 1 asserts the enumeration-based union C is a valid safe set, and the text claims existence of a smooth h, but no construction or proof is given. The h actually used in Eq. (7) is not derived from that set, so the axiom is doing unsupported work.
  • domain assumption The kinematic model used in the QP barrier constraint adequately represents the plant for safety purposes, and the NMPC tracks the corrected reference with negligible error.
    The CBF in Sec. 4.2.1 is built on the kinematic components (Eq. 4), while the real plant follows the 6-DOF dynamics (Eq. 5) via an NMPC tracking layer (Sec. 4.2.2). Forward invariance of the kinematic CBF does not automatically transfer to the closed-loop system with tracking error and model mismatch.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Designing Control Barrier Function via Probabilistic Enumeration for Safe Reinforcement Learning Navigation." pith.science (2026). https://pith.science/paper/QMJMTRYT

@misc{pith2026250421643,
  author       = {Pith},
  title        = {Pith review of: Designing Control Barrier Function via Probabilistic Enumeration for Safe Reinforcement Learning Navigation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QMJMTRYT}},
  note         = {Machine review of arXiv:2504.21643}
}
read the original abstract

Achieving safe autonomous navigation systems is critical for deploying robots in dynamic and uncertain real-world environments. In this paper, we propose a hierarchical control framework leveraging neural network verification techniques to design control barrier functions (CBFs) and policy correction mechanisms that ensure safe reinforcement learning navigation policies. Our approach relies on probabilistic enumeration to identify unsafe regions of operation, which are then used to construct a safe CBF-based control layer applicable to arbitrary policies. We validate our framework both in simulation and on a real robot, using a standard mobile robot benchmark and a highly dynamic aquatic environmental monitoring task. These experiments demonstrate the ability of the proposed solution to correct unsafe actions while preserving efficient navigation behavior. Our results show the promise of developing hierarchical verification-based systems to enable safe and robust navigation behaviors in complex scenarios.

Figures

Figures reproduced from arXiv: 2504.21643 by the authors.

Figure 1
Figure 1. Overview of the proposed approach. deployment of safe policies and correcting their decisions when detecting an unsafe situation remains an open research question. We address this gap by leveraging the advantages of probabilistic verification and CBF-based control, introducing a framework that guarantees safe DRL-based navigation policies and correct policy actions in potentially unsafe situations ( [PITH_FULL_IMAG… view at source ↗
Figure 2
Figure 2. Illustrative example of a safety property in aquatic mapless navigation: the left shows an unsafe state near [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Block diagram of the hierarchical architecture. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Explanatory example of the relationship between the enumeration process and the CBF. On the left, we report [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Linear and angular velocities tracking via NMPC. Red dashed lines are the reference [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Indoor mobile navigation scenario and aquatic navigation task. On the right, density maps of unsafe regions [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Training and evaluation results. On the left, we report the learning phase using PPO, PPO_penalty, and [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Real-world experiments setup with PPOLag (first row) and PPOLag+CBF (second row). [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 21 canonical work pages

  1. [17]

    Enumerating safe regions in deep neural networks with provable probabilistic guarantees

    Luca Marzari, Davide Corsi, Enrico Marchesini, Alessandro Farinelli, and Ferdinando Cicalese. Enumerating safe regions in deep neural networks with provable probabilistic guarantees. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 21387–21394, 2024

    work page 2024

  2. [1]

    Achieving sustainable aquaculture: Historical and current perspectives and future needs and challenges

    Claude E Boyd, Louis R D’Abramo, Brent D Glencross, David C Huyben, Lorenzo M Juarez, George S Lockwood, Aaron A McNevin, Albert GJ Tacon, Fabrice Teletchea, Joseph R Tomasso Jr, et al. Achieving sustainable aquaculture: Historical and current perspectives and future needs and challenges. Journal of the world aquaculture society, 51(3):578–633, 2020

    work page 2020

  3. [2]

    Aspects of precision agriculture

    Francis J Pierce and Peter Nowak. Aspects of precision agriculture. Advances in agronomy, 67:1–85, 1999

    work page 1999

  4. [3]

    The challenges and opportunities for the use of robotic autonomous robotic systems in support of the blue economy

    Damien Guihen. The challenges and opportunities for the use of robotic autonomous robotic systems in support of the blue economy. In International Conference on Offshore Mechanics and Arctic Engineering, volume 86922. American Society of Mechanical Engineers, 2023

    work page 2023

  5. [4]

    Dec-mcts: Decentralized planning for multi-robot active perception

    Graeme Best, Oliver M Cliff, Timothy Patten, Ramgopal R Mettu, and Robert Fitch. Dec-mcts: Decentralized planning for multi-robot active perception. The International Journal of Robotics Research, 38(2-3):316–337, 2019

    work page 2019

  6. [5]

    A markov decision process approach for decentralized uav formation path planning

    Francesco Trotti, Alessandro Farinelli, and Riccardo Muradore. A markov decision process approach for decentralized uav formation path planning. In 2024 European Control Conference (ECC), pages 436–441. IEEE, 2024

    work page 2024

  7. [6]

    Path re-planning with stochastic obstacle modeling: A monte carlo tree search approach

    Francesco Trotti, Alessandro Farinelli, and Riccardo Muradore. Path re-planning with stochastic obstacle modeling: A monte carlo tree search approach. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 8017–8022. IEEE, 2024

    work page 2024

  8. [7]

    An online path planner based on pomdp for uavs

    Francesco Trotti, Alessandro Farinelli, and Riccardo Muradore. An online path planner based on pomdp for uavs. In 2023 European Control Conference (ECC), pages 1–6. IEEE, 2023

    work page 2023

Show all 28 references
  1. [8]

    Safety gymnasium: A unified safe reinforcement learning benchmark

    Jiaming Ji, Borong Zhang, Jiayi Zhou, Xuehai Pan, Weidong Huang, Ruiyang Sun, Yiran Geng, Yifan Zhong, Josef Dai, and Yaodong Yang. Safety gymnasium: A unified safe reinforcement learning benchmark. In Thirty-seventh Conference on Neural Information Processing Systems Datasets...

    2023

  2. [9]

    Online safety property collection and refinement for safe deep reinforcement learning in mapless navigation

    Luca Marzari, Enrico Marchesini, and Alessandro Farinelli. Online safety property collection and refinement for safe deep reinforcement learning in mapless navigation. In 2023 IEEE International Conference on Robotics and Automation (ICRA), pages 7133–7139. IEEE, 2023

    2023

  3. [10]

    Omnisafe: An infrastructure for accelerating safe reinforcement learning research

    Jiaming Ji, Jiayi Zhou, Borong Zhang, Juntao Dai, Xuehai Pan, Ruiyang Sun, Weidong Huang, Yiran Geng, Mickel Liu, and Yaodong Yang. Omnisafe: An infrastructure for accelerating safe reinforcement learning research. Journal of Machine Learning Research, 25(285):1–6, 2024

    2024

  4. [11]

    Donti, Changliu Liu, and Enrico Marchesini

    Luca Marzari, Priya L. Donti, Changliu Liu, and Enrico Marchesini. Improving policy optimization via ε-retrain. In Proceedings of the 2025 International Conference on Autonomous Agents and Multiagent Systems, 2025

    2025

  5. [12]

    Intriguing properties of neural networks

    Christian Szegedy, Wojciech Zaremba, Ilya Sutskever, Joan Bruna, Dumitru Erhan, Ian Goodfellow, and Rob Fergus. Intriguing properties of neural networks. arXiv preprint arXiv:1312.6199, 2013

    2013 arXiv

  6. [13]

    Verifying learning-based robotic navigation systems

    Guy Amir, Davide Corsi, Raz Yerushalmi, Luca Marzari, David Harel, Alessandro Farinelli, and Guy Katz. Verifying learning-based robotic navigation systems. In 29th International Conference, TACAS 2023 , pages 607–627. Springer, 2023. 10

    2023

  7. [14]

    Algorithms for verifying deep neural networks

    Changliu Liu, Tomer Arnon, Christopher Lazarus, Christopher Strong, Clark Barrett, Mykel J Kochenderfer, et al. Algorithms for verifying deep neural networks. Foundations and Trends® in Optimization, 4(3-4):244–404, 2021

    2021

  8. [15]

    Proven: Verifying robustness of neural networks with a probabilistic approach

    Lily Weng, Pin-Yu Chen, Lam Nguyen, Mark Squillante, Akhilan Boopathy, Ivan Oseledets, and Luca Daniel. Proven: Verifying robustness of neural networks with a probabilistic approach. In International Conference on Machine Learning, pages 6727–6736. PMLR, 2019

    2019

  9. [16]

    The #dnn-verification problem: Counting unsafe inputs for deep neural networks

    Luca Marzari, Davide Corsi, Ferdinando Cicalese, and Alessandro Farinelli. The #dnn-verification problem: Counting unsafe inputs for deep neural networks. In Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, IJCAI-23, pages 217–224, 2023

    2023

  10. [18]

    Control barrier function based quadratic programs for safety critical systems

    Aaron D Ames, Xiangru Xu, Jessy W Grizzle, and Paulo Tabuada. Control barrier function based quadratic programs for safety critical systems. IEEE Transactions on Automatic Control, 62(8):3861–3876, 2016

    2016

  11. [19]

    Safe control under input limits with neural control barrier functions

    Simin Liu, Changliu Liu, and John Dolan. Safe control under input limits with neural control barrier functions. In Conference on Robot Learning, pages 1970–1980. PMLR, 2023

    1970

  12. [20]

    How to train your neural control barrier function: Learning safety filters for complex input-constrained systems

    Oswin So, Zachary Serlin, Makai Mann, Jake Gonzales, Kwesi Rutledge, Nicholas Roy, and Chuchu Fan. How to train your neural control barrier function: Learning safety filters for complex input-constrained systems. In 2024 IEEE International Conference on Robotics and Automation...

    2024

  13. [21]

    Verification of neural control barrier functions with symbolic derivative bounds propagation

    Hanjiang Hu, Yujie Yang, Tianhao Wei, and Changliu Liu. Verification of neural control barrier functions with symbolic derivative bounds propagation. arXiv preprint arXiv:2410.16281, 2024

    2024 arXiv

  14. [22]

    Responsive safety in reinforcement learning by pid lagrangian methods

    Adam Stooke, Joshua Achiam, and Pieter Abbeel. Responsive safety in reinforcement learning by pid lagrangian methods. In International Conference on Machine Learning (ICML), 2020

    2020

  15. [23]

    Safe deep reinforcement learning by verifying task-level properties

    Enrico Marchesini, Luca Marzari, Alessandro Farinelli, and Christopher Amato. Safe deep reinforcement learning by verifying task-level properties. In Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems, page 1466–1475, 2023

    2023

  16. [24]

    Towards aircraft autonomy using a pomdp-based planner

    Francesco Trotti, Alessandro Farinelli, and Riccardo Muradore. Towards aircraft autonomy using a pomdp-based planner. In 2024 American Control Conference (ACC), pages 2399–2404. IEEE, 2024

    2024

  17. [25]

    Modelling the dynamics of ships with different propulsion systems for control purpose

    Witold Gierusz. Modelling the dynamics of ships with different propulsion systems for control purpose. Polish Maritime Research, (1):31–36, 2016

    2016

  18. [26]

    Nonlinear model predictive control

    Lars Grüne, Jürgen Pannek, Lars Grüne, and Jürgen Pannek. Nonlinear model predictive control. Springer, 2017

    2017

  19. [27]

    A smooth converse lyapunov theorem for robust stability

    Yuandan Lin, Eduardo D Sontag, and Yuan Wang. A smooth converse lyapunov theorem for robust stability. SIAM Journal on Control and Optimization, 34(1):124–160, 1996

    1996

  20. [28]

    Proximal policy optimization algorithms

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347, 2017. 11

    2017 arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.