CoChat
Sign Up
Cited by 97Human Motion and AnimationEvacuation and Crowd Dynamics3D Shape Modeling and Analysis

The DOI matches its Crossref record; 8 of the 8 references it cites were found in a registry or an index.

Trace and Pace: Controllable Pedestrian Animation via Guided Trajectory Diffusion

Open access
Rempe, Davis, Luo, Zhengyi, Peng, Xue Bin, Yuan, Ye, Kitani, Kris, Kreis, Karsten, Fidler, Sanja, Litany, Or·2023·2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
DOI 10.1109/cvpr52729.2023.01322
Open this paper in CoChat — free accountOpen PDF

Abstract

We introduce a method for generating realistic pedestrian trajectories and full-body animations that can be controlled to meet user-defined goals. We draw on recent advances in guided diffusion modeling to achieve test-time controllability of trajectories, which is normally only associated with rule-based systems. Our guided diffusion model allows users to constrain trajectories through target waypoints, speed, and specified social groups while accounting for the surrounding environment context. This trajectory diffusion model is integrated with a novel physics-based humanoid controller to form a closed-loop, full-body pedestrian animation system capable of placing large crowds in a simulated environment with varying terrains. We further propose utilizing the value function learned during RL training of the animation controller to guide diffusion to produce trajectories better suited for particular scenarios such as collision avoidance and traversing uneven terrain. Video results are available on the project page.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/cvpr52729.2023.01322

Cites

8 of 8 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Retrieval-augmented generation approaches typically employ dense vector similarity but lose fine-grained temporal information and provide limited interpretability.

    “RAG approaches (Liu et al. 2024; Rempe et al. 2023) typically employ dense vector similarity but lose fine-grained temporal information and provide limited interpretability.”

    From ARTEM: Enhancing Large Language Model Agents with Spatial-Temporal Episodic Memory (2026) · cites this paper as Rempe et al. 2023

    The architecture and ablation passages (A.1.2 and D.2; page numbers unavailable) discuss pedestrian trajectories, feature vectors, and diffusion guidance, not retrieval-augmented generation. Searches for RAG, dense-vector similarity, temporal-information loss, and interpretability found no passage addressing the attributed claim. (LLM, conf=0.76)

    Experimental — accuracy not yet measured

  • Recent approaches to memory tasks in large language models, such as in-context learning, retrieval-augmented generation, and fine-tuning, are inadequate for tasks requiring chronological awareness of stored information.

    “Recent approaches to handle memory tasks in LLMs, such as in-context learning, retrieval-augmented generation (RAG), and fine-tuning, may resolve the long-term retention issues, but are still inadequate to handle tasks requiring chronological awareness of the stored information.”

    From ARTEM: Enhancing Large Language Model Agents with Spatial-Temporal Episodic Memory (2026) · cites this paper as Rempe et al. 2023

    The Related Work section discusses pedestrian planning and animation, while the VAE Baseline Details section concerns trajectory generation; page numbers are unavailable in our copy. Searches for in-context learning, retrieval-augmented generation, LLM memory, and chronological awareness found no passage addressing the claim. (LLM, conf=0.78)

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Flagged first, then by citations

Research works printed but not linked: 56

  1. 10.1038/35035023 · DOI
  2. 10.1145/3528223.3530067 · DOI
  3. 10.1145/3424636.3426894 · DOI
  4. ho, 2022, Imagen video: High definition video generation with diffusion mod-els, ArXiv Preprint
  5. yu, 2018, Learning sym-metric and low-energy locomotion, ACM Transactions on Graphics (TOG), 37, 1
  6. 10.1103/PhysRevE.51.4282 · DOI
  7. xu, 2022, Bits: Bi-level imitation for traffic simulation, ArXiv Preprint
  8. 10.1103/PhysRevE.62.1805 · DOI · Paper page
  9. song, 2019, Generative modeling by esti-mating gradients of the data distribution, Advances in neural information processing systems, 32
  10. hassan, 0, Stochas-tic scene-aware motion prediction, Proceedings of the IEEE/CVF International Conference on Computer Vision, 11374
  11. van den berg, 0, Recip-rocal velocity obstacles for real-time multi-agent navigation, 2008 IEEE international conference on robotics and au-tomation, 1928
  12. 10.1109/CVPR.2018.00240 · DOI · Paper page
  13. vahdat, 2021, Score-based generative modeling in latent space, Advances in neural information processing systems, 34, 11287
  14. ho, 2020, Denoising diffusion probabilistic models, Advances in neural information processing systems, 33, 6840
  15. 10.1145/3072959.3073663 · DOI
  16. ho, 2022, Video dif-fusion models, ArXiv Preprint
  17. ren, 2017, Group modeling: A uni-fied velocity-based approach, Computer Graphics Forum, 36, 45, 10.1111/cgf.12993 · DOI
  18. 10.1109/CVPR52688.2022.01679 · DOI
  19. rudin, 2021, Learning to walk in minutes using massively parallel deep reinforcement learning
  20. 10.1177/0278364920917446 · DOI
  21. 10.1007/978-3-030-69532-3_29 · DOI
  22. 10.1145/3450626.3459670 · DOI
  23. 10.1109/ICCV48922.2021.01129 · DOI
  24. priisalu, 2022, Generating scenarios with diverse pedes-trian behaviors for autonomous vehicle testing, Confer-ence on Robot Learning, 1247
  25. salzmann, 2020, Trajectron++: Dynamically-feasible trajec-tory forecasting with heterogeneous data, European Con-ference on Computer Vision, 683
  26. fu, 2022, Deep whole-body control: Learning a unified policy for manipulation and locomotion, ArXiv
  27. 10.1109/CVPR52688.2022.01660 · DOI
  28. cao, 2020, Long-term human motion prediction with scene context, European Conference on Computer Vision, 387
  29. 10.1007/978-3-030-69532-3_25 · DOI
  30. chai, 0, Multipath: Multiple probabilistic anchor trajec-tory hypotheses for behavior prediction, Conference on Robot Learning (CoRL), 86
  31. 10.1145/3528223.3530110 · DOI
  32. meng, 2022, On distillation of guided diffusion models, ArXiv Preprint
  33. mangalam, 0, From goals, waypoints & paths to long term hu-man trajectory forecasting, Proceedings of the IEEE/CVF International Conference on Computer Vision, 15233
  34. 10.1145/3528233.3530712 · DOI
  35. nichol, 0, Improved denoising diffusion probabilistic models, International Conference on Machine Learning, 8162
  36. luo, 2022, Em-bodied scene-aware human pose estimation, Advances in neural information processing systems
  37. luo, 2021, Dynamics-regulated kinematic policy for egocentric pose es-timation, Advances in neural information processing systems
  38. makoviychuk, 2021, Isaac gym High performance gpu-based physics simulation for robot learning
  39. 10.1109/ICCV.2019.00554 · DOI · Paper page
  40. 10.1007/978-3-642-19457-3_1 · DOI
  41. alahi, 0, So-ciallstm: Human trajectory prediction in crowded spaces, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 961
  42. 10.1145/3072959.3073602 · DOI
  43. 10.1109/CVPR.2017.233 · DOI
  44. 10.1145/3274247.3274510 · DOI
  45. 10.1111/j.1467-8659.2007.01089.x · DOI
  46. lemonari, 2022, Authoring virtual crowds: A sur-vey, Computer Graphics Forum, 41, 677, 10.1111/cgf.14506 · DOI
  47. janner, 0, Planning with diffusion for flexible behavior synthe-sis, International Conference on Machine Learning (ICML)
  48. 10.1109/ICRA48891.2023.10161463 · DOI
  49. 10.1109/CVPR52688.2022.01983 · DOI
  50. 10.1145/2601097.2601170 · DOI
  51. 10.1145/3272127.3275079 · DOI
  52. 10.1145/3386569.3392422 · DOI
  53. 10.1109/ITSC.2018.8569938 · DOI
  54. 10.1109/CVPR46437.2021.00708 · DOI
  55. zeng, 2022, Lion: Latent point diffusion models for 3d shape generation, Advances in Neural IInformation Processing Systems
  56. 10.1109/ICCV48922.2021.00967 · DOI

Cited by

Sorted by citations

Metadata

Authors
Rempe, D., Luo, Z., Peng, X. B., Yuan, Y., Kitani, K., Kreis, K., Fidler, S., & Litany, O.
Affiliations
  • Nvidia (United Kingdom) (GB)
  • Carnegie Mellon University (US)
ORCID
Venue
2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Publisher
IEEE
Published
2023-06
Pages
13756-13766
ISSN
1063-6919 (ISSN-L)
DOI
10.1109/cvpr52729.2023.01322
OpenAlex
W4386066131
Cited by
97
References
8
Citations by year
2026: 13 · 2025: 46 · 2024: 37 · 2023: 1
FWCI
41.53
Type
conference paper
Language
en
Open access
Licence
Locations
  • IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University)
Funders
NVIDIA
Topics
  • Physical Sciences › Computer Science › Computer Graphics and Computer-Aided Design › Human Motion and Animation
  • Social Sciences › Safety Research › Evacuation and Crowd Dynamics
  • Physical Sciences › Computer Science › Computer Graphics and Computer-Aided Design › 3D Shape Modeling and Analysis
Keywords
guided diffusion, trajectory controllability, crowd simulation, uneven terrain traversal, collision avoidance

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review