pith. sign in

arxiv: 2607.00776 · v1 · pith:AGQDTML7new · submitted 2026-07-01 · 💻 cs.RO · cs.SY· eess.SY

From Prediction Uncertainty to Conformalized Distance Fields for Safe Motion Planning

Jaeuk Shin , Yoonseok Ra , Insoon Yang This is my paper

Pith reviewed 2026-07-02 11:27 UTC · model grok-4.3

classification 💻 cs.RO cs.SYeess.SY
keywords functional conformal predictiondistance fieldsafe motion planningmodel predictive controldynamic obstaclesprediction uncertaintyfunctional PCAdistribution-free bounds
0
0 comments X

The pith

Functional conformal prediction on the full distance field yields a uniform, distribution-free safety bound for trajectories.

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

The paper shows that conformalizing the entire predicted distance field at once, rather than scalar per-obstacle scores, produces a field-level lower bound from which safety follows uniformly for any satisfying trajectory. This bound is obtained by fitting an offline envelope via functional PCA and Gaussian-mixture inductive conformal prediction, then refining it online with a lightweight adaptive update on low-dimensional coefficients. The approach is tractable because the residual distance field is empirically low-rank and approximately time-invariant. When embedded as a tightened constraint in sampling-based model predictive control, the method scales to scenes with hundreds of dynamic obstacles while remaining insensitive to obstacle count.

Core claim

The functional conformal prediction framework conformalizes the predicted distance field at once to obtain a field-level lower bound. Safety follows uniformly because any trajectory satisfying the constraint is certified safe, independent of how the control space is sampled. The bound is constructed offline via functional PCA and Gaussian-mixture inductive conformal prediction on the low-dimensional coefficients, then updated online by a lightweight adaptive procedure that maintains coverage under distribution shift.

What carries the argument

Functional conformal prediction (FCP) applied directly to the residual distance field, which is made decomposable in coefficient space by its empirically low-rank and approximately time-invariant structure.

If this is right

  • Any trajectory obeying the conformalized field constraint is certified safe regardless of sampling density.
  • Per-step computation stays largely independent of the number of dynamic obstacles.
  • Long-run field coverage is retained under distribution shift via the adaptive online update.
  • The resulting FCP-MPC reaches goals in dense 3D scenes where pointwise conformal baselines become overly conservative or computationally prohibitive.

Where Pith is reading between the lines

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

  • The low-rank decomposition could be reused to conformalize other predicted spatial quantities such as velocity or cost fields.
  • The uniform safety property independent of sampling may allow direct substitution into continuous optimization planners.
  • The adaptive functional update could be extended to handle non-stationary obstacle dynamics beyond the tested pedestrian and quadrotor cases.

Load-bearing premise

The residual distance field must be empirically low-rank and approximately time-invariant so the conformal bound can be decomposed in coefficient space.

What would settle it

A test set in which the empirical coverage of the field-level lower bound falls below the nominal target level would falsify the distribution-free safety claim.

Figures

Figures reproduced from arXiv: 2607.00776 by Insoon Yang, Jaeuk Shin, Yoonseok Ra.

Figure 1
Figure 1. Figure 1: In the offline stage, snapshots of residual distance fields are collected, and a set of basis [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: An example snapshot of the quadrotor simulation considered in this paper. [PITH_FULL_IMAGE:figures/full_fig_p021_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The residual field is approximately time-invariant. Top: raw camera frames of the univ scene at an early and a late time, cropped to the walkable region. Bottom: per-cell means of the residual distance field St+i|t estimated from the temporally disjoint early and late halves of the recording over the same region (color: mean residual in meters, shared diverging scale). These two halves agree closely ( [PI… view at source ↗
Figure 4
Figure 4. Figure 4: Per-cell residual means of univ from the temporally disjoint early and late halves, plotted against each other (one point per spatial cell; the dashed line is the identity). They line up closely (Pearson r = 0.71), quantifying the time-invariance visualized in [PITH_FULL_IMAGE:figures/full_fig_p023_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The residual score field is low-rank. The four leading functional principal components ψ1–ψ4 of the univ residual fields (the fraction of variance each explains is shown in parentheses) are smooth, scene-frame spatial modes rather than high-frequency noise. A handful of them capture most of the field, letting the envelope be learned offline. The cumulative variance across all scenes is reported in [PITH_F… view at source ↗
Figure 6
Figure 6. Figure 6: Cumulative variance explained by the leading functional principal components on each [PITH_FULL_IMAGE:figures/full_fig_p025_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Conformalized distance field on a representative [PITH_FULL_IMAGE:figures/full_fig_p026_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Representative closed-loop trajectories on one scene of each of the five ETH–UCY [PITH_FULL_IMAGE:figures/full_fig_p028_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: True and conformal safety bounds in 3D, zoomed onto a single obstacle. The solid blue [PITH_FULL_IMAGE:figures/full_fig_p029_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Closed-loop trajectory comparison in the dense 3D quadrotor task ( [PITH_FULL_IMAGE:figures/full_fig_p030_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Per-step planning cost versus Nobs in the 3D task: mean control time per planning step (log scale, ms) as Nobs grows from 10 to 280, for every controller under the shared planner. FCP-MPC (ours) reports the soft variant; the hard-filter variant is faster still and likewise flat in Nobs. ECP-MPC is one to two orders of magnitude slower throughout, and its cost grows with Nobs. CC- and ACP-MPC also scale wi… view at source ↗
read the original abstract

Safe motion planning in dynamic environments requires reasoning about the uncertainty in predicted obstacle motion without sacrificing real-time performance. Existing conformal approaches conformalize a scalar score that aggregates per-obstacle prediction errors, losing spatial coherence and scaling poorly with scene density. We instead conformalize the entire predicted distance field at once. This functional conformal prediction (FCP) framework yields a distribution-free, field-level lower bound, from which safety follows uniformly: any trajectory satisfying the resulting constraint is certified safe, independent of how the control space is sampled. The key enabler is that the residual distance field is empirically low-rank and approximately time-invariant, which makes the bound decomposable in coefficient space. An envelope is fitted offline via functional PCA and a Gaussian-mixture inductive conformal procedure, then refined online by a lightweight adaptive functional conformal (AFCP) update on a low-dimensional vector. This keeps the per-step cost largely insensitive to obstacle count and retains long-run field coverage under distribution shift. We embed the envelope as a tightened safety constraint in a sampling-based model predictive controller, FCP-MPC. On the ETH--UCY pedestrian benchmarks and a dense 3D quadrotor task with up to 280 dynamic obstacles, FCP-MPC attains a favorable balance of safety, feasibility, and efficiency, reaching goals where pointwise and egocentric conformal baselines become too conservative or too expensive, while keeping per-step computation far below online uncertainty-reasoning baselines.

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 proposes a functional conformal prediction (FCP) framework to obtain a distribution-free, field-level lower bound on predicted distance fields for safe motion planning. It exploits an empirical low-rank and approximately time-invariant structure of the residual distance field to enable offline functional PCA plus GMM inductive conformal prediction, followed by a lightweight adaptive functional conformal (AFCP) online update; the resulting envelope is embedded as a tightened constraint in sampling-based MPC (FCP-MPC). Experiments on ETH-UCY pedestrian data and a dense 3D quadrotor scenario with up to 280 obstacles report favorable safety-feasibility-efficiency trade-offs relative to pointwise and egocentric conformal baselines.

Significance. If the low-rank/time-invariance approximation error can be controlled and the conformal coverage transfers to the true field, the approach supplies a uniform safety certificate independent of control-space sampling density while keeping per-step cost largely insensitive to obstacle count. This addresses a practical limitation of scalar conformal methods in dense dynamic scenes and demonstrates concrete gains on standard benchmarks.

major comments (2)
  1. [Abstract] Abstract: the distribution-free field-level lower bound and uniform safety claim for arbitrary trajectories rest on the statement that 'the residual distance field is empirically low-rank and approximately time-invariant, which makes the bound decomposable in coefficient space.' No quantification of truncation or time-invariance error is supplied; if this error is non-negligible in regions that affect collision checking, the conformal coverage guarantee applies only to the projected field and the safety certificate is no longer distribution-free.
  2. [Abstract] Abstract: the claim that 'long-run field coverage under distribution shift' is retained by the AFCP update is stated without reference to a theorem or lemma establishing that the adaptive procedure preserves marginal coverage when the low-rank decomposition is only approximate.
minor comments (1)
  1. The abstract asserts that per-step computation remains 'far below online uncertainty-reasoning baselines,' but does not indicate whether this comparison includes the offline PCA/GMM fitting cost amortized over the run.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive major comments. We address each point below and will incorporate revisions to strengthen the error analysis and theoretical support for the coverage claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the distribution-free field-level lower bound and uniform safety claim for arbitrary trajectories rest on the statement that 'the residual distance field is empirically low-rank and approximately time-invariant, which makes the bound decomposable in coefficient space.' No quantification of truncation or time-invariance error is supplied; if this error is non-negligible in regions that affect collision checking, the conformal coverage guarantee applies only to the projected field and the safety certificate is no longer distribution-free.

    Authors: We agree that the current manuscript presents the low-rank and time-invariance properties empirically without quantitative error bounds, which leaves open the possibility that coverage holds only for the projected field. In the revision we will add a dedicated subsection (and corresponding appendix figures) that reports (i) the cumulative explained variance from functional PCA truncation on both the ETH-UCY and quadrotor datasets and (ii) the L2 norm of the time-invariance residual over sliding windows. These metrics will be evaluated specifically in the vicinity of obstacle surfaces to confirm that approximation error remains negligible where collision checking occurs. The safety certificate will be restated to apply to trajectories that satisfy the conformal envelope on the low-rank reconstruction, with the added error analysis bounding the difference to the true field. revision: yes

  2. Referee: [Abstract] Abstract: the claim that 'long-run field coverage under distribution shift' is retained by the AFCP update is stated without reference to a theorem or lemma establishing that the adaptive procedure preserves marginal coverage when the low-rank decomposition is only approximate.

    Authors: The manuscript asserts long-run coverage retention via the AFCP update but does not supply a formal lemma that accounts for inexact low-rank decomposition. We will add a short lemma in the appendix that decomposes the coverage gap into the standard inductive conformal deviation plus an additive term controlled by the approximation error (already quantified in the new error analysis). The lemma will show that marginal coverage is retained asymptotically whenever the approximation error is bounded, thereby justifying the distribution-shift claim under the empirically observed conditions. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The derivation applies standard conformal prediction theory to coefficients obtained from an empirical functional PCA decomposition. The low-rank and time-invariance properties are presented explicitly as empirical observations that enable the decomposition, not as results derived from the safety bound itself. No equation reduces the field-level lower bound to a fitted parameter or self-citation by construction, and the coverage guarantee remains distribution-free under the stated procedure. The chain is therefore self-contained against external conformal prediction results.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

Abstract-only; the central claim rests on the empirical low-rank and time-invariance property of residual distance fields plus standard conformal prediction coverage guarantees. No explicit free parameters or invented entities are named.

free parameters (2)
  • number of functional PCA components
    Chosen to capture the low-rank structure of the residual distance field; value not stated in abstract.
  • GMM mixture components and conformal quantile
    Parameters of the inductive conformal procedure fitted offline; not quantified in abstract.
axioms (2)
  • domain assumption Residual distance field is empirically low-rank and approximately time-invariant
    Explicitly identified in the abstract as the key enabler that allows decomposition in coefficient space.
  • standard math Standard conformal prediction coverage guarantees apply to the functional setting after PCA projection
    Implicit in the use of inductive conformal prediction on the coefficient vector.

pith-pipeline@v0.9.1-grok · 5796 in / 1399 out tokens · 23593 ms · 2026-07-02T11:27:23.390376+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

45 extracted references · 45 canonical work pages

  1. [1]

    Chance-constrained optimal path planning with obstacles,

    L. Blackmore, M. Ono, and B. C. Williams, “Chance-constrained optimal path planning with obstacles,”IEEE Transactions on Robotics, vol. 27, no. 6, pp. 1080–1094, 2011

    work page 2011

  2. [2]

    Risk-aware motion planning and control using cvar- constrained optimization,

    A. Hakobyan, G. C. Kim, and I. Yang, “Risk-aware motion planning and control using cvar- constrained optimization,”IEEE Robotics and Automation letters, vol. 4, no. 4, pp. 3924–3931, 2019

    work page 2019

  3. [3]

    STEP: Stochastic traversability evaluation and planning for risk-aware off-road navigation,

    D. D. Fan, K. Otsu, Y. Kubo, A. Dixit, J. Burdick, and A.-A. Agha-Mohammadi, “STEP: Stochastic traversability evaluation and planning for risk-aware off-road navigation,”Robotics: Science and Systems XVII, 2021

    work page 2021

  4. [4]

    Wasserstein distributionally robust motion control for collision avoidance using conditional value-at-risk,

    A. Hakobyan and I. Yang, “Wasserstein distributionally robust motion control for collision avoidance using conditional value-at-risk,”IEEE Transactions on Robotics, vol. 38, no. 2, pp. 939–957, 2021

    work page 2021

  5. [5]

    V. Vovk, A. Gammerman, and G. Shafer,Algorithmic Learning in a Random World. Springer, 2005

    work page 2005

  6. [6]

    Conformal prediction: A gentle introduction,

    A. N. Angelopoulos and S. Bates, “Conformal prediction: A gentle introduction,”Foundations and Trends in Machine Learning, vol. 16, no. 4, pp. 494–591, 2023

    work page 2023

  7. [7]

    Adaptive conformal inference under distribution shift,

    I. Gibbs and E. Cand` es, “Adaptive conformal inference under distribution shift,” inAdvances in Neural Information Processing Systems, vol. 34, 2021, pp. 1660–1672

    work page 2021

  8. [8]

    Safe planning in dynamic envi- ronments using conformal prediction,

    L. Lindemann, M. Cleaveland, G. Shim, and G. J. Pappas, “Safe planning in dynamic envi- ronments using conformal prediction,”IEEE Robotics and Automation Letters, vol. 8, no. 8, pp. 5116–5123, 2023

    work page 2023

  9. [9]

    Adaptive conformal prediction for motion planning among dynamic agents,

    A. Dixit, L. Lindemann, S. X. Wei, M. Cleaveland, G. J. Pappas, and J. W. Burdick, “Adaptive conformal prediction for motion planning among dynamic agents,” inProceedings of The 5th Annual Learning for Dynamics and Control Conference, vol. 211, 2023, pp. 300–314

    work page 2023

  10. [10]

    Perceive with confidence: Statistical safety assurances for navigation with learning-based perception,

    Z. Mei, A. Dixit, M. Booker, E. Zhou, M. Storey-Matsutani, A. Z. Ren, O. Shorinwa, and A. Majumdar, “Perceive with confidence: Statistical safety assurances for navigation with learning-based perception,”The International Journal of Robotics Research, vol. 45, no. 6, pp. 938–967, 2025

    work page 2025

  11. [11]

    Conformal deci- sion theory: Safe autonomous decisions from imperfect predictions,

    J. Lekeufack, A. N. Angelopoulos, A. Bajcsy, M. I. Jordan, and J. Malik, “Conformal deci- sion theory: Safe autonomous decisions from imperfect predictions,” inIEEE International Conference on Robotics and Automation, 2024, pp. 11 668–11 675

    work page 2024

  12. [12]

    Safe, out-of-distribution-adaptive MPC with conformalized neural network ensembles,

    J. L. Contreras, O. Shorinwa, and M. Schwager, “Safe, out-of-distribution-adaptive MPC with conformalized neural network ensembles,” inProceedings of The 7th Annual Learning for Dynamics and Control Conference, 2025

    work page 2025

  13. [13]

    Safe planning in unknown environments using conformalized semantic maps,

    D. S. Sundarsingh, Y. Li, T. Tang, G. J. Pappas, N. Atanasov, and Y. Kantaros, “Safe planning in unknown environments using conformalized semantic maps,”IEEE Robotics and Automation Letters, vol. 11, no. 4, pp. 4921–4928, 2026

    work page 2026

  14. [14]

    Simultaneous localiza- tion, mapping and moving object tracking,

    C.-C. Wang, C. Thorpe, S. Thrun, M. Hebert, and H. Durrant-Whyte, “Simultaneous localiza- tion, mapping and moving object tracking,”The International Journal of Robotics Research, vol. 26, no. 9, pp. 889–916, 2007

    work page 2007

  15. [15]

    Bayesian occupancy filtering for multitarget tracking: An automotive application,

    C. Cou´ e, C. Pradalier, C. Laugier, T. Fraichard, and P. Bessi` ere, “Bayesian occupancy filtering for multitarget tracking: An automotive application,”The International Journal of Robotics Research, vol. 25, no. 1, pp. 19–30, 2006

    work page 2006

  16. [16]

    Modeling and tracking the driving environment with a particle-based occupancy grid,

    R. Danescu, F. Oniga, and S. Nedevschi, “Modeling and tracking the driving environment with a particle-based occupancy grid,”IEEE Transactions on Intelligent Transportation Systems, vol. 12, no. 4, pp. 1331–1342, 2011

    work page 2011

  17. [17]

    Grid-based mapping and tracking in dynamic environments using a uniform evidential environment representation,

    G. Tanzmeister, J. Thomas, D. Wollherr, and M. Buss, “Grid-based mapping and tracking in dynamic environments using a uniform evidential environment representation,” inIEEE International Conference on Robotics and Automation, 2014, pp. 6090–6095

    work page 2014

  18. [18]

    A random finite set approach for dynamic occupancy grid maps with real-time application,

    D. Nuss, S. Reuter, M. Thom, T. Yuan, G. Krehl, M. Maile, A. Gern, and K. Dietmayer, “A random finite set approach for dynamic occupancy grid maps with real-time application,”The International Journal of Robotics Research, vol. 37, no. 8, pp. 841–866, 2018

    work page 2018

  19. [19]

    Continuous occupancy mapping in dynamic environments using particles,

    G. Chen, W. Dong, P. Peng, J. Alonso-Mora, and X. Zhu, “Continuous occupancy mapping in dynamic environments using particles,”IEEE Transactions on Robotics, vol. 40, pp. 64–84, 2023

    work page 2023

  20. [20]

    DeepSDF: Learning con- tinuous signed distance functions for shape representation,

    J. J. Park, P. Florence, J. Straub, R. Newcombe, and S. Lovegrove, “DeepSDF: Learning con- tinuous signed distance functions for shape representation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 165–174

    work page 2019

  21. [21]

    iSDF: Real-time neural signed distance fields for robot perception,

    J. Ortiz, A. Clegg, J. Dong, E. Sucar, D. Novotny, M. Zollhoefer, and M. Mukadam, “iSDF: Real-time neural signed distance fields for robot perception,” inProceedings of Robotics: Sci- ence and Systems, New York City, NY, USA, Jun. 2022

    work page 2022

  22. [22]

    KinectFusion: Real-time dense surface mapping and tracking,

    R. A. Newcombe, S. Izadi, O. Hilliges, D. Molyneaux, D. Kim, A. J. Davison, P. Kohli, J. Shotton, S. Hodges, and A. Fitzgibbon, “KinectFusion: Real-time dense surface mapping and tracking,” inIEEE International Symposium on Mixed and Augmented Reality, 2011, pp. 127–136

    work page 2011

  23. [23]

    Trajectron++: Dynamically- feasible trajectory forecasting with heterogeneous data,

    T. Salzmann, B. Ivanovic, P. Chakravarty, and M. Pavone, “Trajectron++: Dynamically- feasible trajectory forecasting with heterogeneous data,” inEuropean Conference on Computer Vision, 2020, pp. 683–700

    work page 2020

  24. [24]

    Probabilistic collision checking with chance constraints,

    N. E. Du Toit and J. W. Burdick, “Probabilistic collision checking with chance constraints,” IEEE Transactions on Robotics, vol. 27, no. 4, pp. 809–815, 2011

    work page 2011

  25. [25]

    Robot motion planning in dynamic, uncertain environments,

    ——, “Robot motion planning in dynamic, uncertain environments,”IEEE Transactions on Robotics, vol. 28, no. 1, pp. 101–115, 2011

    work page 2011

  26. [26]

    Optimization-based collision avoidance,

    X. Zhang, A. Liniger, and F. Borrelli, “Optimization-based collision avoidance,”IEEE Trans- actions on Control Systems Technology, vol. 29, no. 3, pp. 972–983, 2020

    work page 2020

  27. [27]

    Computing large convex regions of obstacle-free space through semidefinite programming,

    R. Deits and R. Tedrake, “Computing large convex regions of obstacle-free space through semidefinite programming,” inAlgorithmic Foundations of Robotics XI: Selected Contributions of the Eleventh International Workshop on the Algorithmic Foundations of Robotics, 2015, pp. 109–124

    work page 2015

  28. [28]

    Collision-free encoding for chance-constrained nonconvex path planning,

    M. da Silva Arantes, C. F. M. Toledo, B. C. Williams, and M. Ono, “Collision-free encoding for chance-constrained nonconvex path planning,”IEEE Transactions on Robotics, vol. 35, no. 2, pp. 433–448, 2019

    work page 2019

  29. [29]

    Distributionally robust optimization with unscented transform for learning-based motion control in dynamic environments,

    A. Hakobyan and I. Yang, “Distributionally robust optimization with unscented transform for learning-based motion control in dynamic environments,” inIEEE International Conference on Robotics and Automation, 2023, pp. 3225–3232

    work page 2023

  30. [30]

    Distributionally robust risk map for learning-based motion planning and control: A semidefinite programming approach,

    ——, “Distributionally robust risk map for learning-based motion planning and control: A semidefinite programming approach,”IEEE Transactions on Robotics, vol. 39, no. 1, pp. 718– 737, 2022

    work page 2022

  31. [31]

    A conformal prediction approach to explore functional data,

    J. Lei, A. Rinaldo, and L. Wasserman, “A conformal prediction approach to explore functional data,”Annals of Mathematics and Artificial Intelligence, vol. 74, no. 1–2, pp. 29–43, 2015

    work page 2015

  32. [32]

    Egocentric conformal prediction for safe and efficient navigation in dynamic cluttered environments,

    J. Shin, J. Lee, and I. Yang, “Egocentric conformal prediction for safe and efficient navigation in dynamic cluttered environments,” inIEEE Conference on Decision and Control, 2025, pp. 4910–4917

    work page 2025

  33. [33]

    Frontier-based exploration using multiple robots,

    B. Yamauchi, “Frontier-based exploration using multiple robots,” inProceedings of the Second International Conference on Autonomous Agents, 1998, pp. 47–53

    work page 1998

  34. [34]

    AgentFormer: Agent-aware transformers for socio-temporal multi-agent forecasting,

    Y. Yuan, X. Weng, Y. Ou, and K. M. Kitani, “AgentFormer: Agent-aware transformers for socio-temporal multi-agent forecasting,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 9813–9823

    work page 2021

  35. [35]

    HiVT: Hierarchical vector transformer for multi- agent motion prediction,

    Z. Zhou, L. Ye, J. Wang, K. Wu, and K. Lu, “HiVT: Hierarchical vector transformer for multi- agent motion prediction,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 8823–8833

    work page 2022

  36. [36]

    MoFlow: One-step flow matching for human trajectory forecasting via implicit maximum likelihood estimation based distillation,

    Y. Fu, Q. Yan, L. Wang, K. Li, and R. Liao, “MoFlow: One-step flow matching for human trajectory forecasting via implicit maximum likelihood estimation based distillation,” inPro- ceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 17 282–17 293

    work page 2025

  37. [37]

    KoopCast: Trajectory forecasting via Koopman operators,

    J. Lee, J. Shin, G. Kim, J. Han, and I. Yang, “KoopCast: Trajectory forecasting via Koopman operators,” inIEEE International Conference on Robotics and Automation, 2026

    work page 2026

  38. [38]

    Crowds by example,

    A. Lerner, Y. Chrysanthou, and D. Lischinski, “Crowds by example,” inComputer Graphics Forum, vol. 26, no. 3. Wiley Online Library, 2007, pp. 655–664

    work page 2007

  39. [39]

    You’ll never walk alone: Modeling social behavior for multi-target tracking,

    S. Pellegrini, A. Ess, K. Schindler, and L. Van Gool, “You’ll never walk alone: Modeling social behavior for multi-target tracking,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2009, pp. 261–268

    work page 2009

  40. [40]

    Information theoretic MPC for model-based reinforcement learning,

    G. Williams, N. Wagener, B. Goldfain, P. Drews, J. M. Rehg, B. Boots, and E. A. Theodorou, “Information theoretic MPC for model-based reinforcement learning,” inInternational Con- ference on Robotics and Automation (ICRA), 2017, pp. 1714–1721

    work page 2017

  41. [41]

    Inevitable collision states — a step towards safer robots?

    T. Fraichard and H. Asama, “Inevitable collision states — a step towards safer robots?” Advanced Robotics, vol. 18, no. 10, pp. 1001–1024, 2004

    work page 2004

  42. [42]

    Feasibility issues in linear model predictive control,

    P. O. M. Scokaert and J. B. Rawlings, “Feasibility issues in linear model predictive control,” AIChE Journal, vol. 45, no. 8, pp. 1649–1659, 1999

    work page 1999

  43. [43]

    Soft constraints and exact penalty functions in model predictive control,

    E. C. Kerrigan and J. M. Maciejowski, “Soft constraints and exact penalty functions in model predictive control,” inProceedings of the UKACC International Conference on Control, 2000

    work page 2000

  44. [44]

    Nocedal and S

    J. Nocedal and S. J. Wright,Numerical Optimization, 2nd ed., ser. Springer Series in Opera- tions Research and Financial Engineering. Springer, 2006

    work page 2006

  45. [45]

    Learning to fly—a gym environment with PyBullet physics for reinforcement learning of multi-agent quadcopter control,

    J. Panerati, H. Zheng, S. Zhou, J. Xu, A. Prorok, and A. P. Schoellig, “Learning to fly—a gym environment with PyBullet physics for reinforcement learning of multi-agent quadcopter control,” inIEEE/RSJ International Conference on Intelligent Robots and Systems, 2021, pp. 7512–7519

    work page 2021