pith. sign in

arxiv: 2606.06014 · v1 · pith:IGZB46YFnew · submitted 2026-06-04 · 💻 cs.AI · cs.RO

PLAN-S: Bridging Planning with Latent Style Dynamics for Autonomous Driving World Models

Xiaoyun Qiu , Jingtao He , Yijie Chen , Yusong Huang , Haotian Wang , Yixuan Wang , Xinhu Zheng This is my paper

Pith reviewed 2026-06-28 01:12 UTC · model grok-4.3

classification 💻 cs.AI cs.RO
keywords autonomous drivinglatent world modelstrajectory planningsemantic cost mapsdriving stylesnuScenesNAVSIMend-to-end planning
0
0 comments X

The pith

Decoding a style-conditioned four-channel semantic cost map from latent representations allows upstream fusion that reduces trajectory errors and collision rates in autonomous driving planners.

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

The paper seeks to resolve the compactness-controllability issue in latent world models for end-to-end autonomous driving, where entangled latents make it difficult to explicitly handle risk, drivability, and driving style preferences before generating trajectories. PLAN-S introduces a bridge that decodes a style-conditioned four-channel semantic cost map from the latent state and feeds it into the planner through attention or reward fusion interfaces while keeping the host model frozen. This setup enables supervision and modulation of style dynamics at the cost-map level rather than only at the final trajectory. Experiments on nuScenes and NAVSIM report lower L2 errors across horizons, reduced collision rates, and higher PDMS scores, with ablations attributing gains primarily to the cost pathway.

Core claim

PLAN-S bridges latent world models to planning by decoding a style-conditioned, four-channel semantic cost map from the latent representation. The cost map is conditioned on ego state and driving style and is consumed upstream of the planning decision through attention-level fusion for regression planners and reward-level fusion for anchor-score planners. Validation on two frozen host architectures shows consistent metric gains: 0.55 m average L2 and 42 percent relative reduction in 3 s collision rate on nuScenes, and 89.4 PDMS on NAVSIM for the rule-cost variant, with complementary gains from the learned-cost variant on challenging scenes.

What carries the argument

The style-conditioned four-channel semantic cost map decoded from latent representations and fused upstream of planning via attention-level or reward-level interfaces.

If this is right

  • L2 trajectory error decreases at every prediction horizon relative to the frozen baseline.
  • 3-second collision rate drops by 42 percent on nuScenes.
  • Rule-cost variant reaches 89.4 PDMS on NAVSIM while learned-cost variant adds gains on hard scenes.
  • Cost pathway contributes most directly to safer trajectory selection according to ablations.
  • Diverse, spatially consistent cost maps can be generated for different driving styles.

Where Pith is reading between the lines

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

  • The bridge architecture could be inserted into other latent planners without retraining the world model backbone.
  • Changing the style conditioning input at inference time might enable online style switching without additional training.
  • Hybrid systems that combine the decoded cost map with external rule sets could be tested for further safety margins.
  • The approach may transfer to longer-horizon forecasting if the latent encoder preserves style information over extended sequences.

Load-bearing premise

The latent representations already encode sufficient disentangled information about risk, drivability, and style preferences that a four-channel semantic cost map can be decoded and fused upstream of planning without degrading the host planner.

What would settle it

Re-running the nuScenes and NAVSIM evaluations with the cost-map decoder removed or replaced by random maps while keeping hosts frozen, and finding no reduction in L2 or collision rate, would falsify the claim that the decoded cost map drives the observed improvements.

Figures

Figures reproduced from arXiv: 2606.06014 by Haotian Wang, Jingtao He, Xiaoyun Qiu, Xinhu Zheng, Yijie Chen, Yixuan Wang, Yusong Huang.

Figure 1
Figure 1. Figure 1: Overview of the PLAN-S framework. (a) The frozen perception encoder and latent world model produce current and future bird’s-eye-view (BEV) features. (b) The trainable cost-map decoder produces a four-channel semantic cost map from the BEV latent, conditioned on ego state and driving style via dual AdaFiLM. The cost map then guides planning through matched coupling interfaces for anchor-score and regressio… view at source ↗
Figure 2
Figure 2. Figure 2: PDMS stratified by reproduced WoTE baseline difficulty on NAVSIM [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: shows a representative NAVSIM navtest scene at a curved urban intersection. The top row provides the multi-view camera inputs and the front-view trajectory overlay, where green denotes the model prediction and red denotes the GT trajectory. The bottom row compares BEV planning outputs. Compared with DiffusionDrive, PLAN-S keeps a smoother predicted trajectory with larger clearance from nearby agents. The r… view at source ↗
Figure 4
Figure 4. Figure 4: Style-conditioned cost-map comparison on a NAVSIM [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
read the original abstract

Latent world models (LWMs) have strengthened end-to-end autonomous driving by forecasting compact scene dynamics for downstream planning. However, existing LWM-based planners usually generate trajectories directly from entangled latent representations. This compact latent-to-planner pathway lacks explicit modeling of risk, drivability, and diverse style preferences, making driving-style dynamics difficult to supervise, inspect, or modulate before a final trajectory is selected. We propose PLAN-S (PLANning with latent Style dynamics), a planner-facing bridge that addresses this compactness-controllability dilemma by decoding a style-conditioned, four-channel semantic cost map from the latent representation. The cost map is conditioned on ego state and driving style and is consumed up-stream of the planning decision through two host-side interfaces: attention-level fusion for regression planners and reward-level fusion for anchor-score planners. We validate PLAN-S on two architecturally distinct hosts, ResWorld on nuScenes and WoTE on NAVSIM, while keeping the host backbones frozen to isolate the contribution of the proposed bridge. On nuScenes, PLAN-S reduces L2 at every horizon over the baseline, with 0.55 m average L2 and a 42% relative reduction in the 3 s collision rate. On NAVSIM, the rule-cost variant reaches 89.4 Predictive Driver Model Score (PDMS), while the learned cost variant provides complementary gains on baseline-challenging scenes. Ablations show that the cost pathway contributes most directly to safer trajectory selection. Qualitative results further show that PLAN-S can produce diverse cost maps, with spatially consistent variations aligned to different driving styles.

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

3 major / 1 minor

Summary. The paper proposes PLAN-S, a bridge module that decodes a style-conditioned four-channel semantic cost map from frozen latent world model representations. The cost map is conditioned on ego state and driving style and fused upstream of frozen host planners (ResWorld on nuScenes via attention-level fusion; WoTE on NAVSIM via reward-level fusion) to enable explicit modeling of risk, drivability, and style preferences. Reported results include 0.55 m average L2 error with 42% relative 3 s collision-rate reduction on nuScenes and 89.4 PDMS (rule-cost variant) on NAVSIM, with ablations attributing gains primarily to the cost pathway and qualitative results showing style-aligned cost-map variations.

Significance. If the experimental claims hold under full scrutiny, the work demonstrates a modular, host-agnostic way to improve controllability and safety metrics in LWM-based planners without retraining backbones. The frozen-host protocol and dual fusion interfaces are strengths that isolate the bridge contribution and support style modulation.

major comments (3)
  1. [Experiments] Experiments section: the reported gains (0.55 m avg L2, 42 % collision reduction on nuScenes; 89.4 PDMS on NAVSIM) are presented without error bars, run-to-run variance, or statistical significance tests, leaving open whether the improvements exceed baseline variability.
  2. [Methods] Methods / latent-representation analysis: the central claim that the latent already encodes sufficient disentangled risk/drivability/style information for a four-channel cost map to be decoded and fused upstream is load-bearing, yet no probing, mutual-information analysis, or controlled visualization of the latent factors is described to substantiate disentanglement versus post-hoc mapping.
  3. [Ablations] Ablation studies: while the text states that ablations show the cost pathway contributes most to safer selection, no table or quantitative deltas are supplied for the individual components (style conditioning, four-channel decoding, fusion type), preventing verification that the bridge—not ancillary changes—drives the reported metrics.
minor comments (1)
  1. [Abstract] Abstract: the phrase 'spatially consistent variations aligned to different driving styles' is used without defining a quantitative measure of consistency or diversity.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback on experimental reporting, latent analysis, and ablation details. We address each major comment below and will revise the manuscript to incorporate the suggested improvements.

read point-by-point responses

  1. Referee: Experiments section: the reported gains (0.55 m avg L2, 42 % collision reduction on nuScenes; 89.4 PDMS on NAVSIM) are presented without error bars, run-to-run variance, or statistical significance tests, leaving open whether the improvements exceed baseline variability.

    Authors: We agree that the absence of error bars, variance across runs, and statistical tests limits the strength of the claims. In the revised manuscript we will report results from multiple independent training and evaluation runs, include standard deviations, and add statistical significance tests (e.g., paired t-tests) to confirm that the observed improvements exceed baseline variability. revision: yes

  2. Referee: Methods / latent-representation analysis: the central claim that the latent already encodes sufficient disentangled risk/drivability/style information for a four-channel cost map to be decoded and fused upstream is load-bearing, yet no probing, mutual-information analysis, or controlled visualization of the latent factors is described to substantiate disentanglement versus post-hoc mapping.

    Authors: We acknowledge that additional analysis is needed to substantiate the claim of sufficient disentangled information in the latent space. While the downstream performance and qualitative cost-map variations provide supporting evidence, we will add probing experiments, including controlled visualizations of latent factors and mutual-information estimates between latent dimensions and risk/drivability/style attributes, in the revised version. revision: yes

  3. Referee: Ablation studies: while the text states that ablations show the cost pathway contributes most to safer selection, no table or quantitative deltas are supplied for the individual components (style conditioning, four-channel decoding, fusion type), preventing verification that the bridge—not ancillary changes—drives the reported metrics.

    Authors: We agree that a quantitative ablation table with explicit deltas for each component is necessary for verification. We will expand the ablation section in the revised manuscript to include a detailed table reporting metrics for ablations of style conditioning, channel count, and fusion type, with clear quantitative comparisons to the full model. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The manuscript describes PLAN-S as an auxiliary decoding bridge that produces a style-conditioned four-channel cost map from a frozen latent world model and fuses it upstream of two distinct host planners (ResWorld, WoTE) whose backbones remain frozen. No equations, parameter-fitting steps, or self-citation chains are supplied that would reduce the reported L2, collision-rate, or PDMS gains to quantities defined by the same fitted values. The isolation claim rests on external benchmark numbers rather than internal redefinition, satisfying the criteria for a self-contained empirical addition.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central addition is the newly proposed PLAN-S module and the assumption that latent states contain decodable style and risk information; no free parameters or invented physical entities are mentioned.

axioms (1)
  • domain assumption Latent representations produced by existing world models already contain sufficient information about driving style, risk, and drivability to be decoded into a useful four-channel semantic cost map.
    This premise is required for the decoder to produce actionable cost maps that improve planning.
invented entities (1)
  • PLAN-S bridge no independent evidence
    purpose: Decode style-conditioned semantic cost map from latent state for upstream fusion with planners.
    New architectural component introduced by the paper.

pith-pipeline@v0.9.1-grok · 5839 in / 1479 out tokens · 19587 ms · 2026-06-28T01:12:49.557995+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

42 extracted references · 7 canonical work pages · 3 internal anchors

  1. [1]

    Occworld: Learning a 3d occupancy world model for autonomous driving,

    W. Zheng, W. Chen, Y . Huang, B. Zhang, Y . Duan, and J. Lu, “Occworld: Learning a 3d occupancy world model for autonomous driving,” in European conference on computer vision. Springer, 2024, pp. 55–72

    2024

  2. [2]

    Driveworld: 4d pre-trained scene understanding via world models for autonomous driving,

    C. Min, D. Zhao, L. Xiao, J. Zhao, X. Xu, Z. Zhu, L. Jin, J. Li, Y . Guo, J. Xinget al., “Driveworld: 4d pre-trained scene understanding via world models for autonomous driving,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2024, pp. 15 522–15 533

    2024

  3. [3]

    Bevworld: A multimodal world model for autonomous driving via unified bev latent space,

    Y . Zhang, S. Gong, K. Xiong, X. Ye, X. Tan, F. Wang, J. Huang, H. Wu, and H. Wang, “Bevworld: A multimodal world model for autonomous driving via unified bev latent space,”arXiv preprint arXiv:2407.05679, 2024

    work page arXiv 2024

  4. [4]

    Resworld: Temporal residual world model for end-to-end autonomous driving,

    J. Zhang, Z. Fu, Q. Liu, Y . Wanget al., “Resworld: Temporal residual world model for end-to-end autonomous driving,” inThe Fourteenth International Conference on Learning Representations, 2026

    2026

  5. [5]

    Planning-oriented autonomous driving,

    Y . Hu, J. Yang, L. Chen, K. Li, C. Sima, X. Zhu, S. Chai, S. Du, T. Lin, W. Wanget al., “Planning-oriented autonomous driving,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 17 853–17 862

    2023

  6. [6]

    Vad: Vectorized scene representation for efficient autonomous driving,

    B. Jiang, S. Chen, Q. Xu, B. Liao, J. Chen, H. Zhou, Q. Zhang, W. Liu, C. Huang, and X. Wang, “Vad: Vectorized scene representation for efficient autonomous driving,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 8340–8350

    2023

  7. [7]

    Genad: Generative end-to-end autonomous driving,

    W. Zheng, R. Song, X. Guo, C. Zhang, and L. Chen, “Genad: Generative end-to-end autonomous driving,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 87–104

    2024

  8. [8]

    Hydra-MDP: End-to-end Multimodal Planning with Multi-target Hydra-Distillation

    Z. Li, K. Li, S. Wang, S. Lan, Z. Yu, Y . Ji, Z. Li, Z. Zhu, J. Kautz, Z. Wu et al., “Hydra-mdp: End-to-end multimodal planning with multi-target hydra-distillation,”arXiv preprint arXiv:2406.06978, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  9. [9]

    Diffusiondrive: Truncated diffusion model for end-to-end autonomous driving,

    B. Liao, S. Chen, H. Yin, B. Jiang, C. Wang, S. Yan, X. Zhang, X. Li, Y . Zhang, Q. Zhanget al., “Diffusiondrive: Truncated diffusion model for end-to-end autonomous driving,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 12 037–12 047

    2025

  10. [10]

    End-to-end driving with online trajectory evaluation via bev world model,

    Y . Li, Y . Wang, Y . Liu, J. He, L. Fan, and Z. Zhang, “End-to-end driving with online trajectory evaluation via bev world model,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 27 137–27 146. SUBMITTED TO IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 11

    2025

  11. [11]

    Styledrive: Towards driving-style aware benchmarking of end-to-end autonomous driving,

    R. Hao, B. Jing, H. Yu, and Z. Nie, “Styledrive: Towards driving-style aware benchmarking of end-to-end autonomous driving,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 40, no. 6, 2026, pp. 4627–4635

    2026

  12. [12]

    Driving with A Thousand Faces: A Benchmark for Closed-Loop Personalized End-to-End Autonomous Driving

    X. Dong, R. Li, X. Han, Z. Wu, J. Wang, J. Chen, Q. Jiang, S. Yiu, X. Zhu, and Y . Ma, “Driving with a thousand faces: A benchmark for closed-loop personalized end-to-end autonomous driving,”arXiv preprint arXiv:2602.18757, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  13. [13]

    Drive my way: Preference alignment of vision-language-action model for personalized driving,

    Z. Wang, H. Jiang, S. Dong, Y . Wang, H. Qiu, and J. Li, “Drive my way: Preference alignment of vision-language-action model for personalized driving,”arXiv preprint arXiv:2603.25740, 2026

    work page arXiv 2026

  14. [14]

    Film: Visual reasoning with a general conditioning layer,

    E. Perez, F. Strub, H. De Vries, V . Dumoulin, and A. Courville, “Film: Visual reasoning with a general conditioning layer,” inProceedings of the AAAI Conference on Artificial Intelligence, 2018

    2018

  15. [15]

    nuscenes: A multimodal dataset for autonomous driving,

    H. Caesar, V . Bankiti, A. H. Lang, S. V ora, V . E. Liong, Q. Xu, A. Krishnan, Y . Pan, G. Baldan, and O. Beijbom, “nuscenes: A multimodal dataset for autonomous driving,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 11 621–11 631

    2020

  16. [16]

    Navsim: Data-driven non-reactive autonomous vehicle simulation and benchmarking,

    D. Dauner, M. Hallgarten, T. Li, X. Weng, Z. Huang, Z. Yang, H. Li, I. Gilitschenski, B. Ivanovic, M. Pavoneet al., “Navsim: Data-driven non-reactive autonomous vehicle simulation and benchmarking,”Ad- vances in Neural Information Processing Systems, vol. 37, pp. 28 706– 28 719, 2024

    2024

  17. [17]

    Model-based imitation learning for urban driving,

    A. Hu, G. Corrado, N. Griffiths, Z. Murez, C. Gurau, H. Yeo, A. Kendall, R. Cipolla, and J. Shotton, “Model-based imitation learning for urban driving,”Advances in Neural Information Processing Systems, vol. 35, pp. 20 703–20 716, 2022

    2022

  18. [18]

    Scalable diffusion models with transformers,

    W. Peebles and S. Xie, “Scalable diffusion models with transformers,” in IEEE/CVF International Conference on Computer Vision (ICCV), 2023, pp. 4195–4205

    2023

  19. [19]

    GAIA-1: A Generative World Model for Autonomous Driving

    A. Hu, L. Russell, H. Yeo, Z. Murez, G. Fedoseev, A. Kendall, J. Shotton, and G. Corrado, “Gaia-1: A generative world model for autonomous driving,”arXiv preprint arXiv:2309.17080, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  20. [20]

    Driving into the future: Multiview visual forecasting and planning with world model for autonomous driving,

    Y . Wang, J. He, L. Fan, H. Li, Y . Chen, and Z. Zhang, “Driving into the future: Multiview visual forecasting and planning with world model for autonomous driving,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2024, pp. 14 749–14 759

    2024

  21. [21]

    Lift, splat, shoot: Encoding images from arbitrary camera rigs by implicitly unprojecting to 3d,

    J. Philion and S. Fidler, “Lift, splat, shoot: Encoding images from arbitrary camera rigs by implicitly unprojecting to 3d,” inEuropean conference on computer vision. Springer, 2020, pp. 194–210

    2020

  22. [22]

    Bevformer: Learning bird’s-eye-view representation from multi-camera images via spatiotemporal transformers,

    Z. Li, W. Wang, H. Li, E. Xie, C. Sima, T. Lu, Q. Yu, and J. Dai, “Bevformer: Learning bird’s-eye-view representation from multi-camera images via spatiotemporal transformers,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 47, no. 3, pp. 2020–2036, 2024

    2020

  23. [23]

    Bevdepth: Acquisition of reliable depth for multi-view 3d object de- tection,

    Y . Li, Z. Ge, G. Yu, J. Yang, Z. Wang, Y . Shi, J. Sun, and Z. Li, “Bevdepth: Acquisition of reliable depth for multi-view 3d object de- tection,” inProceedings of the AAAI conference on artificial intelligence, vol. 37, no. 2, 2023, pp. 1477–1485

    2023

  24. [24]

    Trajectory-guided control prediction for end-to-end autonomous driving: A simple yet strong baseline,

    P. Wu, X. Jia, L. Chen, J. Yan, H. Li, and Y . Qiao, “Trajectory-guided control prediction for end-to-end autonomous driving: A simple yet strong baseline,” inAdvances in Neural Information Processing Systems, 2022

    2022

  25. [25]

    Goalflow: Goal-driven flow matching for multimodal trajec- tories generation in end-to-end autonomous driving,

    Z. Xing, X. Zhang, Y . Hu, B. Jiang, T. He, Q. Zhang, X. Long, and W. Yin, “Goalflow: Goal-driven flow matching for multimodal trajec- tories generation in end-to-end autonomous driving,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 1602–1611

    2025

  26. [26]

    Sparsedrive: End-to-end autonomous driving via sparse scene representation,

    W. Sun, X. Lin, Y . Shi, C. Zhang, H. Wu, and S. Zheng, “Sparsedrive: End-to-end autonomous driving via sparse scene representation,” in2025 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2025, pp. 8795–8801

    2025

  27. [27]

    Parting with misconceptions about learning-based vehicle motion planning,

    D. Dauner, M. Hallgarten, A. Geiger, and K. Chitta, “Parting with misconceptions about learning-based vehicle motion planning,” inCon- ference on Robot Learning. PMLR, 2023, pp. 1268–1281

    2023

  28. [28]

    Mp3: A unified model to map, perceive, predict and plan,

    S. Casas, A. Sadat, and R. Urtasun, “Mp3: A unified model to map, perceive, predict and plan,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 14 403–14 412

    2021

  29. [29]

    End-to-end interpretable neural motion planner,

    W. Zeng, W. Luo, S. Suo, A. Sadat, B. Yang, S. Casas, and R. Urtasun, “End-to-end interpretable neural motion planner,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 8660–8669

    2019

  30. [30]

    St-p3: End-to-end vision-based autonomous driving via spatial-temporal feature learning,

    S. Hu, L. Chen, P. Wu, H. Li, J. Yan, and D. Tao, “St-p3: End-to-end vision-based autonomous driving via spatial-temporal feature learning,” inEuropean Conference on Computer Vision. Springer, 2022, pp. 533– 549

    2022

  31. [31]

    Learning to navigate intersections with unsupervised driver trait infer- ence,

    S. Liu, P. Chang, H. Chen, N. Chakraborty, and K. Driggs-Campbell, “Learning to navigate intersections with unsupervised driver trait infer- ence,” in2022 International Conference on Robotics and Automation (ICRA). IEEE, 2022, pp. 3576–3582

    2022

  32. [32]

    V ADv2: End-to-end vectorized autonomous driving via probabilistic planning,

    B. Jiang, S. Chen, H. Gao, B. Liao, Q. Zhang, W. Liu, and X. Wang, “V ADv2: End-to-end vectorized autonomous driving via probabilistic planning,” inThe Fourteenth International Conference on Learning Representations, 2026

    2026

  33. [33]

    Is ego status all you need for open-loop end-to-end autonomous driving?

    Z. Li, Z. Yu, S. Lan, J. Li, J. Kautz, T. Lu, and J. M. Alvarez, “Is ego status all you need for open-loop end-to-end autonomous driving?” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 14 864–14 873

    2024

  34. [34]

    Scene as occupancy,

    W. Tong, C. Sima, T. Wang, L. Chen, S. Wu, H. Deng, Y . Gu, L. Lu, P. Luo, D. Linet al., “Scene as occupancy,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 8406–8415

    2023

  35. [35]

    Para- drive: Parallelized architecture for real-time autonomous driving,

    X. Weng, B. Ivanovic, Y . Wang, Y . Wang, and M. Pavone, “Para- drive: Parallelized architecture for real-time autonomous driving,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 15 449–15 458

    2024

  36. [36]

    Navigation-guided sparse scene representation for end-to-end autonomous driving,

    P. Li and D. Cui, “Navigation-guided sparse scene representation for end-to-end autonomous driving,”arXiv preprint arXiv:2409.18341, 2024

    work page arXiv 2024

  37. [37]

    Transfuser: Imitation with transformer-based sensor fusion for au- tonomous driving,

    K. Chitta, A. Prakash, B. Jaeger, Z. Yu, K. Renz, and A. Geiger, “Transfuser: Imitation with transformer-based sensor fusion for au- tonomous driving,”IEEE transactions on pattern analysis and machine intelligence, vol. 45, no. 11, pp. 12 878–12 895, 2022

    2022

  38. [38]

    Enhancing end-to-end autonomous driving with latent world model,

    Y . Li, L. Fan, J. He, Y . Wang, Y . Chen, Z. Zhang, and T. Tan, “Enhancing end-to-end autonomous driving with latent world model,” in The Thirteenth International Conference on Learning Representations, 2025

    2025

  39. [39]

    Drama: An efficient end-to-end motion planner for autonomous driving with mamba,

    C. Yuan, Z. Zhang, J. Sun, S. Sun, Z. Huang, C. D. W. Lee, D. Li, Y . Han, A. Wong, K. P. Teeet al., “Drama: An efficient end-to-end motion planner for autonomous driving with mamba,”arXiv preprint arXiv:2408.03601, 2024

    work page arXiv 2024

  40. [40]

    Geobev: Learn- ing geometric bev representation for multi-view 3d object detection,

    J. Zhang, Y . Zhang, Y . Qi, Z. Fu, Q. Liu, and Y . Wang, “Geobev: Learn- ing geometric bev representation for multi-view 3d object detection,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 9, 2025, pp. 9960–9968

    2025

  41. [41]

    Decoupled weight decay regularization,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” inInternational Conference on Learning Representations, 2019. Xiaoyun Qiureceived the B.S. and M.S. degrees in transportation engineering from Harbin Institute of Technology, Harbin, China, in 2019 and 2021, respectively. She is currently working toward the Ph.D. degree in intelligent t...

    2019

  42. [42]

    degree in Intelligent Trans- portation from the Hong Kong University of Science and Technology (Guangzhou) in 2024, where he is currently pursuing the Ph.D

    He received M.S. degree in Intelligent Trans- portation from the Hong Kong University of Science and Technology (Guangzhou) in 2024, where he is currently pursuing the Ph.D. degree with the Intel- ligent Transportation Thrust. His current research interests include autonomous driving, cooperative perception and prediction, and intelligent transporta- tion...

    2024