Pith. sign in

REVIEW 3 major objections 5 minor 38 references

Imitation planners that ace standard driving tests collapse under new cities and actuation noise; the tested RL planner does not.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-10 16:51 UTC pith:DTDB56MQ

load-bearing objection Useful dual-track closed-loop stress suite (DeepPlan conversion + AWGN/OU noise) with clear IL fragility numbers; the threshold re-calibration is a real soft spot but does not erase the contribution. the 3 major comments →

arxiv 2607.07844 v1 pith:DTDB56MQ submitted 2026-07-08 cs.RO cs.AIcs.LG

Shift & Drift: A Zero-Shot Benchmark for Generalizable and Robust Autonomous Driving Motion Planning

Alessandro Canevaro , Hang Yu , Julian Schmidt , Peizheng Li , Silvan Lindner , Wilhelm Stork , Georg Martius , Julian Jordan This is my paper
classification cs.RO cs.AIcs.LG
keywords autonomous drivingmotion planningzero-shot generalizationdistribution shiftimitation learningreinforcement learningclosed-loop simulationactuation noise
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Closed-loop motion planners for autonomous driving look strong when tested on the same geographic and traffic distributions they were trained on, but that leaves two practical questions unanswered: do they still work on novel city layouts with dense pedestrians and cyclists, and can they recover when small execution errors accumulate? This paper introduces Shift & Drift, a dual-track benchmark that answers both. One track converts high-precision aerial recordings from four German cities and San Francisco into the standard nuPlan simulator so models trained only on North American and Singaporean data can be evaluated zero-shot on 1,182 new scenarios. The second track injects white Gaussian jitter and temporally correlated Ornstein–Uhlenbeck noise into the vehicle’s acceleration and steering commands. Across rule-based, imitation-learning, and reinforcement-learning planners, the paper shows that high imitation scores on the familiar benchmark hide large failures under semantic shift and under persistent drift, while the evaluated RL planner degrades far more gracefully. The result is a concrete empirical trade-off between how faithfully a planner copies expert driving and how resilient it remains once the world or the actuators leave the training distribution.

Core claim

Imitation-learning planners that score highly on the standard in-distribution nuPlan validation set (closed-loop scores roughly 84–93) suffer large zero-shot drops on the new DeepPlan scenarios (overall scores roughly 34–37 in non-reactive mode, with one model falling 76 percent in Munich) and degrade further under temporally correlated actuation noise, whereas the evaluated reinforcement-learning planner retains substantially higher safety and progress (DeepPlan scores roughly 70–73, at most an 8 percent decay under high noise). The paper therefore claims an empirical trade-off between imitation fidelity and closed-loop resilience under semantic and state-distribution shift.

What carries the argument

Shift & Drift dual-track benchmark: (1) DeepPlan, a conversion of the aerial DeepScenario Open 3D dataset into 1,182 nuPlan-compatible scenarios for zero-shot semantic-shift evaluation, and (2) controlled injection of AWGN and Ornstein–Uhlenbeck noise into ego acceleration and steering-rate commands to measure recovery from state-distribution drift.

Load-bearing premise

The hand-relaxed safety and comfort thresholds used to score the new European pedestrian-heavy scenarios correctly measure planning quality rather than quietly masking failures or rearranging the ranking among the tested planners.

What would settle it

Re-score all five planners on the full DeepPlan set with the original unadjusted nuPlan thresholds and check whether the large gap between the RL planner and the imitation planners shrinks, vanishes, or reverses.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper introduces Shift & Drift, a dual-track closed-loop benchmark for object-level autonomous-driving motion planners. The Semantic Shift Track converts the aerial DSC3D dataset into a nuPlan-compatible suite (DeepPlan: 1,182 scenarios across four German cities and San Francisco) to enable zero-shot evaluation of planners trained only on nuPlan’s North-American/Singaporean data, stressing novel topologies and dense pedestrian–cyclist interactions. The State-Distribution Drift Track injects AWGN and Ornstein–Uhlenbeck actuation noise into the ego command vector to quantify recovery from compounding execution error. Five SOTA planners (PDM-Closed, PlanTF, PLUTO, Diffusion Planner, CaRL) are evaluated under both tracks. The central empirical claim is that high-ID imitation-learning methods suffer large CLS drops under semantic shift (e.g., Diffusion Planner Val14 89.87 → Munich 21.50) and further collapse under temporally correlated OU noise, whereas the evaluated RL planner CaRL degrades more gracefully (DeepPlan CLS ~70–73; ≤8 % decay under high noise), revealing a trade-off between imitation fidelity and closed-loop resilience.

Significance. If the reported rankings hold under scrutiny of the scoring protocol, the work supplies a timely, large-scale, publicly released stress test that goes beyond i.i.d. nuPlan splits and handcrafted long-tail suites such as interPlan. The conversion pipeline that turns occlusion-free aerial trajectories into standardized nuPlan logs and maps, the dual-axis design (semantic + state-distribution), and the multi-seed noise results are concrete community assets. Explicit release of code and data further strengthens the contribution. The empirical contrast between IL fragility and RL resilience is actionable for paradigm selection and hybrid design, even if limited to the five evaluated agents.

major comments (3)
  1. [§III-C, Table II] §III-C and Table II: The DeepPlan CLS numbers that underwrite the central zero-shot claim (Table III) and the combined-stress claim (Table IV) rest on substantially relaxed hard-constraint thresholds (drivable-area violation 0.3→3.0 m, min TTC 1.0→0.5 s, relaxed jerk/acceleration bounds). The sole justification is that log-playback CLS falls from ~90 to ~65 under defaults. Because CLS zeros an entire scenario on any hard safety violation, a 10× relaxation of the curb bound and a 50 % relaxation of TTC can convert many near-miss or curb-clip events into non-zero scores. No planner-level CLS (or NCR/PER/TTC) under the original nuPlan thresholds, nor any continuous sensitivity sweep, is reported. Without that ablation it is impossible to separate genuine semantic-shift fragility from an artifact of re-calibration that may differentially favor or penalize the five methods.
  2. [§III-A.4] §III-A.4 (Post-Processing): Scenarios are discarded if the log-playback planner itself produces at-fault collisions, drivable-area violations, or insufficient progress, followed by manual route correction. The paper does not report the fraction discarded, the geographic or interaction-density distribution of the discarded set, or any comparison of planner rankings on the unfiltered versus filtered suite. Because the filter is defined by the same safety/progress criteria that later appear in the score, it risks systematically removing precisely the long-tail topologies that the Semantic Shift Track claims to stress, thereby inflating absolute CLS values and potentially compressing relative gaps.
  3. [§IV-B, Abstract, §V] §IV-B / Table III and abstract: The paradigm-level language (“IL methods \ldots exhibit significant failures”, “the evaluated reinforcement-learning-based planner demonstrates more graceful degradation”) is supported by only a single RL agent (CaRL). While the manuscript is careful in places to say “the evaluated” RL planner, the abstract and conclusions still frame an empirical trade-off between imitation and reinforcement learning as paradigms. With a single RL data point the claim remains suggestive rather than comparative; either additional closed-loop RL baselines or a clearer restriction of language to the specific agents is required for the load-bearing generalization.
minor comments (5)
  1. [Fig. 1] Fig. 1 caption and panel label contain the typo “State-Dristribution Drift Track” (missing ‘t’).
  2. [§III-B.2] §III-B.2: The choice of OU mean-reversion rate θ = 2.0 s⁻¹ (correlation time 0.5 s) is stated without reference to measured actuator or vehicle dynamics; a short justification or sensitivity note would help readers assess realism.
  3. [Table I] Table I lists “Scenario Count 1118” for nuPlan while the text elsewhere uses Val14; a consistent reference (or explicit note that 1118 is the full Val14 split) would avoid confusion.
  4. [Fig. 3] Fig. 3 qualitative panels are informative but the bottom-row captions are dense; labeling each sub-panel (a–e) and referring to them in the text would improve readability.
  5. [§III-B] Eqs. (1)–(4) use both continuous-time SDE and discrete Euler–Maruyama forms; explicitly stating the simulation Δt used for integration would make the noise process fully reproducible from the text alone.

Circularity Check

0 steps flagged

Empirical benchmark paper with no derivation chain; scores are external simulator outputs, not rearrangements of fitted inputs or self-defined quantities.

full rationale

Shift & Drift introduces a dual-track evaluation suite (DeepPlan conversion of DSC3D aerial logs into nuPlan format + stochastic AWGN/OU actuation noise) and reports closed-loop scores of five existing planners (PDM-Closed, PlanTF, PLUTO, Diffusion Planner, CaRL) under those conditions. There is no algebraic derivation, no parameter fitted to a subset then re-labeled as a prediction, no uniqueness theorem, and no ansatz smuggled via self-citation. The composite CLS, NCR, PER, etc., are produced by the independent nuPlan simulator on held-out converted scenarios; the same metric equations are applied uniformly. The observation that CaRL’s reward (progress/safety/comfort) is aligned with those metrics is an explanatory remark, not a circular reduction of the reported numbers. Threshold re-calibration in Table II is a modeling choice whose effect on rankings is unablated, but that is an assumption/correctness issue, not circularity. No load-bearing step reduces by construction to the paper’s own inputs. Score 0 is therefore the correct, non-manufactured finding.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 2 invented entities

The central claims rest on standard closed-loop AD evaluation practice plus several modeling choices that define the stress tests. Free parameters are the hand-chosen noise intensities, OU mean-reversion rate, episode windowing, and the relaxed DeepPlan metric thresholds. Domain axioms include the validity of the aerial-to-nuPlan conversion and the use of AWGN/OU as proxies for real actuation error. Invented entities are the DeepPlan suite and the Shift & Drift dual-track protocol themselves; both are operationalized by released conversion code and scenario counts rather than postulated unobservables.

free parameters (4)
  • AWGN/OU noise intensity levels (σ_δ̇, σ_a)
    Three hand-chosen intensity triples (Low/Mid/High: σ_δ̇ ∈ {0.1,0.2,0.3} rad/s, σ_a ∈ {0.5,1.0,1.5} m/s²) define the entire State-Distribution Drift Track; results and ranking under drift depend on these values.
  • OU mean-reversion rate θ
    θ=2.0 s⁻¹ (correlation time 0.5 s) is chosen by hand to produce multi-frame drift; different θ would change how hard recovery is for open-loop-style IL policies.
  • DeepPlan metric threshold overrides (Table II)
    Driving-direction, drivable-area, TTC, and comfort thresholds were relaxed after log-playback scored poorly under defaults; comparative CLS on DeepPlan depends on these adjusted cutoffs.
  • Episode windowing and ego-selection criteria
    18 s windows with 6 s overlap, vehicle-only ego, peak speed >1 m/s, no bounding-box overlap, and frame-60 as t=0 determine which of the raw DSC3D recordings become the 1,182 evaluation scenarios.
axioms (4)
  • domain assumption nuPlan closed-loop score (weighted safety/progress/comfort with hard zeroing on critical violations) is a valid primary measure of planner quality under both tracks.
    All quantitative claims in §IV are reported as CLS and component metrics under this protocol; alternative metrics could reorder methods.
  • domain assumption Converting occlusion-free aerial DSC3D trajectories and OpenDRIVE maps into nuPlan logs/maps preserves physically valid multi-agent interaction context for closed-loop planning.
    Load-bearing for the Semantic Shift Track (§III-A); conversion artifacts or map misalignment would invalidate zero-shot scores.
  • domain assumption AWGN and independent OU processes on acceleration and steering-rate commands are adequate standardized proxies for real execution/actuation error.
    Stated explicitly in §III-B as stress-test proxies rather than hardware replicas; the robustness ranking is only as meaningful as this modeling choice.
  • ad hoc to paper Post-conversion automated filtering (discard log-playback collisions, drivable-area violations, insufficient progress) plus manual route correction yields a fair evaluation set rather than a biased easy subset.
    §III-A post-processing; filtering on log-playback success can remove the hardest residual cases and affect absolute difficulty.
invented entities (2)
  • DeepPlan evaluation suite independent evidence
    purpose: Provide 1,182 nuPlan-compatible scenarios from DSC3D aerial data across Munich, Stuttgart, Sindelfingen, Berlin, and San Francisco for zero-shot semantic-shift testing.
    Constructed by the paper’s conversion pipeline; independent evidence is the released scenario count and geographic split, but validity as a planning benchmark depends on the conversion axioms above.
  • Shift & Drift dual-track protocol independent evidence
    purpose: Jointly stress-test planners on semantic (cross-dataset) and state-distribution (actuation noise) axes under one framework.
    The dual-track design is the paper’s methodological contribution; it is operationalized by the released code and noise definitions rather than a new physical entity.

pith-pipeline@v1.1.0-grok45 · 18455 in / 3833 out tokens · 45950 ms · 2026-07-10T16:51:36.095045+00:00 · methodology

0 comments
read the original abstract

While closed-loop motion planners trained on large-scale, object-level datasets, e.g., nuPlan, demonstrate strong in-distribution (ID) performance, their generalization to novel urban topologies and recovery mechanisms following execution perturbations remain under-explored. To address this, we present Shift & Drift, a novel dual-track benchmark designed to rigorously stress-test motion planners across two critical axes of distribution shift: (1) The Semantic Shift Track leverages a novel conversion pipeline that transforms the aerial, DeepScenario Open 3D dataset into the nuPlan simulation framework. This enables zero-shot evaluation of planners trained on North American and Singaporean data against 1,182 scenarios spanning four German cities and the US city of San Francisco featuring dense pedestrian-cyclist interactions. (2) The State-Distribution Drift Track injects stochastic perturbations into the ego vehicle's dynamics to quantify robustness against compounding execution errors. Based on this, we systematically evaluate the failure modes of diverse planning paradigms under semantic and state-distribution shifts. While imitation learning methods achieve high scores in ID benchmarks, they exhibit significant failures under semantic shift, particularly in pedestrian-dense environments, and suffer from persistent drift when subjected to temporally correlated actuation noise. In contrast, the evaluated reinforcement-learning-based planner demonstrates more graceful degradation, maintaining higher safety and progress metrics across both tracks. Our findings reveal an empirical trade-off between imitation fidelity and closed-loop resilience, providing the community with a rigorous benchmark to evaluate progress toward reliable deployment.

Figures

Figures reproduced from arXiv: 2607.07844 by Alessandro Canevaro, Georg Martius, Hang Yu, Julian Jordan, Julian Schmidt, Peizheng Li, Silvan Lindner, Wilhelm Stork.

Figure 1
Figure 1. Figure 1: Overview of our dual-track benchmark: Shift & Drift. Track 1: Semantic Shift Track converts aerial DSC3D data [6] into nuPlan [7] standard for zero-shot evaluation. Track 2: State-Distribution Drift Track injects different types of noise into actuation. [4], models that excel in known urban layouts often suffer from geographic overfitting, where the policy implicitly memorizes map-specific features or loca… view at source ↗
Figure 2
Figure 2. Figure 2: Average number of vehicles, pedestrian and bicycles [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison of planner trajectories across DeepPlan scenarios. We show the ground truth (Log) alongside PlanTF, PDM-Closed, PLUTO, Diffusion Planner, and CaRL. Top row: During a right turn with high pedestrian density, Diffusion Planner and PDM-Closed proceed aggressively, resulting in collisions, while PLUTO and CaRL successfully yield. PlanTF fails to commit to forward progress. Middle row: In… view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative analysis of trajectory stability under actuation perturbations. From left to right: (1) Reference trajectories for Log, CaRL, and Diffusion Planner under nominal conditions; (2) CaRL under nominal conditions (solid) vs. high-intensity OU drift (dashed); (3) Diffusion Planner under nominal conditions (solid) vs. high-intensity AWGN jitter (dashed); (4) Diffusion Planner under nominal conditions … view at source ↗

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

38 extracted references · 38 canonical work pages · 5 internal anchors

  1. [1]

    A survey of deep learning techniques for autonomous driving,

    S. Grigorescu, B. Trasnea, T. Cocias, and G. Macesanu, “A survey of deep learning techniques for autonomous driving,”Journal of Field Robotics, 2019

    work page 2019

  2. [2]

    Self-driving cars: A survey,

    C. Badue, R. Guidolini, R. V . Carneiro, P. Azevedo, V . B. Cardoso, A. Forechi, L. Jesus, R. Berriel, T. M. Paix ˜ao, F. Mutz, L. de Paula Veronese, T. Oliveira-Santos, and A. F. De Souza, “Self-driving cars: A survey,”Expert Systems with Applications, 2021

    work page 2021

  3. [3]

    A survey of deep rl and il for autonomous driving policy learning,

    Z. Zhu and H. Zhao, “A survey of deep rl and il for autonomous driving policy learning,”IEEE T-ITS, 2022

    work page 2022

  4. [4]

    Exploring the limitations of behavior cloning for autonomous driving,

    F. Codevilla, E. Santana, A. Lopez, and A. Gaidon, “Exploring the limitations of behavior cloning for autonomous driving,” inICCV, 2019

    work page 2019

  5. [5]

    Causal confusion in imitation learning,

    P. de Haan, D. Jayaraman, and S. Levine, “Causal confusion in imitation learning,”NeurIPS, 2019

    work page 2019

  6. [6]

    Highly accurate and diverse traffic data: The deepscenario open 3d dataset,

    O. Dhaouadi, J. Meier, L. Wahl, J. Kaiser, L. Scalerandi, N. Wan- delburg, Z. Zhou, N. Berinpanathan, H. Banzhaf, and D. Cremers, “Highly accurate and diverse traffic data: The deepscenario open 3d dataset,” inIEEE IV, 2025

    work page 2025

  7. [7]

    Nuplan: A closed-loop ml-based planning benchmark for autonomous vehicles,

    H. Caesar, J. Kabzan, K. Tan, and et al., “Nuplan: A closed-loop ml-based planning benchmark for autonomous vehicles,” inCVPR ADP3 Workshop, 2021

    work page 2021

  8. [8]

    Wod-e2e: Waymo open dataset for end-to-end driving in challenging long-tail scenarios,

    R. Xu, H. Lin, W. Jeon, H. Feng, Y . Zou, L. Sun, J. Gorman, E. Tolstaya, S. Tang, B. White, B. Sapp, M. Tan, J.-J. Hwang, and D. Anguelov, “Wod-e2e: Waymo open dataset for end-to-end driving in challenging long-tail scenarios,”arXiv preprint arxiv:2510.26125, 2025

    work page arXiv 2025

  9. [9]

    Traphic: Tra- jectory prediction in dense and heterogeneous traffic using weighted interactions,

    R. Chandra, U. Bhattacharya, A. Bera, and D. Manocha, “Traphic: Tra- jectory prediction in dense and heterogeneous traffic using weighted interactions,” inCVPR, 2019

    work page 2019

  10. [10]

    A survey on autonomous driving datasets: Statistics, annotation quality, and a future outlook,

    M. Liu, E. Yurtsever, J. Fossaert, X. Zhou, W. Zimmer, Y . Cui, B. L. Zagar, and A. C. Knoll, “A survey on autonomous driving datasets: Statistics, annotation quality, and a future outlook,”IEEE T-IV, 2024

    work page 2024

  11. [11]

    Alvinn: An autonomous land vehicle in a neural network,

    D. A. Pomerleau, “Alvinn: An autonomous land vehicle in a neural network,” inNeurIPS, 1988

    work page 1988

  12. [12]

    A reduction of imitation learning and structured prediction to no-regret online learning,

    S. Ross, G. Gordon, and D. Bagnell, “A reduction of imitation learning and structured prediction to no-regret online learning,” inAISTATS, 2011

    work page 2011

  13. [13]

    CARLA: An open urban driving simulator,

    A. Dosovitskiy, G. Ros, F. Codevilla, A. Lopez, and V . Koltun, “CARLA: An open urban driving simulator,” inCoRL, 2017, pp. 1–16

    work page 2017

  14. [14]

    On the theory of the brownian motion,

    G. E. Uhlenbeck and L. S. Ornstein, “On the theory of the brownian motion,”Phys. Rev., 1930

    work page 1930

  15. [15]

    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,” in CoRL, 2023

    work page 2023

  16. [16]

    Rethinking Imitation-based Planner for Autonomous Driving

    J. Cheng, Y . Chen, X. Mei, B. Yang, B. Li, and M. Liu, “Rethink- ing imitation-based planner for autonomous driving,”arXiv preprint arXiv:2309.10443, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  17. [17]

    PLUTO: Pushing the Limit of Imitation Learning-based Planning for Autonomous Driving

    J. Cheng, Y . Chen, and Q. Chen, “Pluto: Pushing the limit of imita- tion learning-based planning for autonomous driving,”arXiv preprint arXiv:2404.14327, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  18. [18]

    Diffusion-based planning for autonomous driving with flexible guidance,

    Y . Zheng, R. Liang, K. Zheng, J. Zheng, L. Mao, J. Li, W. Gu, R. Ai, S. E. Li, X. Zhan, and J. Liu, “Diffusion-based planning for autonomous driving with flexible guidance,” inICLR, 2025

    work page 2025

  19. [19]

    Carl: Learning scalable planning policies with simple rewards,

    B. Jaeger, D. Dauner, J. Beißwenger, S. Gerstenecker, K. Chitta, and A. Geiger, “Carl: Learning scalable planning policies with simple rewards,” inCoRL, 2025

    work page 2025

  20. [20]

    Are we ready for autonomous driving? the kitti vision benchmark suite,

    A. Geiger, P. Lenz, and R. Urtasun, “Are we ready for autonomous driving? the kitti vision benchmark suite,” inCVPR, 2012

    work page 2012

  21. [21]

    Argoverse 2: Next generation datasets for self-driving perception and forecasting,

    B. Wilson, W. Qi, T. Agarwal, J. Lambert, J. Singh, S. Khandelwal, B. Pan, R. Kumar, A. Hartnett, J. K. Pontes, D. Ramanan, P. Carr, and J. Hays, “Argoverse 2: Next generation datasets for self-driving perception and forecasting,” inNeurIPS Datasets and Benchmarks, 2021

    work page 2021

  22. [22]

    Scenarionet: Open-source platform for large-scale traffic scenario simulation and modeling,

    Q. Li, Z. Peng, L. Feng, Z. Liu, C. Duan, W. Mo, and B. Zhou, “Scenarionet: Open-source platform for large-scale traffic scenario simulation and modeling,”NeurIPS, 2023

    work page 2023

  23. [23]

    opendd: A large-scale roundabout drone dataset,

    A. Breuer, J.-A. Term ¨ohlen, S. Homoceanu, and T. Fingscheidt, “opendd: A large-scale roundabout drone dataset,” inITSC, 2020

    work page 2020

  24. [24]

    Citysim: A drone-based vehicle trajectory dataset for safety-oriented research and digital twins,

    O. Zheng, M. Abdel-Aty, L. Yue, A. Abdelraouf, Z. Wang, and N. Mahmoud, “Citysim: A drone-based vehicle trajectory dataset for safety-oriented research and digital twins,”TRR, 2024

    work page 2024

  25. [25]

    The highd dataset: A drone dataset of naturalistic vehicle trajectories on german highways for validation of highly automated driving systems,

    R. Krajewski, J. Bock, L. Kloeker, and L. Eckstein, “The highd dataset: A drone dataset of naturalistic vehicle trajectories on german highways for validation of highly automated driving systems,” inITSC, 2018

    work page 2018

  26. [26]

    INTERACTION Dataset: An INTERnational, Adversarial and Cooperative moTION Dataset in Interactive Driving Scenarios with Semantic Maps

    W. Zhan, L. Sun, D. Wang, H. Shi, A. Clausse, M. Naumann, J. K ¨ummerle, H. K ¨onigshof, C. Stiller, A. de La Fortelle, and M. Tomizuka, “INTERACTION dataset: An international, adversarial and cooperative motion dataset in interactive driving scenarios with semantic maps,”CoRR, vol. abs/1910.03088, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1910

  27. [27]

    Deepurban: Interaction-aware trajectory pre- diction and planning for automated driving by aerial imagery,

    C. Selzer and F. Flohr, “Deepurban: Interaction-aware trajectory pre- diction and planning for automated driving by aerial imagery,” inITSC, 2024

    work page 2024

  28. [28]

    Proximal Policy Optimization Algorithms

    J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Proximal policy optimization algorithms.”CoRR, vol. abs/1707.06347, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  29. [29]

    Congested traffic states in empirical observations and microscopic simulations,

    M. Treiber, A. Hennecke, and D. Helbing, “Congested traffic states in empirical observations and microscopic simulations,”Physical Review E, 2000

    work page 2000

  30. [30]

    General lane-changing model mobil for car-following models,

    A. Kesting, M. Treiber, and D. Helbing, “General lane-changing model mobil for car-following models,”TRR, 2007

    work page 2007

  31. [31]

    Bench2drive: Towards multi-ability benchmarking of closed-loop end-to-end autonomous driving,

    X. Jia, Z. Yang, Q. Li, Z. Zhang, and J. Yan, “Bench2drive: Towards multi-ability benchmarking of closed-loop end-to-end autonomous driving,” inNeurIPS Datasets and Benchmarks Track, 2024

    work page 2024

  32. [32]

    Pseudo-simulation for autonomous driving,

    W. Cao, M. Hallgarten, T. Li, D. Dauner, X. Gu, C. Wang, Y . Miron, M. Aiello, H. Li, I. Gilitschenski, B. Ivanovic, M. Pavone, A. Geiger, and K. Chitta, “Pseudo-simulation for autonomous driving,” inCoRL, 2025

    work page 2025

  33. [33]

    nuplan-r: A closed-loop planning benchmark for autonomous driving via reactive multi-agent simulation,

    M. Peng, R. Yao, X. Guo, and J. Ma, “nuplan-r: A closed-loop planning benchmark for autonomous driving via reactive multi-agent simulation,”arXiv preprint arxiv:2511.10403, 2025

    work page arXiv 2025

  34. [34]

    When planners meet reality: How learned, reactive traffic agents shift nuplan benchmarks,

    S. Hagedorn, L. Donkov, A. Distelzweig, and A. P. Condurache, “When planners meet reality: How learned, reactive traffic agents shift nuplan benchmarks,”arXiv preprint arxiv:2510.14677, 2025

    work page arXiv 2025

  35. [35]

    Can Vehicle Motion Planning Generalize to Realistic Long-tail Scenarios?

    M. Hallgarten, J. Zapata, M. Stoll, K. Renz, and A. Zell, “Can vehicle motion planning generalize to realistic long-tail scenarios?”arXiv preprint arXiv:2404.07569, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  36. [36]

    Safebench: A benchmarking platform for safety evaluation of autonomous vehicles,

    C. Xu, W. Ding, W. Lyu, Z. Liu, S. Wang, Y . He, H. Hu, D. Zhao, and B. Li, “Safebench: A benchmarking platform for safety evaluation of autonomous vehicles,” inNeurIPS Datasets and Benchmarks, 2022

    work page 2022

  37. [37]

    Advsim: Generating safety-critical scenarios for self- driving vehicles,

    J. Wang, A. Pun, J. Tu, S. Manivasagam, A. Sadat, S. Casas, M. Ren, and R. Urtasun, “Advsim: Generating safety-critical scenarios for self- driving vehicles,” inCVPR, 2021

    work page 2021

  38. [38]

    Opendrive: Open dynamic road information for vehicle environment,

    OpenDrive, “Opendrive: Open dynamic road information for vehicle environment,” 2000. [Online]. Available: https://www.opendrive.com

    work page 2000