Pith. sign in

REVIEW 4 major objections 7 minor 42 references

Diffusion Policies for Generative Modeling of Spacecraft Trajectories

T0 review · 4 major / 7 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read This paper claims that composing a trajectory diffusion model with risk maps and constraint energies at inference time generates constraint-abiding, multi-modal spacecraft landing trajectories without retraining.

desk verdict A clean engineering demonstration of compositional diffusion for 6-DoF landing-site selection, but the 'few-shot/OOD' claims outrun the experiments and the inference-time composition is only validated on hand-tuned single-configuration demos. read the letter →

arxiv 2501.00915 v1 pith:OVMWKPDE submitted 2025-01-01 cs.RO cs.LGcs.SYeess.SYmath.OC

Julia Briden , Breanna Johnson , Richard Linares , Abhishek Cauligi This is my paper · ORCID
classification cs.ROcs.LGcs.SYeess.SYmath.OC
keywords diffusionmodelstrajectorygenerationpowereddescentguidancecompositionalgenerativemodelingenergy-basedmulti-landingsiteselectionsuccessiveconvexificationspacecraftautonomy
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

Current machine-learning methods for spacecraft trajectory generation learn a single parameter-to-solution mapping, so any change in mission constraints forces a new dataset and a new model. This paper argues that diffusion models avoid that limitation by representing the trajectory distribution as a probability density that can be composed, at inference time, with other densities encoding costs and constraints. The central demonstration is a 6-degree-of-freedom powered descent guidance model trained on successive-convexification solutions: negation composition with a differentiable glideslope energy cuts constraint violations, inpainting enforces waypoint and boundary equality constraints exactly, and composing with a low-risk landing-site map steers samples away from high-risk sites. If correct, a single learned dynamics model can generate the multi-modal, constraint-abiding initial guesses that warm-start trajectory optimization, removing the need to retrain for each new landing site or constraint.

What carries the argument

The central object is the compositional diffusion model, which treats a trained trajectory diffusion model as an energy-based model: the learned energy $E_\theta(\mathbf{x},t)$ is proportional to the negative log-likelihood, and the score function is $\boldsymbol{\epsilon}_\theta(\mathbf{x}_t,t)=\nabla E_\theta(\mathbf{x}_t,t)/\sqrt{1-\bar\alpha_t}$. Composition combines independent models through product $p_\theta(\mathbf{x})\prod_k g_k$, negation $p_\theta(\mathbf{x})/h(\mathbf{x})^\alpha$, or inpainting with a Dirac-delta constraint $h(\tau)=\delta_{g_t(s_0,a_0,\dots)}$. Constraint energies are hand-crafted differentiable functions, e.g., the glideslope energy $E_{\text{constraint}}=-\lambda_{\text{penalty}}\,\text{violation}^2$ and the obstacle energy built from a normalized risk map with bilinear interpolation and a sinusoidal time scaling $s(t)=\sin(\pi t/T)$. These energies are summed with weights ($\alpha_1=1.3$, $\alpha_2=0.3$) to form the composed energy, whose gradient drives reverse diffusion; the reverse process is annealed ULA MCMC without full MCMC iterations, so samples approximate rather than exactly match the composed distribution.

What would settle it

Re-run the glideslope composition experiment and measure the fraction of trajectory segments violating the constraint for the base and composed models; if the composed model does not reproducibly cut the violation rate below the base model's roughly 10%, or if the reduction disappears when the penalty weight $\lambda_{\text{penalty}}$ is perturbed, the composition assumption is not doing the claimed work.

Watch

Extended reading notes

Core claim

The paper's discovery is that a diffusion model trained only on unconstrained 6-DoF powered-descent trajectories, produced by successive convexification (SCvx), already encodes the dynamics well enough that new mission requirements can be imposed at sampling time by composing probability densities. Formally, the composed distribution is $p_\theta(\mathbf{x}) \propto \prod_i p^i_\theta(\mathbf{x}\mid\text{env})\prod_j c_j\prod_k g_k$, where $p^i_\theta$ are independently trained trajectory diffusion models, $c_j$ are cost distributions, and $g_k$ are constraint distributions. Using energy-based formulations, the glideslope constraint is encoded as a differentiable violation energy under negation composition, reducing glideslope violations to below 5% of trajectory segments versus roughly 10% for the base model; inpainting forces waypoint equality constraints exactly while preserving multi-modality; and composition with a risk map reduces landings in high-risk regions by about 40% (from 11% to under 7%). The authors claim this is the first use of generative composition and multi-modality for high-dimensional multi-landing-site trajectory generation, with all samples produced without retraining and used as initial guesses for the optimizer.

Load-bearing premise

The approach assumes that composing independently trained diffusion models by simple energy addition and negation, with only reverse diffusion acting as sampling, yields samples close enough to the true constrained distribution that the new constraints are satisfied.

Editorial extensions

If this is right

  • A single trajectory diffusion model can be reused for new landing-site maps and constraints at inference time, eliminating the need to regenerate datasets and retrain for each mission variation.
  • Negation composition with a differentiable glideslope energy cuts glideslope violations from about 10% of segments to below 5%.
  • Inpainting enforces boundary and waypoint equality constraints exactly while preserving multi-modal trajectory choices, such as passing left or right of a waypoint.
  • Composing with a normalized landing-risk map reduces landings in regions with over 10% risk from 11% to under 7%.
  • Generated samples act as initial guesses for SCvx-based powered descent guidance, yielding dynamically feasible trajectories in a fraction of the time needed to generate optimizer solutions (about 1.17 minutes for 1,000 diffusion samples versus roughly one hour for 1,000 optimizer trajectories).

Reading between the lines

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

  • If the composition approximation holds, the same recipe should transfer to other differentiable constraint energies, such as no-fly zones, plume-impingement limits, or communication line-of-sight constraints, since the framework only requires a differentiable energy layer.
  • The paper leaves open whether a small number of MCMC refinement steps would close the gap between the approximate reverse-diffusion samples and the exact composed distribution; a testable extension is to measure constraint violation and sample diversity as a function of the number of MCMC steps.
  • The hand-tuned weights ($\alpha_1=1.3$, $\alpha_2=0.3$, point weights 1 and 2) suggest the method's robustness should be checked against variations in these constants, since a constraint-satisfaction result that survives weight perturbation would be much stronger than one tuned per scenario.
  • The reported distributional bias of the base diffusion model (e.g., negative-x bias up to 0.7 and positive-y bias less than 1.8) implies that composed constraints may inherit or amplify base-model bias; an extension would be to measure landing-site selection bias against the risk map rather than only aggregate violation rates.
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

4 major / 7 minor

Summary. This paper trains a denoising diffusion probabilistic model on 6-degree-of-freedom powered-descent guidance trajectories generated by a successive convexification solver, with random initial conditions and random final horizontal positions. It then uses energy-based composition, specifically negation composition with hand-crafted glideslope and risk-map energy functions, to modify the generated distribution at inference time, and uses inpainting to enforce equality constraints such as waypoints. The reported results are distributional comparisons between diffusion-generated and optimizer-generated trajectories, a glideslope-violation comparison (4.5% vs. 10%), a qualitative inpainting demonstration with multimodality, and a risk-map comparison (<7% vs. 11% high-risk landings). The paper claims this is the first generative-composition approach to high-dimensional multi-landing-site trajectory generation and that the generated samples are useful as initial guesses for powered-descent guidance.

Significance. If the claims hold, this would be a practically valuable contribution: a single dynamics model could be reused for multiple mission constraints and landing-site selections without retraining, and the demonstrated multimodality is genuinely useful for sampling-based warm starting. The paper is honest about its main limitation, namely that reverse diffusion does not sample the exact composed distribution, and the inpainting portion is exact by construction. The risk-map composition is a clear and nontrivial demonstration. However, the experimental support is currently narrow: all constraint-satisfaction claims come from single configurations with hand-tuned energy weights, no error bars, no sensitivity analysis, no warm-start solver study, and no evaluation outside the training distribution. The significance is therefore conditional on substantially strengthening the empirical evaluation.

major comments (4)
  1. [Section IV.A, Eqs. (18e), (23e), and Table 2] The boundary conditions fix the full final position vector to zero, but the accompanying text states that 'only the z component of the final state is constrained to be zero, while the x and y components are sampled,' and Table 2 lists r_x,N and r_y,N as uniform on [-4,4]. This internal contradiction affects the definition of the multi-landing-site dataset and the diffusion model's training target; please clarify which boundary conditions were actually used to generate the training data.
  2. [Abstract and Section V] The abstract claims that the method efficiently adapts out-of-distribution data and problem variations in a few-shot framework and that the samples enable computationally efficient trajectory generation, but the paper contains no out-of-distribution test (e.g., initial conditions outside the ranges in Table 2, different glideslope angles, or unseen obstacle layouts) and no warm-start study in which diffusion samples initialize SCvx or an NLP solver. Section V.A reports only raw sampling times (1.17 min for 1000 diffusion samples versus about 5.67 min for 100 optimizer solutions). Because the warm-start benefit is a central claimed benefit, this experiment or a tempering of the claim is necessary.
  3. [Section V.B, Eq. (31); Section V.D, Eq. (34); Section III.B] The reported glideslope improvement (4.5% vs. 10%) and high-risk-landing improvement (<7% vs. 11%) are single point estimates from one configuration of hand-set weights (alpha_1=1.3, alpha_2=0.3, lambda_penalty, lambda, tau, w_k, s(t)), with no confidence intervals, no seed variation, and no sensitivity analysis. Because Section III.B explicitly states that reverse diffusion does not sample the exact composed distribution, and Eq. (13) bounds score error rather than constraint-violation probability, the manuscript needs violation statistics over repeated sampling runs and over ranges of these parameters and problem parameters before the 'constraint-abiding' claim is supported.
  4. [Section III.A.1 and Section V.A.1] The derivation of the score loss, the score-error bound in Eq. (13), the cosine beta scheduling, and the initialization of the trained model all come from reference [14], which is authored by the first author of this paper. The manuscript does not state whether [14] is publicly or archivally available, and it does not include the derivation or the initialization hyperparameters. Since these are load-bearing for the trained model and the theoretical justification of composition, please either provide the required derivation and training details in an appendix or confirm that [14] is accessible and version-stable.
minor comments (7)
  1. [Nomenclature] The entry defining alpha_t as the product of beta_t is inconsistent with the sampling equation (7), where alpha_t appears as the product of (1 - beta_t); please correct the definition.
  2. [Figure 2 caption] The caption says 'multi-model structure' but should say 'multi-modal structure.'
  3. [Section VI] The claim of 'reducing constraint violations by 40%' does not follow exactly from 11% to less than 7% (approximately 36%); please report the measured values and the calculation.
  4. [Eqs. (13)-(14)] The score-error bound is a single inequality split across two numbered equations; renumbering as one equation would avoid the impression that there are two separate bounds.
  5. [Section V.A] The phrase 'transported convolutional layers' should be 'transposed convolutional layers.'
  6. [Section V.C] The claim that inpainting generates trajectories that 'exactly meet all equality constraints' would be stronger with a numerical error table (e.g., max absolute constraint error over a sample set), since Figure 17 alone is qualitative.
  7. [Section V.A, Table 3] The U-Net description lists only the conv2_d weight shapes; providing the full architecture details, including skip connections, normalization, and any attention layers, or a pointer to code, would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the trajectory model is benchmarked against SCvx data and the composed models are compared to the un-composed baseline, so the load-bearing claims do not reduce to their inputs.

full rationale

Tracing the paper's derivation chain: the core trajectory model is trained on SCvx-generated solutions of the 6-DoF powered-descent problem (Eqs. 22-27, Table 2), and its fidelity is assessed against SCvx samples (Figures 5-14), an external benchmark independent of any parameter fitted in the paper. The compositional step (Eq. 16) is an explicitly assumed product of independent densities, and the constraint-satisfaction results are measured against the un-composed baseline: negation composition lowers glideslope violations from about 10% to 4.5% (Section V.B) and high-risk landings from 11% to less than 7% (Section V.D). These are demonstrations of an approximate sampler, not predictions forced by construction: reverse diffusion is stochastic, the composed energy only biases the distribution (Eqs. 31-36), and Section III.B explicitly concedes that reverse diffusion does not sample from the exact composed distribution. The hand-set weights and the lack of sensitivity analysis are correctness and generalization risks, not circularity. Self-citations to Ref. [14] appear for the score-error bound (Eq. 13), the negation-over-product recommendation, the cosine beta schedule, and model initialization; these support design choices and background, but the paper re-demonstrates the main qualitative effects in its own experiments, so no load-bearing claim reduces to a self-citation. No step was found where a stated prediction equals an input by definition.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central results rest on four families of assumptions: standard diffusion and DDPM theory, the compositional independence premise inherited from [13], the optimality and feasibility of SCvx training data, and a set of hand-tuned energy weights and thresholds used to demonstrate constraint satisfaction. No new physical entities are introduced. The free parameters are numerous and several numeric values are not reported, which makes the quantitative claims hard to reproduce or interpret as predictions.

free parameters (7)
  • alpha_1 (trajectory energy weight) = 1.3
    Chosen manually for glideslope negation composition; not predicted or justified.
  • alpha_2 (constraint energy weight) = 0.3
    Chosen manually; score magnitude balance affects whether composition changes samples, as the paper itself notes.
  • lambda_penalty (glideslope energy scale) = not stated
    Introduced in Eq. (31); the behavior of Eq. (31) depends on it and no value is reported.
  • risk map penalty coefficient lambda = not stated
    From Eq. (34); controls obstacle gradient strength and is not given a value or sensitivity study.
  • risk threshold tau = not stated
    From Eq. (34); determines what risk level triggers penalty; no value or sensitivity analysis is reported.
  • point weights w_k = 1.0 for start/end, 2.0 for middle
    Described as found empirically in Section V.D; this is a free parameter fit to the evaluation metric.
  • risk_map_gaussian_sigma = 0.5
    Chosen for the risk map in Table 4; affects the smoothness of the obstacle risk distribution.
assumptions (5)
  • standard math DDPM forward process q and reverse process p_theta with Gaussian noise schedule are valid.
    Background equations (5)-(7) and (11)-(12); not re-derived.
  • domain assumption Independent diffusion models can be multiplied or negated as probability densities, with normalization constants ignored.
    Assumed in Eq. (16), citing [13]; no convergence or identifiability analysis for this trajectory application.
  • domain assumption SCvx solutions used for training are representative of the true 6 DoF powered descent trajectory distribution.
    Section IV.A; dataset and solver setup follow [34], with no verification against independent ground truth.
  • domain assumption Negation composition with reverse diffusion improves over product composition for powered descent constraints.
    Section V.B, citing [14]; the paper says it was shown experimentally in [14], not proven here.
  • ad hoc to paper The handcrafted glideslope and obstacle energy functions are sufficiently smooth and differentiable for score composition and gradients.
    Needed for Eq. (12) and Eq. (38); a smooth ReLU approximation is used, but no principled derivation or calibration procedure is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Diffusion Policies for Generative Modeling of Spacecraft Trajectories." pith.science (2026). https://pith.science/paper/OVMWKPDE

@misc{pith2026250100915,
  author       = {Pith},
  title        = {Pith review of: Diffusion Policies for Generative Modeling of Spacecraft Trajectories},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OVMWKPDE}},
  note         = {Machine review of arXiv:2501.00915}
}
read the original abstract

Machine learning has demonstrated remarkable promise for solving the trajectory generation problem and in paving the way for online use of trajectory optimization for resource-constrained spacecraft. However, a key shortcoming in current machine learning-based methods for trajectory generation is that they require large datasets and even small changes to the original trajectory design requirements necessitate retraining new models to learn the parameter-to-solution mapping. In this work, we leverage compositional diffusion modeling to efficiently adapt out-of-distribution data and problem variations in a few-shot framework for 6 degree-of-freedom (DoF) powered descent trajectory generation. Unlike traditional deep learning methods that can only learn the underlying structure of one specific trajectory optimization problem, diffusion models are a powerful generative modeling framework that represents the solution as a probability density function (PDF) and this allows for the composition of PDFs encompassing a variety of trajectory design specifications and constraints. We demonstrate the capability of compositional diffusion models for inference-time 6 DoF minimum-fuel landing site selection and composable constraint representations. Using these samples as initial guesses for 6 DoF powered descent guidance enables dynamically feasible and computationally efficient trajectory generation.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 33 canonical work pages

  1. [14]

    Compositional Diffusion Models for Powered Descent Trajectory Generation with Flexible Constraints,

    Briden, J., Du, Y., Zucchelli, E. M., and Linares, R., “Compositional Diffusion Models for Powered Descent Trajectory Generation with Flexible Constraints,”IEEE Aerospace Conference, 2025

    work page 2025

  2. [1]

    Survey of Numerical Methods for Trajectory Optimization,

    Betts, J. T., “Survey of Numerical Methods for Trajectory Optimization,”AIAA Journal of Guidance, Control, and Dynamics, Vol. 21, No. 2, 1998, pp. 193–207. 25 Fig. 19 Trajectory and energy distributions for the composed trajectory diffusion model

    work page 1998

  3. [2]

    Survey of convex optimization for aerospace applications,

    Liu, X., Lu, P., and Pan, B., “Survey of convex optimization for aerospace applications,”Astrodynamics, Vol. 1, No. 1, 2017, pp. 23 – 40

    work page 2017

  4. [3]

    Evaluation of the Multicore Performance Capabilities of the Next Generation Flight Computers,

    Solé, M., Wolf, J., Rodriguez, I., Jover, A., Trompouki, M. M., Kosmidis, L., and Steenari, D., “Evaluation of the Multicore Performance Capabilities of the Next Generation Flight Computers,”Digital Avionics Systems Conference, 2023

    work page 2023

  5. [4]

    Spacecraft Autonomy Challenges for Next Generation Space 26 Missions,

    Starek, J. A., Açıkmeşe, B., Nesnas, I. A. D., and Pavone, M., “Spacecraft Autonomy Challenges for Next Generation Space 26 Missions,”Advances in Control System Technology for Aerospace Applications, Springer, 2016, Chap. 1

    work page 2016

  6. [5]

    End-to-End Constrained Optimization Learning: A Survey,

    Kotary, J., Fioretto, F., Van Hentenryck, P., and Wilder, B., “End-to-End Constrained Optimization Learning: A Survey,”Int. Joint Conf. on Artificial Intelligence, 2021

    work page 2021

  7. [6]

    Tutorial on Amortized Optimization,

    Amos, B., “Tutorial on Amortized Optimization,”Foundations and Trends in Machine Learning, Vol. 16, No. 5, 2023, p. 732

    work page 2023

  8. [7]

    End-to-EndLearningtoWarm-StartforReal-TimeQuadraticOptimization,

    Sambharya,R.,Hall,G.,Amos,B.,andStellato,B.,“End-to-EndLearningtoWarm-StartforReal-TimeQuadraticOptimization,” Learning for Dynamics & Control, 2023

    work page 2023

Show all 42 references
  1. [8]

    Improving Computational Efficiency for Powered Descent Guidance via Transformer-based Tight Constraint Prediction,

    Briden, J., Gurga, T., Johnson, B. J., Cauligi, A., and Linares, R., “Improving Computational Efficiency for Powered Descent Guidance via Transformer-based Tight Constraint Prediction,”AIAA Scitech Forum, 2024

    2024

  2. [9]

    PRISM: Recurrent neural networks and presolve methods for fast mixed-integer optimal control,

    Cauligi, A., Chakrabarty, A., Di Cairano, S., and Quirynen, R., “PRISM: Recurrent neural networks and presolve methods for fast mixed-integer optimal control,”Learning for Dynamics & Control, 2022

    2022

  3. [10]

    Generative Modeling by Estimating Gradients of the Data Distribution,

    Song, Y., and Ermon, S., “Generative Modeling by Estimating Gradients of the Data Distribution,”Conf. on Neural Information Processing Systems, 2019

    2019

  4. [11]

    Improved Techniques for Training Score-Based Generative Models,

    Song, Y., and Ermon, S., “Improved Techniques for Training Score-Based Generative Models,”Conf. on Neural Information Processing Systems, 2020

    2020

  5. [12]

    Planning with Diffusion for Flexible Behavior Synthesis,

    Janner, M., Du, Y., Tenenbaum, J. B., and Levine, S., “Planning with Diffusion for Flexible Behavior Synthesis,”Int. Conf. on Machine Learning, 2022

    2022

  6. [13]

    Compositional Generative Modeling: A Single Model is Not All You Need,

    Du, Y., and Kaelbling, L., “Compositional Generative Modeling: A Single Model is Not All You Need,”Int. Conf. on Machine Learning, 2024

    2024

  7. [15]

    Large scale model predictive control with neural networks and primal active sets,

    Chen, S. W., Wang, T., Atanasov, N., Kumar, V., and Morari, M., “Large scale model predictive control with neural networks and primal active sets,”Automatica, Vol. 135, 2022, p. 109947

    2022

  8. [16]

    Is Conditional Generative Modeling All You Need for Decision-Making?

    Ajay, A., Du, Y., Gupta, A., Tenenbaum, J. B., Jaakkola, T., and Agrawal, P., “Is Conditional Generative Modeling All You Need for Decision-Making?”Int. Conf. on Learning Representations, 2023

    2023

  9. [17]

    Diffusion Policy: Visuomotor Policy Learning via Action Diffusion,

    Chi, C., Feng, S., Du, Y., Xu, Z., Cousineau, E., Burchfiel, B., and Song, S., “Diffusion Policy: Visuomotor Policy Learning via Action Diffusion,”Robotics: Science and Systems, 2023

    2023

  10. [18]

    Motion Planning Diffusion: Learning and Planning of Robot Motions with Diffusion Models,

    Carvalho, J., Le, A. T., Baierl, M., Koert, D., and Peters, J., “Motion Planning Diffusion: Learning and Planning of Robot Motions with Diffusion Models,”IEEE/RSJ Int. Conf. on Intelligent Robots & Systems, 2023

    2023

  11. [19]

    NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration,

    Sridhar, A., Shah, D., Glossop, C., and Levine, S., “NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration,” Proc. IEEE Conf. on Robotics and Automation, 2024

    2024

  12. [20]

    Amortized Global Search for Efficient Preliminary Trajectory Design with Deep Generative Models,

    Li, A., Sinha, A., and Beeson, R., “Amortized Global Search for Efficient Preliminary Trajectory Design with Deep Generative Models,”ArXiv Preprint, 2023. Available athttps://arxiv.org/abs/2308.03960

    2023 arXiv

  13. [21]

    Denoising Heat-Inspired Diffusion with Insulators for Collision Free Motion Planning,

    Chang, J., Ryu, H., Kim, J., Yoo, S., Seo, J., Prakash, N., Choi, J., and Horowitz, R., “Denoising Heat-Inspired Diffusion with Insulators for Collision Free Motion Planning,”NeurIPS 2023 Workshop on Diffusion Models, 2023. Available at https://arxiv.org/abs/2310.12609

    2023 arXiv

  14. [22]

    Compositional Diffusion-Based Continuous Constraint Solvers,

    Yang, Z., Mao, J., Du, Y., Wu, J., Tenenbaum, J. B., Lozano-Pérez, T., and Kaelbling, L. P., “Compositional Diffusion-Based Continuous Constraint Solvers,”Conf. on Robot Learning, 2023

    2023

  15. [23]

    SafeDiffuser: SafePlanningwithDiffusionProbabilisticModels,

    Xiao,W.,Wang,T.-H.,Gan,C.,andRus,D.,“SafeDiffuser: SafePlanningwithDiffusionProbabilisticModels,” ArXivPreprint,

  16. [24]

    Sampling Constrained Trajectories Using Composable Diffusion Models,

    Power, T., Soltani-Zarrin, R., Iba, S., and Berenson, D., “Sampling Constrained Trajectories Using Composable Diffusion Models,”IEEE/RSJ Int. Conf. on Intelligent Robots & Systems: Workshop on Differentiable Probabilistic Robotics: Emerging Perspectives on Robot Learning, 2023

    2023

  17. [25]

    Trajectory Generation, Control, and Safety with Denoising Diffusion Probabilistic Models,

    Botteghi, N., Califano, F., Poel, M., and Brune, C., “Trajectory Generation, Control, and Safety with Denoising Diffusion Probabilistic Models,”ArXiv Preprint, 2023. Available athttps://arxiv.org/abs/2306.15512. 27

    2023 arXiv

  18. [26]

    Efficient and Guaranteed-Safe Non-Convex Trajectory Optimization with Constrained Diffusion Model,

    Li, A., Ding, Z., Dieng, A. B., and Beeson, R., “Efficient and Guaranteed-Safe Non-Convex Trajectory Optimization with Constrained Diffusion Model,”ArXiv Preprint, 2024. Available athttps://arxiv.org/abs/2403.05571

    2024 arXiv

  19. [27]

    Stochastic Optimal Control Matching,

    Domingo-Enrich, C., Han, J., Amos, B., Bruna, J., and Chen, R. T. Q., “Stochastic Optimal Control Matching,”Conf. on Neural Information Processing Systems, 2024

    2024

  20. [28]

    Implicit Behavioral Cloning,

    Florence, P., Lynch, C., Zeng, A., Ramirez, O. A., Wahid, A., Downs, L., Wong, A., Lee, J., Mordatch, I., and Tompson, J., “Implicit Behavioral Cloning,”Conf. on Robot Learning, 2022

    2022

  21. [29]

    Fast Multi-Robot Motion Planning via Imitation Learning of Mixed-Integer Programs,

    Srinivasan, M., Chakrabarty, A., Quirynen, R., Yoshikawa, N., Mariyama, T., and Di Cairano, S., “Fast Multi-Robot Motion Planning via Imitation Learning of Mixed-Integer Programs,”IFAC-Papers Online, Vol. 54, No. 20, 2021, pp. 598–604

    2021

  22. [30]

    Implicitgenerationandmodelingwithenergybasedmodels,

    Du,Y.,andMordatch,I.,“Implicitgenerationandmodelingwithenergybasedmodels,” Conf.onNeuralInformationProcessing Systems, 2019

    2019

  23. [31]

    Exponential convergence of Langevin distributions and their discrete approximations,

    Roberts, G. O., and Tweedie, R. L., “Exponential convergence of Langevin distributions and their discrete approximations,” Bernoulli, Vol. 2, No. 4, 1996, pp. 341–363

    1996

  24. [32]

    Reduce, Reuse, Recycle: Compositional Generation with Energy-Based Diffusion Models and MCMC,

    Du, Y., Durkan, C., Strudel, R., Tenenbaum, J. B., Dieleman, S., Fergus, R., Sohl-Dickstein, J., Doucet, A., and Grathwohl, W. S., “Reduce, Reuse, Recycle: Compositional Generation with Energy-Based Diffusion Models and MCMC,”Int. Conf. on Machine Learning, 2023

    2023

  25. [33]

    HALO: Hazard-Aware Landing Optimization for Autonomous Systems,

    Hayner, C. R., Buckner, S. C., Broyles, D., Madewell, E., Leung, K., and Açıkmeşe, B., “HALO: Hazard-Aware Landing Optimization for Autonomous Systems,”Proc. IEEE Conf. on Robotics and Automation, 2023

    2023

  26. [34]

    Successive Convexification for 6-DoF Mars Rocket Powered Landing with Free-Final-Time,

    Szmuk, M., and Açıkmeşe, B., “Successive Convexification for 6-DoF Mars Rocket Powered Landing with Free-Final-Time,” AIAA Scitech Forum, 2018

    2018

  27. [35]

    Successive Convexification for Real-Time Six-Degree-of-Freedom Powered Descent Guidance with State-Triggered Constraints,

    Szmuk, M., Reynolds, T. P., and Açıkmeşe, B., “Successive Convexification for Real-Time Six-Degree-of-Freedom Powered Descent Guidance with State-Triggered Constraints,”AIAA Journal of Guidance, Control, and Dynamics, Vol. 43, No. 8, 2020, pp. 1399–1413

    2020

  28. [36]

    CVXPY:A Python-Embedded ModelingLanguage forConvex Optimization,

    Diamond, S., and Boyd, S., “CVXPY:A Python-Embedded ModelingLanguage forConvex Optimization,”Journal of Machine Learning Research, Vol. 17, No. 83, 2016, pp. 1–5

    2016

  29. [37]

    A rewriting system for convex optimization problems,

    Agrawal, A., Verschueren, R., Diamond, S., and Boyd, S., “A rewriting system for convex optimization problems,”Journal of Control and Decision, Vol. 5, No. 1, 2018, pp. 42–60

    2018

  30. [38]

    ECOS: An SOCP solver for embedded systems,

    Domahidi, A., Chu, E., and Boyd, S., “ECOS: An SOCP solver for embedded systems,”European Control Conference, 2013

    2013

  31. [39]

    Model Based Reinforcement Learning for Atari,

    Kaiser, L., Babaeizadeh, M., Miłos, P., Osiński, B., Campbell, R. H., Czechowski, K., Erhan, D., Finn, C., Kozakowski, P., Levine, S., Mohiuddin, A., Sepassi, R., Tucker, G., and Michalewski, H., “Model Based Reinforcement Learning for Atari,”Int. Conf. on Learning Representat...

    2020

  32. [40]

    Improved Denoising Diffusion Probabilistic Models,

    Nichol, A. Q., and Dhariwal, P., “Improved Denoising Diffusion Probabilistic Models,”Int. Conf. on Machine Learning, 2021

    2021

  33. [41]

    Scikit-learn: Machine Learning in Python,

    Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, E., “Scikit-learn: Machine Learning in Python,”Journal of...

    2011

  34. [2023]

    Available athttps://arxiv.org/abs/2306.00148

    arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.