Pith. sign in

REVIEW 3 major objections 5 minor 34 references

MAD-PINN: A Decentralized Physics-Informed Machine Learning Framework for Safe and Optimal Multi-Agent Control

T0 review · 3 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read A single physics-informed neural network, trained on just three agents, can co-optimize safety and performance for swarms of hundreds.

desk verdict A practically useful decentralized PINN framework for multi-agent SC-OCP, but the safety guarantee is not actually proven — the gap between centralized training and decentralized execution is real and unaddressed. read the letter →

arxiv 2509.23960 v2 pith:MYRKIJYZ submitted 2025-09-28 cs.RO cs.AI

Manan Tayal , Aditya Singh , Shishir Kolathaya , Somil Bansal This is my paper
classification cs.ROcs.AI
keywords multi-agentsafecontrolphysics-informedneuralnetworkstate-constrainedoptimalHamilton-Jacobireachabilitydecentralizedpolicyepigraphreformulationcollisionavoidancescalableautonomy
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

The paper claims that the hard problem of multi-agent state-constrained optimal control can be split: train one value function on a small homogeneous system, then deploy it independently on each agent using only local observations. The value function is the solution of an epigraph-reformulated Hamilton-Jacobi-Bellman PDE, approximated by a physics-informed neural network with the terminal safety boundary hard-coded. During deployment, agents pick their most safety-critical neighbors using a pairwise reachability value function, then execute the policy in a receding horizon. If correct, this shows that safe, low-cost navigation can scale to hundreds of agents without retraining, because observation size is fixed and each agent runs the same policy.

What carries the argument

The load-bearing object is the epigraph-reformulated auxiliary value function V̂(t, o, z): a function of state and an auxiliary cost-bound z that is the unique viscosity solution to an HJB PDE (equation 8). It is approximated by a physics-informed neural network whose output is structured as terminal condition + (T−t)·residual, so the safety terminal boundary max(φ(o)−z, g(o)) is satisfied exactly. A second learned object, the pairwise safety value function from HJ reachability, ranks neighbor risk and selects the fixed-size neighbor set. Policy synthesis inverts V̂ via binary search over z to get the minimal admissible cost, then extracts the feedback control from the Hamiltonian minimizer.

What would settle it

Run the 256-agent experiment with more random seeds and log every collision; if a nontrivial fraction of collisions involve at least one pair of agents that were not selected as neighbors in the receding-horizon update, the locality assumption is violated. Alternatively, construct a head-on crossing where two non-neighbor agents both converge on the ego agent's future position from outside the observation radius; if the ego agent collides, the claimed globally consistent safety is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a single distributed value function — trained on a three-agent system — can be deployed in a decentralized manner to any number of homogeneous agents and still co-optimize safety and performance. The paper argues this works because the decentralized SC-OCP has fixed observation size, making the problem dimension independent of total agent count; the epigraph reformulation encodes both cost and safety in one auxiliary value function; and the reachability-based neighbor selection prioritizes safety-critical interactions. Experiments report near-100% safety rates up to 16 agents and above 96% safety rates at 256 agents, with cumulative cost increasing only proportional

Load-bearing premise

The framework assumes that interactions with agents outside the chosen neighbor set can be ignored without compromising safety; no theorem or bound is provided to guarantee this in dense crowds.

Editorial extensions

If this is right

  • If the value function truly generalizes across agent count, then adding agents to a swarm costs no new training: the same network and the same local observation code handle any N.
  • Because neighbor selection is based on reachability rather than raw distance, agents anticipate collisions from fast-approaching or velocity-aligned neighbors, not just nearby ones.
  • The receding-horizon execution makes the policy robust to model mismatch and sensor noise, since the neighbor set and optimization are recomputed online.
  • Safety and performance are unified in one scalar trade-off z: the agent picks the smallest cost bound that still avoids the unsafe set, so conservatism is controlled not by reward weights but by the constraint structure itself.

Reading between the lines

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

  • The paper's locality assumption — that ignoring agents beyond the selected n is safe — is argued by homogeneity and fixed observation size, but no formal error bound is given; a natural next step is a certificate that bounds safety-violation probability as a function of density and n.
  • If the method transfers to heterogeneous agents or learned dynamics, it would effectively turn multi-agent SC-OCP into a per-agent optimal-control problem with a learned interaction-aware cost, which could apply to warehouse robotics and aerial swarms.
  • The reported degradation from 100% safe scenarios at 3 agents to 85% at 256 suggests a testable scaling law: one could measure how safe-scenario percentage decays with agent density and use that to pick the required neighbor count for a target safety level.
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 MAD-PINN, a decentralized physics-informed machine learning framework for multi-agent state-constrained optimal control (SC-OCP). The method trains, on a small 3-agent system, an epigraph-based auxiliary value function satisfying an HJB PDE, using a boundary-aware PINN. For deployment, each agent selects a fixed number of safety-critical neighbors via a DeepReach pairwise value function and executes a receding-horizon policy derived from the shared local value function. Experiments on multi-agent drone navigation compare against DEF-MARL, SafeMARL, and MPPI, reporting superior cumulative-cost/safety trade-offs, high safety rates up to 256 agents, and an ablation supporting the reachability-based neighbor selection. The central claim is that a single locally trained value function, deployed per-agent with local observations, yields near-collision-free, low-cost navigation for arbitrarily many homogeneous agents.

Significance. If the safety guarantee were rigorously established, the paper would make a useful contribution: it demonstrates that an SC-OCP value function trained on a small cluster can be deployed in a decentralized way with good empirical performance, and the receding-horizon neighbor selection is a sensible engineering idea. The experimental comparison is thorough, with standard errors across seeds, and the ablation in Table III directly supports the value of the proposed neighbor-selection mechanism. The boundary-aware network structure that exactly encodes the terminal condition is also a clean design choice. However, the paper's load-bearing claim of 'strict safety guarantees' and 'globally consistent' solutions is not supported by the presented theory. The missing arguments are not cosmetic: they concern whether the decentralized execution actually solves or approximates the intended safe-optimal control problem, and whether the learned PINN approximation carries any certificate. These gaps currently put the formal framing of the paper ahead of its evidence.

major comments (3)
  1. [Section III-C and Section II-A] The bridge from centralized local training to decentralized execution is missing. In Eq. (4), the value function V_d is defined for a single local cluster, and the certificate in Eqs. (5)-(6) applies to the joint control u_d of that cluster. In Section III-C, each agent i extracts the i-component of the joint minimizer for its own observation. But other agents are not executing i's minimizer; they execute controls from their own local Hamiltonians. The realized multi-agent trajectory is therefore not the joint-optimal trajectory for any single cluster, so the local certificate \hat V_d <= 0 does not apply to the actual system. The paper provides no fixed-point, equilibrium, or bounded-mismatch argument to close this gap. This is independent of PINN approximation error and of truncation to n neighbors. The empirical degradation from 100% safe scenarios at N=3 to 86% at N=16 (Table I) and
  2. [Section II-B, Eq. (8), and Section III-A] The formal safety claim relies on the auxiliary value function being the unique viscosity solution of the HJB PDE (8), which is only true under assumptions A1-A4 from [23]. The paper neither states these assumptions nor verifies that the double-integrator drone system satisfies them. Moreover, the learned \hat V_theta is a PINN approximation minimizing the PDE residual (10) at sampled collocation points; no convergence, verification, or worst-case error bound is provided. The conclusion section acknowledges this by listing 'quantify the approximation error of the auxiliary value function' as future work. Consequently, the abstract's claim of 'strict safety guarantees' is not established. At minimum, the paper should either prove or explicitly cite a verification result for \hat V_theta, or reframe the contribution as an empirical heuristic without formal guarantees.
  3. [Section II-A and Section IV-D] The assumption that fixed-size local observations of n neighbors fully determine each agent's safe-optimal decision is asserted but not analyzed. Section II-A claims this yields a 'globally consistent solution' independent of N, but no theorem or bound quantifies the effect of ignoring non-neighbor agents. Receding-horizon re-selection of neighbors does not eliminate the possibility that an ignored agent causes a collision. The empirical results show this effect directly: safe-scenario rate drops from 100% at N=3 to 86% at N=16 and 85% at N=256. A formal or at least quantitative characterization of this degradation is needed if the paper is to claim scalability with guaranteed safety.
minor comments (5)
  1. [Section IV-C] The experimental setup says 'all algorithms over 100 distinct initial conditions across 5 seeds', but Section IV-D.2 and Table II use only '20 distinct initial conditions'. Please state the number of seeds for the large-scale experiments as well.
  2. [Section II-B] The assumptions A1-A4 from [23] are referenced but never listed. Since the viscosity solution result in Eq. (8) depends on them, the paper should at least state them in an appendix and comment on their satisfaction for the considered dynamics.
  3. [Eq. (10)] The notation in the loss function is inconsistent: H(t, x̂) is defined with x̂ but the second line uses H(ˆok, tk) and H(ˆo, t). Please standardize the argument order.
  4. [Abstract and Section I] The phrases 'strict safety guarantees' and 'formal safety guarantees' are used despite the reported safety rates being below 100% in several configurations (e.g., 98.3% safety rate at N=16 in Table I). Consider distinguishing between a certified guarantee and an empirical safety rate.
  5. [Figure 4] The heatmap caption says 'with respect to the ego agent's position coordinates', but the text does not specify the fixed states of the other agents beyond their positions/velocities. Please clarify what is being plotted and how the goal location is encoded.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the value function is learned from the HJB PDE and evaluated against external collision metrics; self-citations are tools, not load-bearing premises.

full rationale

The paper's derivation chain is self-contained. The epigraph reformulation (Eqs. 5-6) and the HJB PDE characterization (Eqs. 8-9) come from external references ([23], [25]), and the network parameterization enforces the terminal condition exactly by construction, with the residual loss (Eq. 10) being the PDE residual. The policy is then extracted from the learned value function via the Hamiltonian, a standard and non-circular step. The decentralized value function is trained on a fixed-size local cluster and deployed per-agent; no fitted parameter is later renamed as a prediction, and the safety metrics are external (actual collision distances), not the value function's own level sets. The self-citations ([24], [28], [29]) provide prior tools and training strategies, but the central multi-agent claim is evaluated against external baselines (DEF-MARL, SafeMARL, MPPI) and an ablation study, so the results are empirically falsifiable. The main gap—the absence of a proof that decentralized composition of locally optimal controls preserves the local safety certificate—is a soundness/completeness limitation, not circularity. A minor burden is the reliance on self-authored components for the PINN machinery, but this does not make the central result reduce to its inputs.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim relies on a chain of assumptions: homogeneity, local-observation sufficiency, the epigraph/HJB characterization from [23], and PINN convergence. The free parameters are hand-chosen design values that affect the safety-performance trade-off. No new physical entities are introduced; the auxiliary variable z is a mathematical artifact borrowed from the epigraph reformulation of [23],[25].

free parameters (4)
  • local neighbor count n = 2
    The value function is trained with 3 agents (2 neighbors) and deployed with fixed n=2; this design choice determines state dimension and safety horizon, with no analysis of sufficiency.
  • observation radius r_obs = 0.5
    Hand-chosen; defines the local observation field in training and deployment, and agents ignore interactions beyond this radius.
  • training horizon T = 0.2s
    Hand-chosen for the SC-OCP; receding-horizon replanning compensates, but the learned value function is optimal only relative to this horizon.
  • neural network architecture and learning rate = MLP 3x256, sine activation, Adam lr 2e-5
    Hand-selected with no sensitivity analysis or convergence guarantee.
assumptions (5)
  • domain assumption All agents are homogeneous with identical dynamics, costs, and observation structure
    Section II states a 'homogeneous multi-agent system' and separable identical costs, which justifies a single shared value function V_d for all agents.
  • ad hoc to paper Local observation of n neighbors fully determines each agent's safe-optimal decision; interactions outside the neighbor set can be ignored
    Section II-A claims fixed observation size makes dimension independent of N and yields a 'globally consistent solution', but no theorem establishes this; it is the central sufficiency assumption.
  • standard math Epigraph reformulation and HJB-PDE characterization of SC-OCP hold under assumptions A1-A4 from [23]
    Section II-B cites [23]; the assumptions are not stated or checked for the double-integrator drone system.
  • domain assumption Neural network training converges to the unique viscosity solution of the HJB-PDE
    Section III-A adopts the curriculum/PINN scheme; no approximation-error bound is provided, and the paper lists quantifying approximation error as future work.
  • ad hoc to paper Reachability-based neighbor selection using pairwise V_s preserves global safety when neighbors are re-selected in a receding horizon
    Section III-B and III-C; empirically motivated, not proven.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MAD-PINN: A Decentralized Physics-Informed Machine Learning Framework for Safe and Optimal Multi-Agent Control." pith.science (2026). https://pith.science/paper/MYRKIJYZ

@misc{pith2026250923960,
  author       = {Pith},
  title        = {Pith review of: MAD-PINN: A Decentralized Physics-Informed Machine Learning Framework for Safe and Optimal Multi-Agent Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MYRKIJYZ}},
  note         = {Machine review of arXiv:2509.23960}
}
read the original abstract

Co-optimizing safety and performance in large-scale multi-agent systems remains a fundamental challenge. Existing approaches based on multi-agent reinforcement learning (MARL), safety filtering, or Model Predictive Control (MPC) either lack strict safety guarantees, suffer from conservatism, or fail to scale effectively. We propose MAD-PINN, a decentralized physics-informed machine learning framework for solving the multi-agent state-constrained optimal control problem (MASC-OCP). Our method leverages an epigraph-based reformulation of SC-OCP to simultaneously capture performance and safety, and approximates its solution via a physics-informed neural network. Scalability is achieved by training the SC-OCP value function on reduced-agent systems and deploying them in a decentralized fashion, where each agent relies only on local observations of its neighbours for decision-making. To further enhance safety and efficiency, we introduce an Hamilton-Jacobi (HJ) reachability-based neighbour selection strategy to prioritize safety-critical interactions, and a receding-horizon policy execution scheme that adapts to dynamic interactions while reducing computational burden. Experiments on multi-agent navigation tasks demonstrate that MAD-PINN achieves superior safety-performance trade-offs, maintains scalability as the number of agents grows, and consistently outperforms state-of-the-art baselines.

Figures

Figures reproduced from arXiv: 2509.23960 by the authors.

Figure 1
Figure 1. We propose MAD-PINN – a framework for safe and optimal multi-agent control. MAD-PINN is divided into two phases: (1) Centralized training, where we learn the auxiliary epigraph-based value function Vˆ θ using a boundary-aware PINN, and the pairwise safety value function Vs via DeepReach; (2) Decentralized deployment, where each agent selects its safety-critical neighbours using Vs and executes the policy in a recedi… view at source ↗
Figure 2
Figure 2. Comparison of cumulative costs and safety rates across testing environments with [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Snapshots of multi-agent navigation trajectories at different times using MAD-PINN and baselines. Agents are [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Heatmap of the learned value function with respect to the ego agent’s position coordinates. The other two agents are at [−0.3, 0.4] and [−0.3, −0.4], both moving with velocity [−1, 0], while the ego agent moves with velocity [1, 0] toward its goal at [−0.5, 0]. which a…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

34 extracted references · 1 canonical work pages

  1. [23]

    A general hamilton- jacobi framework for non-linear state-constrained control problems,

    A. Altarovici, O. Bokanowski, and H. Zidani, “A general hamilton- jacobi framework for non-linear state-constrained control problems,” ESAIM: Control, Optimisation and Calculus of Variations, vol. 19, no. 2, pp. 337–357, 2013

    2013

  2. [1]

    Predictive control of aerial swarms in cluttered environments,

    E. Soria, F. Schiano, and D. Floreano, “Predictive control of aerial swarms in cluttered environments,”Nature Machine Intelligence, vol. 3, no. 6, pp. 545–554, 2021

    2021

  3. [2]

    Multi-agent reinforcement learning in intelligent transportation systems: A comprehensive survey,

    R. Donatus, K. Ter, O.-O. Ajayi, and D. Udekwe, “Multi-agent reinforcement learning in intelligent transportation systems: A comprehensive survey,” 2025. [Online]. Available: https://arxiv.org/ abs/2508.20315

    arXiv 2025

  4. [3]

    Distributed optimization in multi-agent robotics for industry 4.0 warehouses,

    A. Kattepur, H. K. Rath, A. Simha, and A. Mukherjee, “Distributed optimization in multi-agent robotics for industry 4.0 warehouses,” inProceedings of the 33rd Annual ACM Symposium on Applied Computing, 2018, pp. 808–815

    2018

  5. [4]

    Multi-agent actor-critic for mixed cooperative-competitive environments,

    R. Lowe, Y . I. Wu, A. Tamar, J. Harb, O. Pieter Abbeel, and I. Mor- datch, “Multi-agent actor-critic for mixed cooperative-competitive environments,”Advances in neural information processing systems, vol. 30, 2017

    2017

  6. [5]

    The surprising effectiveness of ppo in cooperative multi-agent games,

    C. Yu, A. Velu, E. Vinitsky, J. Gao, Y . Wang, A. Bayen, and Y . Wu, “The surprising effectiveness of ppo in cooperative multi-agent games,”Advances in neural information processing systems, vol. 35, pp. 24 611–24 624, 2022

    2022

  7. [6]

    Scalable multi-agent reinforcement learning through intelligent information aggregation,

    S. Nayak, K. Choi, W. Ding, S. Dolan, K. Gopalakrishnan, and H. Balakrishnan, “Scalable multi-agent reinforcement learning through intelligent information aggregation,” inInternational Conference on Machine Learning. PMLR, 2023, pp. 25 817–25 833

    2023

  8. [7]

    Altman,Constrained Markov Decision Processes, ser

    E. Altman,Constrained Markov Decision Processes, ser. Stochastic Modeling Series. Taylor & Francis, 1999. [Online]. Available: https://books.google.co.in/books?id=3X9S1NM2iOgC

    1999

Show all 34 references
  1. [8]

    Safe multi-agent reinforcement learning for multi-robot control,

    S. Gu, J. G. Kuba, Y . Chen, Y . Du, L. Yang, A. Knoll, and Y . Yang, “Safe multi-agent reinforcement learning for multi-robot control,” Artificial Intelligence, vol. 319, p. 103905, 2023

    2023

  2. [9]

    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

  3. [10]

    On safety and liveness filtering using hamilton–jacobi reachability analysis,

    J. Borquez, K. Chakraborty, H. Wang, and S. Bansal, “On safety and liveness filtering using hamilton–jacobi reachability analysis,”IEEE Transactions on Robotics, vol. 40, pp. 4235–4251, 2024

    2024

  4. [11]

    Data-driven safety filters: Hamilton-jacobi reachability, control barrier functions, and predictive methods for uncertain systems,

    K. P. Wabersich, A. J. Taylor, J. J. Choi, K. Sreenath, C. J. Tom- lin, A. D. Ames, and M. N. Zeilinger, “Data-driven safety filters: Hamilton-jacobi reachability, control barrier functions, and predictive methods for uncertain systems,”IEEE Control Systems Magazine, vol. 43, ...

    2023

  5. [12]

    The safety filter: A unified view of safety-critical control in autonomous systems,

    K.-C. Hsu, H. Hu, and J. F. Fisac, “The safety filter: A unified view of safety-critical control in autonomous systems,”Annual Review of Control, Robotics, and Autonomous Systems, vol. 7, no. V olume 7, 2024, pp. 47–72, 2024. [Online]. Available: https://www.annualreviews.org/...

    2024 doi

  6. [13]

    Resolving conflicting constraints in multi-agent reinforcement learning with layered safety,

    J. J. Choi, J. J. Aloor, J. Li, M. G. Mendoza, H. Balakrishnan, and C. J. Tomlin, “Resolving conflicting constraints in multi-agent reinforcement learning with layered safety,” inProceedings of Robotics: Science and Systems, Los Angeles, USA, June 2025. [Online]. Available: ht...

    2025

  7. [14]

    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,” inConference on Decision and Control (CDC). IEEE, 2024

    2024

  8. [15]

    Model predictive control: Theory and practice—a survey,

    C. E. Garc ´ıa, D. M. Prett, and M. Morari, “Model predictive control: Theory and practice—a survey,”Automatica, vol. 25, no. 3, pp. 335–348, 1989. [Online]. Available: https://www.sciencedirect.com/ science/article/pii/0005109889900022

    1989

  9. [16]

    Gr ¨une, J

    L. Gr ¨une, J. Pannek, L. Gr ¨une, and J. Pannek,Nonlinear model predictive control. Springer, 2017

    2017

  10. [17]

    Information-theoretic model predictive control: Theory and applica- tions to autonomous driving,

    G. Williams, P. Drews, B. Goldfain, J. M. Rehg, and E. A. Theodorou, “Information-theoretic model predictive control: Theory and applica- tions to autonomous driving,”IEEE Transactions on Robotics, vol. 34, no. 6, pp. 1603–1622, 2018

    2018

  11. [18]

    Multi-agent path integral control for interaction-aware motion planning in urban canals,

    L. Streichenberg, E. Trevisan, J. J. Chung, R. Siegwart, and J. Alonso- Mora, “Multi-agent path integral control for interaction-aware motion planning in urban canals,” in2023 IEEE International Conference on Robotics and Automation (ICRA), 2023, pp. 1379–1385

    2023

  12. [19]

    Multi-agent path integral control for interaction-aware motion planning in urban canals,

    ——, “Multi-agent path integral control for interaction-aware motion planning in urban canals,” in2023 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2023, pp. 1379–1385

    2023

  13. [20]

    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

    2024 arXiv

  14. [21]

    Cp-ncbf: A conformal prediction-based approach to syn- thesize verified neural control barrier functions,

    ——, “Cp-ncbf: A conformal prediction-based approach to syn- thesize verified neural control barrier functions,”arXiv preprint arXiv:2503.17395, 2025

    2025 arXiv

  15. [22]

    Optimal control with state-space constraint i,

    H. M. Soner, “Optimal control with state-space constraint i,”SIAM Journal on Control and Optimization, vol. 24, no. 3, pp. 552–561,

  16. [24]

    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,” inForty-second International Conference on Machine Learning, 2025

    2025

  17. [25]

    Boyd and L

    S. Boyd and L. Vandenberghe,Convex optimization. Cambridge university press, 2004

    2004

  18. [26]

    A toolbox of level set methods,

    I. Mitchell, “A toolbox of level set methods,”http://www. cs. ubc. ca/mitchell/ToolboxLS/toolboxLS.pdf, 2004

    2004

  19. [27]

    hj reachability: Hamilton-Jacobi reachability analysis in JAX,

    E. Schmerling, “hj reachability: Hamilton-Jacobi reachability analysis in JAX,”https://github.com/StanfordASL/hj reachability, 2021

    2021

  20. [28]

    Deepreach: A deep learning approach to high-dimensional reachability,

    S. Bansal and C. J. Tomlin, “Deepreach: A deep learning approach to high-dimensional reachability,” in2021 IEEE International Confer- ence on Robotics and Automation (ICRA), 2021, pp. 1817–1824

    2021

  21. [29]

    Exact imposition of safety boundary conditions in neural reachable tubes,

    A. Singh, Z. Feng, and S. Bansal, “Exact imposition of safety boundary conditions in neural reachable tubes,” in2025 IEEE International Conference on Robotics and Automation (ICRA), 2025. [Online]. Available: https://arxiv.org/abs/2404.00814

    2025 arXiv

  22. [30]

    A time-dependent hamilton-jacobi formulation of reachable sets for continuous dynamic games,

    I. M. Mitchell, A. M. Bayen, and C. J. Tomlin, “A time-dependent hamilton-jacobi formulation of reachable sets for continuous dynamic games,”IEEE Transactions on automatic control, vol. 50, no. 7, pp. 947–957, 2005

    2005

  23. [31]

    On reachability and minimum cost optimal control,

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

    2004

  24. [32]

    Solving multi- agent safe optimal control with distributed epigraph form MARL,

    S. Zhang, O. So, M. Black, Z. Serlin, and C. Fan, “Solving multi- agent safe optimal control with distributed epigraph form MARL,” in Proceedings of Robotics: Science and Systems, 2025

    2025

  25. [33]

    Verification of neural reachable tubes via scenario optimization and conformal prediction,

    A. Lin and S. Bansal, “Verification of neural reachable tubes via scenario optimization and conformal prediction,” inProceedings of the 6th Annual Learning for Dynamics & Control Conference, ser. Proceedings of Machine Learning Research, A. Abate, M. Cannon, K. Margellos, and ...

    2024

  26. [1986]

    Available: https://doi.org/10.1137/0324032

    [Online]. Available: https://doi.org/10.1137/0324032

    doi

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.