Pith. sign in

REVIEW 4 major objections 4 minor 46 references

A Unified Framework for Trajectory Prediction with Explicit Planning and Reaction Decomposition

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

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

Pith's one-line read INTraJ claims that trajectory prediction improves when social influence is split into a planning stage, which builds a smooth reference path from future social cues, and a reaction stage, which adds gated local corrections.

desk verdict A useful plug-in refinement method whose planning-reaction story is not actually isolated by the experiments; the gain comes from a learned gate around a smoothed-future reference. read the letter →

arxiv 2608.05673 v1 pith:PWSH6XMC submitted 2026-08-06 cs.AI cs.CVcs.LG

Jiaheng Chen , Jiaxing Li , Tinghe Zhang , Chaopeng Guo This is my paper · ORCID
classification cs.AIcs.CVcs.LG
keywords trajectorypredictionmotionforecastingsocialinteractionmodelingplanningandreactiondecompositionmulti-targetsingle-targetgatedresidualfusionpedestrian
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 tries to establish that social influence in trajectory prediction should be modeled as a two-stage process rather than a single fused latent factor. It claims that an agent first forms a socially shaped overall plan by anticipating other agents' future motions, and then makes local reactive adjustments to that plan. INTraJ implements this idea by decoding a smoothed planning reference from a pure intent seed augmented with future social information, keeping the backbone's full-context prediction, and combining the two through a gated residual. The paper reports consistent improvements in final displacement error and long-horizon consistency across four benchmarks, including autonomous driving and pedestrian crowds, suggesting that the planning-before-reaction decomposition is a transferable modeling principle.

What carries the argument

The Plan-React module carries the argument. It defines the identity $T_\mathrm{final} = T_\mathrm{ideal} + \alpha \cdot S_\mathrm{high}$, where $T_\mathrm{ideal}$ is a planning reference decoded from a socially shaped intent representation, $T_\mathrm{full}$ is the backbone's full-context prediction, and $S_\mathrm{high} = T_\mathrm{full} - T_\mathrm{ideal}$ is the reaction residual. The planning reference is produced by cross-attending a pure intent seed (derived from the focal agent's history and local context) to a future social carrier $z$ — predicted trajectories of other agents in the multi-target paradigm, or a counterfactual latent response from a static-ego branch in the single-target paradigm. A learnable gate $\alpha$ controls how much of the residual is retained per sample, mode, and time step, and a variance-based auxiliary loss prevents the gate from collapsing to a constant.

What would settle it

Run INTraJ with several smoothing strengths for $T_\mathrm{ideal}$; if the FDE improvement is insensitive to smoothing strength or appears even when $T_\mathrm{ideal}$ is replaced by an unfiltered copy of the ground truth, the two-stage interpretation loses support. Equally, train a matched model with identical parameters but no explicit planning reference, predicting $T_\mathrm{final}$ directly from the same socialized intent, and compare FDE: if the gains vanish, the specific decomposition is not the mechanism.

Watch

Extended reading notes

Core claim

The central claim is that trajectory formation follows a planning-before-reaction order: before acting, an agent anticipates the future behavior of surrounding agents and forms a socially shaped plan, and only then makes local reactive adjustments when executing that plan. Concretely, INTraJ computes a planning-level reference trajectory $T_\mathrm{ideal}$ from a pure intent seed updated through cross-attention with a future social carrier (predicted futures of others in multi-target settings, a counterfactual latent response in single-target settings). It keeps the backbone's full-context prediction $T_\mathrm{full}$ and treats the difference $S_\mathrm{high} = T_\mathrm{full} - T_\mathrm{ideal}$ as reaction-stage adjustment, fused by $T_\mathrm{final} = T_\mathrm{ideal} + \alpha S_\mathrm{high}$ with a learned gate $\alpha$. The paper argues that this explicit separation of global passing tendency from local interaction response yields more stable, coherent predictions, and evidences it with gains in FDE across AV2, AV2-ped, ETH/UCY, and SDD.

Load-bearing premise

The planning reference is trained against a mildly smoothed ground-truth trajectory; if that smoothed target is not what an agent would actually plan before reacting, the planning/reaction split may be an artifact of low-pass filtering rather than a real behavioral mechanism.

Editorial extensions

If this is right

  • If the two-stage view is right, then the same decomposition should keep improving any future backbone it is plugged into, since it adds a structured planning head without altering the base predictor.
  • The framework's gains in FDE and long-horizon consistency imply that explicitly separating global passing tendency from local reaction is a useful inductive bias for trajectory stability.
  • Because INTraJ works in both multi-target and single-target paradigms, the planning-before-reaction principle is not tied to a particular architecture or input representation.
  • The consistent improvements on both autonomous driving and pedestrian crowd benchmarks suggest that staged social modeling generalizes across application domains.

Reading between the lines

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

  • An untested consequence of the paper's view is that the reaction residual $S_\mathrm{high}$ should spike at concrete local interaction events, such as yielding or avoiding a collision; one could verify this by checking whether high-residual moments align with time-to-collision minima in the data.
  • Because the planning reference is supervised by a smoothed ground truth, the paper has not ruled out the alternative that the planning stage is merely a low-pass filter; a testable extension would replace the smoothed target with a scene-derived planning signal, such as lane or goal conditioning, and compare.
  • If the decomposition is genuinely causal, perturbing the future social carrier should alter $T_\mathrm{ideal}$ more than $S_\mathrm{high}$; this could be tested by selectively ablating the carrier and measuring each component's change.
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 / 4 minor

Summary. The paper proposes INTraJ, a plug-in framework that decomposes social influence in trajectory prediction into a planning stage and a reaction stage. The planning stage constructs a reference trajectory T_ideal from a pure-intent seed plus a future social carrier, and the reaction stage forms the final prediction as T_final = T_ideal + α·(T_full − T_ideal), where T_full is the backbone prediction and α is a learned gate. The method is instantiated on multi-target (QCNet) and single-target (DeMo) backbones for Argoverse 2, on Resonance for ETH/UCY and SDD, and on a lightweight baseline for AV2-ped. The paper reports consistent improvements over the backbones, with state-of-the-art results on some crowd benchmarks, and provides ablation and efficiency analyses. The manuscript also contains an embedded "Author Response" section that qualifies the interpretation of T_ideal and T_full − T_ideal.

Significance. The two-stage planning/reaction formulation is potentially useful as a structured modeling principle, and the plug-in nature across two prediction paradigms and four benchmarks is an attractive feature. The paper provides public code and a broad empirical evaluation. However, the experiments as reported do not isolate the claimed mechanism from a simpler explanation: the final prediction is a gate-weighted blend of the backbone output and a network trained to regress a smoothed version of the ground-truth future. Because the central claim that "staged social modeling is critical" rests on this isolation, the significance is conditional on additional controls and claim revisions.

major comments (4)
  1. [§4.5, Table 5] The ablation does not support the claim that both the planning stage and the reaction stage are necessary. Compared with the QCNet baseline (minADE1 1.69, minFDE1 4.30), the ideal-only branch degrades minADE1 to 1.75, the ideal-plus-fusion branch without the learned gate matches the baseline (1.69/4.30), and only the full gated model improves both metrics (1.63/4.14). Thus, the entire observed gain on this benchmark comes from the learned gate, and the planning reference alone is harmful. The text's assertion that removing either stage causes a performance drop is contradicted by the table's numbers.
  2. [§3.5, Eq. (12) and Eq. (13)] Because T_ideal is trained to regress a smoothed ground-truth trajectory T_smooth and T_final = α·T_full + (1−α)·T_ideal, the reported gains could in principle be obtained by any low-pass-filtered reference blended with the backbone, without a planning-before-reaction mechanism. The paper does not provide a control experiment with a non-social smoothing reference, nor a sensitivity analysis over the smoothing strength used to define T_smooth. Such a control is necessary to attribute the improvement to the proposed social planning decomposition rather than to low-pass regularization.
  3. [Embedded Author Response] The manuscript's own author response concedes that "T_ideal and T_full−T_ideal should not be interpreted as uniquely identifiable psychological or causal variables." This concession is in tension with the abstract and conclusion, which state that the framework "validates that staged social modeling is critical" and that the residual "recovers local reactive adjustments." The claims in the abstract, Section 1, and Section 5 should be revised to the more limited formulation of a planning-oriented reference and a reaction-oriented correction, with the non-identifiability limitation stated in the main text.
  4. [§4.2–§4.3, Tables 1–3] All reported results are single numbers with no error bars, confidence intervals, or significance tests, and many of the gains are small (for example, Ours (DeMo) vs. DeMo: minADE1 1.48 vs. 1.49, minFDE1 3.71 vs. 3.74; Ours (QCNet) vs. QCNet: minADE6 0.64 vs. 0.65). Given that the central claim is "consistent improvements" across benchmarks, the paper should report variance across at least three seeds or paired significance tests to establish that the improvements are not noise.
minor comments (4)
  1. [Abstract and throughout] The method name is typeset inconsistently as "INTraJ" in the abstract and as "INTraJ" elsewhere; please standardize the spelling.
  2. [§4.7 and Figure 5] The robustness subsection is interrupted by the embedded author response block, with Figure 5 appearing after that block. The figure and the surrounding discussion should be moved into Section 4.7, and the author-response text should be removed from the camera-ready version or integrated into the main paper if it is meant to be part of the submission.
  3. [Table 3 caption] The caption says "the proposed benchmark test split" without identifying the benchmark; please specify that this is the AV2-ped split described in §4.1.
  4. [§3.5, Eq. (12)] The hyperparameter λ_end appears in the planning reference loss but is not listed among the balancing coefficients or given a value; please report its value and whether it is tuned.

Circularity Check

1 steps flagged · score 3.0 of 10

The planning/reaction decomposition is self-definitional: the final trajectory is a gated blend of the backbone output and a smoothed-future reference, so the empirical gains do not independently validate the claimed staged social mechanism.

  1. self definitional [Sec. 3.2, Eqs. (4) and (5)/(13); Sec. 3.5, Eq. (12); Author Response, 'Novelty and scope']
    "We define the structural difference between the two predictions as: S_high = T_full − T_ideal. (4) S_high approximates the local reactive adjustments that deviate from the overall motion plan. [...] T_final = T_ideal + α · S_high. (5) [...] We obtain a smoothed target trajectory T_smooth by applying mild smoothing to the ground-truth future trajectory, and use it as the supervision target: Lref = Ltraj(T_ideal, T_smooth)+λ_end Lend(T_ideal, T_smooth), (12) [...] (Tideal) and (Tfull−Tideal) should not be interpreted as uniquely identifiable psychological or causal variables."

    By Eqs. (4) and (5)/(13), the reaction term is definitionally the residual between two learned outputs and the final forecast is a convex combination: T_final = α·T_full + (1−α)·T_ideal. The planning/reaction decomposition is therefore a partition of the model's own outputs, not an independently measured staging of social influence. Moreover, T_ideal is trained against a smoothed version of the ground-truth future (Eq. 12), so any low-pass regressor could serve as the 'planning reference'; the social carrier is not isolated. Table 5 confirms that without the learned gate the two-stage output collapses to the backbone (T_final = T_full). The authors concede that the terms are not uniquely identifiable psychological or causal variables.

full rationale

The central self-definitional step is that 'reaction' is defined as the residual between the full-context prediction and the planning reference, and the final output is just the gated weighted sum of those two quantities. This makes the planning/reaction conceptual framing a tautological partition of the model's own outputs, and the use of a smoothed ground-truth trajectory to supervise T_ideal means the 'planning reference' is not independently identified as a socially shaped plan. However, no self-citation chain or imported uniqueness theorem is involved, and the empirical comparisons against external benchmarks are genuine: the test-set ADE/FDE numbers are not constructed from the model's definitions. The main circularity burden is therefore interpretive rather than statistical, and the authors' explicit concession that the two terms are not uniquely identifiable psychological or causal variables further limits the severity. Missing sensitivity analyses for smoothing strength and the absence of a non-social low-pass control are correctness risks rather than additional circular steps.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method introduces no new physical entities; it relies on a learned gate and loss weights, and on the assumption that smoothing the ground truth isolates a planning component.

free parameters (3)
  • gating weight alpha = learned per mode and time step
    Equation (5) controls the fusion of planning reference and full prediction; the ablation shows it is the main source of improvement.
  • loss balancing coefficients lambda1, lambda2 = not reported
    Equation (14) weights the planning-reference loss and auxiliary gate regularizer; values are chosen by hand and not documented.
  • smoothing strength for T_smooth = not reported
    Equation (12) uses a mild smoothing of the ground-truth trajectory as the target for T_ideal; the exact low-pass filter is unspecified and no sensitivity study is given.
assumptions (4)
  • domain assumption Agents form motion plans by anticipating others' future behaviors before making local reactive adjustments.
    Introduction, Section 1: this behavioral premise motivates the two-stage architecture, but is not directly tested against alternative interaction orders.
  • domain assumption Future social information can be encoded by predicted neighbor trajectories or by a static-ego latent response.
    Equations (6) and (10): the multi-target and single-target instantiations rely on these carriers being informative; the author response adds robustness analysis but no failure-mode analysis of a strong predictor.
  • domain assumption A smoothed version of the ground-truth future is a suitable supervision target for the planning reference.
    Equation (12), Section 3.5: the reference trajectory is trained against T_smooth; if the smoothing removes dynamics that matter for FDE, the decomposition gains could be an artifact.
  • domain assumption Linear residual fusion with a per-step learned gate preserves multimodal prediction quality.
    Equation (5) and the gating regularizer in Section 3.5: the gate must avoid collapse, and no theoretical guarantee is given that the convex combination improves calibration.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Unified Framework for Trajectory Prediction with Explicit Planning and Reaction Decomposition." pith.science (2026). https://pith.science/paper/PWSH6XMC

@misc{pith2026260805673,
  author       = {Pith},
  title        = {Pith review of: A Unified Framework for Trajectory Prediction with Explicit Planning and Reaction Decomposition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PWSH6XMC}},
  note         = {Machine review of arXiv:2608.05673}
}
read the original abstract

Trajectory prediction has shifted toward structured formulations with explicit social modeling. However, existing methods inadequately distinguish the functional roles of social influence in trajectory planning. Observing that agents typically form motion plans by anticipating others' future behaviors before making local reactive adjustments, we identify social interactions as playing staged roles, namely planning precedes reaction. We propose INTraJ, a unified framework that decomposes social influence into two stages: a planning stage constructs reference trajectories using future social information, and a reaction stage recovers local adjustments from the residual between full-context prediction and the reference. INTraJ supports both multi-target and single-target paradigms. Extensive experiments on four standard benchmarks, including Argoverse 2, Argoverse 2-ped, ETH/UCY, and SDD, demonstrate consistent improvements, particularly in FDE and long-horizon consistency, with state-of-the-art performance achieved in several settings. INTraJ reframes trajectory prediction as a planning-driven two-stage process, validating that staged social modeling is critical for stable predictions. The code is publicly available at https://github.com/11isnotavailable/INTraJ.

Figures

Figures reproduced from arXiv: 2608.05673 by the authors.

Figure 1
Figure 1. Illustration of socially shaped planning. Starting [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of our INTraJ framework. INTraJ unifies the Multi-Target and Single-Target paradigms under a two-stage [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Qualitative results on ETH/UCY. Compared with full, final alleviates the entanglement caused by coupled prediction [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Qualitative results on AV2. ideal is the planning-level reference. full denotes the backbone prediction with full input. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 25 canonical work pages

  1. [1]

    Alexandre Alahi, Kratarth Goel, Vignesh Ramanathan, Alexandre Robicquet, Li Fei-Fei, and Silvio Savarese. 2016. Social LSTM: Human Trajectory Prediction in Crowded Spaces. In2016 IEEE Conference on Computer Vision and Pattern Recognition. IEEE Computer Society, Las Vegas, NV, USA, 961–971. doi:10.1109/ CVPR.2016.110

    work page 2016

  2. [2]

    Joshua Andle, Nicholas Soucy, Simon Socolow, and Salimeh Yasaei Sekeh. 2023. The Stanford Drone Dataset Is More Complex than We Think: An Analysis of Key Characteristics.IEEE Transactions on Intelligent Vehicles8, 2 (2023), 1863–1873. doi:10.1109/TIV.2022.3166642

    arXiv 2023

  3. [5]

    Pranav Singh Chib, Achintya Nath, Paritosh Kabra, Ishu Gupta, and Pravendra Singh. 2024. MS-TIP: Imputation Aware Pedestrian Trajectory Prediction. In Forty-First International Conference on Machine Learning, Ruslan Salakhutdinov, Zico Kolter, Katherine A. Heller, Adrian Weller, Nuria Oliver, Jonathan Scarlett, and Felix Berkenkamp (Eds.). PMLR / OpenRevi...

    work page 2024

  4. [6]

    Pranav Singh Chib and Pravendra Singh. 2025. LG-Traj: LLM Guided Pedestrian Trajectory Prediction. InIEEE/CVF International Conference on Computer Vision. IEEE, Honolulu, HI, USA, 6861–6871. doi:10.1109/ICCVW69036.2025.00709

    arXiv 2025

  5. [7]

    Alexander Cui, Sergio Casas, Kelvin Wong, Simon Suo, and Raquel Urtasun

    work page

  6. [8]

    Yuxiang Fu, Qi Yan, Lele Wang, Ke Li, and Renjie Liao. 2025. MoFlow: One- Step Flow Matching for Human Trajectory Forecasting via Implicit Maximum Likelihood Estimation Based Distillation. InIEEE/CVF Conference on Computer Vision and Pattern Recognition. Computer Vision Foundation / IEEE, Nashville, TN, USA, 17282–17293. doi:10.1109/CVPR52734.2025.01611

    arXiv 2025

  7. [10]

    Xing Gao, Xiaogang Jia, Yikang Li, and Hongkai Xiong. 2023. Dynamic Scenario Representation Learning for Motion Forecasting With Heterogeneous Graph Convolutional Recurrent Networks.IEEE Robotics and Automation Letters8, 5 (2023), 2946–2953. doi:10.1109/LRA.2023.3262150

    arXiv 2023

  8. [11]

    Thomas Gilles, Stefano Sabatini, Dzmitry Tsishkou, Bogdan Stanciulescu, and Fabien Moutarde. 2022. THOMAS: Trajectory Heatmap Output with Learned Multi-Agent Sampling. InInternational Conference on Learning Representations. OpenReview.net, Virtual Event

    work page 2022

Show all 46 references
  1. [12]

    Markus Knoche, Daan de Geus, and Bastian Leibe. 2025. DONUT: A Decoder- Only Model for Trajectory Prediction.CoRRabs/2506.06854 (2025). doi:10.48550/ ARXIV.2506.06854

    2025 doi arXiv

  2. [14]

    Shijie Li, Xun Xu, Si Yong Yeo, and Xulei Yang. 2025. Future-Aware Interaction Network for Motion Forecasting.CoRRabs/2503.06565 (2025). doi:10.48550/ ARXIV.2503.06565

    2025 doi arXiv

  3. [15]

    Yuhang Li, Changsheng Li, Ruilin Lv, Rongqing Li, Ye Yuan, and Guoren Wang

  4. [16]

    Xiaotong Lin, Tianming Liang, Jianhuang Lai, and Jian-Fang Hu. 2024. Progressive Pretext Task Learning for Human Trajectory Prediction. InEuropean Conference on Computer Vision, Ales Leonardis, Elisa Ricci, Stefan Roth, Olga Russakovsky, Torsten Sattler, and Gül Varol (Eds.). ...

    2024

  5. [17]

    Yu Liu, Zhijie Liu, Xiao Ren, Youfu Li, and He Kong. 2026. Intention-Aware Diffusion Model for Pedestrian Trajectory Prediction. InAAAI Conference on Artificial Intelligence, Sven Koenig, Chad Jenkins, and Matthew E. Taylor (Eds.). AAAI Press, Singapore, 18469–18477. doi:10.16...

    2026 doi

  6. [18]

    Sheng, and Lina Yao

    Yao Liu, Zesheng Ye, Rui Wang, Binghao Li, Quan Z. Sheng, and Lina Yao. 2024. Uncertainty-Aware Pedestrian Trajectory Prediction via Distributional Diffusion. Knowledge-Based Systems296 (2024), 111862. doi:10.1016/J.KNOSYS.2024.111862

    2024

  7. [21]

    Karttikeya Mangalam, Harshayu Girase, Shreyas Agarwal, Kuan-Hui Lee, Ehsan Adeli, Jitendra Malik, and Adrien Gaidon. 2020. It Is Not the Journey But the Destination: Endpoint Conditioned Trajectory Prediction. InEuropean Conference on Computer Vision (Lecture Notes in Computer...

    2020 doi

  8. [22]

    Francesco Marchetti, Federico Becattini, Lorenzo Seidenari, and Alberto Del Bimbo. 2024. SMEMO: Social Memory for Trajectory Forecasting.IEEE Trans- actions on Pattern Analysis and Machine Intelligence46, 6 (2024), 4410–4425. doi:10.1109/TPAMI.2024.3356755

    2024

  9. [23]

    Mohamed, Kun Qian, Mohamed Elhoseiny, and Christian G

    Abduallah A. Mohamed, Kun Qian, Mohamed Elhoseiny, and Christian G. Claudel

  10. [24]

    Muleilan Pei, Shaoshuai Shi, Xuesong Chen, Xu Liu, and Shaojie Shen. 2025. Foresight in Motion: Reinforcing Trajectory Prediction with Reward Heuristics. CoRRabs/2507.12083 (2025). doi:10.48550/ARXIV.2507.12083

    2025 doi arXiv

  11. [26]

    Mozhgan Pourkeshavarz, Junrui Zhang, and Amir Rasouli. 2024. DySeT: A Dynamic Masked Self-distillation Approach for Robust Trajectory Prediction. In European Conference on Computer Vision (Lecture Notes in Computer Science), Ales Leonardis, Elisa Ricci, Stefan Roth, Olga Russa...

    2024 doi

  12. [27]

    Ziyang Ren, Ping* Wei, Shangqi Deng, Haowen Tang, Jiapeng Li, and Huan Li

  13. [28]

    Kitani, Dariu M

    Andrey Rudenko, Luigi Palmieri, Michael Herman, Kris M. Kitani, Dariu M. Gavrila, and Kai O. Arras. 2020. Human Motion Trajectory Prediction: A Sur- vey.The International Journal of Robotics Research39, 8 (2020). doi:10.1177/ 0278364920917446

    2020

  14. [29]

    Tim Salzmann, Boris Ivanovic, Punarjay Chakravarty, and Marco Pavone. 2020. Trajectron++: Dynamically-Feasible Trajectory Forecasting with Heterogeneous Data. InEuropean Conference on Computer Vision (Lecture Notes in Computer Science), Andrea Vedaldi, Horst Bischof, Thomas Br...

    2020 doi

  15. [30]

    Liushuai Shi, Le Wang, Chengjiang Long, Sanping Zhou, Wei Tang, Nanning Zheng, and Gang Hua. 2023. Representing Multimodal Behaviors with Mean Location for Pedestrian Trajectory Prediction.IEEE Transactions on Pattern Analysis and Machine Intelligence45, 9 (2023), 11184–11202....

    2023

  16. [31]

    Shaoshuai Shi, Li Jiang, Dengxin Dai, and Bernt Schiele. 2022. Motion Transformer with Global Intention Localization and Local Movement Refinement. InAnnual Conference on Neural Information Processing Systems, Sanmi Koyejo, S. Mohamed, A. Agarwal, Danielle Belgrave, K. Cho, an...

    2022 doi

  17. [32]

    Yuchao Su, Yuanman Li, Wei Wang, Jiantao Zhou, and Xia Li. 2024. A Unified Environmental Network for Pedestrian Trajectory Prediction. InAAAI Conference on Artificial Intelligence, Michael J. Wooldridge, Jennifer G. Dy, and Sriraam Natarajan (Eds.). AAAI Press, Vancouver, Cana...

    2024 doi

  18. [33]

    Nico Uhlemann, Felix Fent, and Markus Lienkamp. 2024. Evaluating Pedestrian Trajectory Prediction Methods With Respect to Autonomous Driving.IEEE Transactions on Intelligent Transportation Systems25, 10 (2024), 13937–13946. doi:10.1109/TITS.2024.3386195

    2024

  19. [34]

    Nico Uhlemann, Yipeng Zhou, Tobias Simeon Mohr, and Markus Lienkamp

  20. [35]

    Mingkun Wang, Xinge Zhu, Changqian Yu, Wei Li, Yuexin Ma, Ruochun Jin, Xiaoguang Ren, Dongchun Ren, Mingxu Wang, and Wenjing Yang. 2023. GANet: Goal Area Network for Motion Forecasting. InIEEE International Conference on MM ’26, November 10–14, 2026, Rio de Janeiro, Brazil Jia...

    2023

  21. [37]

    Benjamin Wilson, William Qi, Tanmay Agarwal, John Lambert, Jagjeet Singh, Siddhesh Khandelwal, Bowen Pan, Ratnesh Kumar, Andrew Hartnett, Jhony Kae- semodel Pontes, Deva Ramanan, Peter Carr, and James Hays. 2021. Argoverse 2: Next Generation Datasets for Self-Driving Perceptio...

    2021

  22. [39]

    Conghao Wong, Ziqian Zou, and Beihao Xia. 2025. Resonance: Learning to Predict Social-Aware Pedestrian Trajectories as Co-Vibrations. InProceedings of the IEEE/CVF International Conference on Computer Vision. Honolulu, HI, USA, 25788–25799

    2025

  23. [40]

    InIEEE/CVF Winter Conference on Applications of Computer Vision

    Snapshot: Towards Application-Centered Models for Pedestrian Trajectory Prediction in Urban Traffic Environments. InIEEE/CVF Winter Conference on Applications of Computer Vision. IEEE, Tucson, AZ, USA, 1062–1072. doi:10.1109/ WACVW65960.2025.00128

  24. [41]

    Heming Yang, Yu Tian, Changyuan Tian, Hongfeng Yu, Wanxuan Lu, Chubo Deng, and Xian Sun. 2025. SoPerModel: Leveraging Social Perception for Multi- Agent Trajectory Prediction.IEEE Trans. Geosci. Remote. Sens.63 (2025), 1–13. doi:10.1109/TGRS.2025.3543661

    2025

  25. [42]

    Ye Yuan, Xinshuo Weng, Yanglan Ou, and Kris Kitani. 2021. AgentFormer: Agent- Aware Transformers for Socio-Temporal Multi-Agent Forecasting. InIEEE/CVF International Conference on Computer Vision. IEEE, Montreal, QC, Canada, 9793–

    2021

  26. [43]

    Bozhou Zhang, Nan Song, and Li Zhang. 2024. DeMo: Decoupling Motion Fore- casting into Directional Intentions and Dynamic States. InAnnual Conference on Neural Information Processing Systems, Amir Globersons, Lester Mackey, Danielle Belgrave, Angela Fan, Ulrich Paquet, Jakub M...

    2024 doi

  27. [44]

    Lu Zhang, Peiliang Li, Sikang Liu, and Shaojie Shen. 2024. SIMPL: A Simple and Efficient Multi-Agent Motion Prediction Baseline for Autonomous Driving.IEEE Robotics and Automation Letters9, 4 (2024), 3767–3774. doi:10.1109/LRA.2024. 3370039

    2024 doi

  28. [45]

    Waslander, Hongsheng Li, and Yu Liu

    Yang Zhou, Hao Shao, Letian Wang, Steven L. Waslander, Hongsheng Li, and Yu Liu. 2024. SmartRefine: A Scenario-Adaptive Refinement Framework for Efficient Motion Prediction. InIEEE/CVF Conference on Computer Vision and Pattern Recog- nition. IEEE, Seattle, WA, USA, 15281–15290...

    2024

  29. [46]

    Beihao Xia, Conghao Wong, Duanquan Xu, Qinmu Peng, and Xinge You. 2025. Another Vertical View: A Hierarchical Network for Heterogeneous Trajectory Prediction via Spectrums.IEEE Transactions on Pattern Analysis and Machine Intelligence47, 11 (2025), 9749–9766. doi:10.1109/TPAMI...

    2025

  30. [47]

    Zikang Zhou, Luyao Ye, Jianping Wang, Kui Wu, and Kejie Lu. 2022. HiVT: Hierarchical Vector Transformer for Multi-Agent Motion Prediction. InIEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, New Orleans, LA, USA, 8813–8823. doi:10.1109/CVPR52688.2022.00862

    2022

  31. [48]

    Yiyao Zhu, Di Luan, and Shaojie Shen. 2023. BiFF: Bi-level Future Fusion with Polyline-based Coordinate for Interactive Trajectory Prediction. InIEEE/CVF International Conference on Computer Vision. IEEE, Paris, France, 8226–8237. doi:10.1109/ICCV51070.2023.00759

    2023

  32. [53]

    Zikang Zhou, Jianping Wang, Yung-Hui Li, and Yu-Kai Huang. 2023. Query- Centric Trajectory Prediction. InIEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, Vancouver, BC, Canada, 17863–17873. doi:10.1109/ CVPR52729.2023.01713

    2023

  33. [2020]

    In2020 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition

    Social-STGCNN: A Social Spatio-Temporal Graph Convolutional Neural Network for Human Trajectory Prediction. In2020 IEEE/CVF Conference on Com- puter Vision and Pattern Recognition. Computer Vision Foundation / IEEE, Seattle, WA, USA, 14412–14420. doi:10.1109/CVPR42600.2020.01443

  34. [2023]

    InIEEE International Conference on Robotics and Automation

    GoRela: Go Relative for Viewpoint-Invariant Motion Forecasting. InIEEE International Conference on Robotics and Automation. IEEE, London, UK, 7801–

  35. [2024]

    InAnnual Conference on Neural Information Processing Systems, Amir Globersons, Lester Mackey, Danielle Belgrave, Angela Fan, Ulrich Paquet, Jakub M

    LaKD: Length-agnostic Knowledge Distillation for Trajectory Prediction with Any Length Observations. InAnnual Conference on Neural Information Processing Systems, Amir Globersons, Lester Mackey, Danielle Belgrave, Angela Fan, Ulrich Paquet, Jakub M. Tomczak, and Cheng Zhang (E...

    doi

  36. [2025]

    InProceedings of the IEEE/CVF International Conference on Computer Vision

    TOTP: Transferable Online Pedestrian Trajectory Prediction with Temporal- Adaptive Mamba Latent Diffusion. InProceedings of the IEEE/CVF International Conference on Computer Vision. Honolulu, HI, USA, 26263–26272

  37. [7807]

    doi:10.1109/ICRA48891.2023.10160984

    2023

  38. [9803]

    doi:10.1109/ICCV48922.2021.00967

    2021

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.