Pith. sign in

REVIEW 3 major objections 4 minor 53 references

A planner that reuses nearby drivers as teachers, refreshes language scene cues only when traffic gets hard, and trains through a differentiable trajectory optimizer tops the nuPlan Hard20 closed-loop scores.

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

2026-07-14 11:43 UTC pith:F5DDFL7K

load-bearing objection Solid hybrid planner with real Hard20 gains; most lift is from residual optimization and agent-centric labels, not the hand-tuned LLM scheduler the abstract foregrounds. the 3 major comments →

arxiv 2607.10438 v1 pith:F5DDFL7K submitted 2026-07-11 cs.RO cs.NI

Large Language Model Enhanced Differentiable Trajectory Planning for IoT-Enabled Autonomous Driving

Shihao Zhang , Jing Yang , Ziyu Song , Zheng Lin , Sunil Prajapat , Zhaochen Xia , Hemant Ghayvat , Haitao Ding
show 2 more authors
Lip Yee Por Ashok Kumar Das
This is my paper
classification cs.RO cs.NI
keywords imitation learningdifferentiable optimizationlarge language modeltrajectory planningconnected autonomous drivingdata augmentationclosed-loop evaluationIoT intelligent transportation
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.

Imitation-learning planners for city driving often miss rare multi-agent interactions, ignore high-level scene meaning under real-time limits, and train without feedback from the constraints that will later refine their trajectories. This paper claims those three gaps can be closed together without new raw data or constant large-model calls. Surrounding vehicles that already interact in complex ways are recentered as extra planning subjects so one log episode yields more hard training cases. A large language model supplies scene-and-instruction features, but a simple complexity score decides how often to refresh them so latency stays controlled. A residual optimizer then reshapes the chosen trajectory under speed, comfort, safety, and bicycle-model costs, and because the solver is differentiable those costs train the upstream network. On the hard closed-loop splits the combined system records the highest overall scores among the compared methods, and a live CARLA-ROS loop confirms it can run online.

Core claim

The authors establish that an imitation planner trained with surrounding-agent-centric trajectory reuse, complexity-aware asynchronous language-model semantic features, and residual differentiable optimization produces safer, more feasible closed-loop trajectories than strong baselines on nuPlan Hard20, reaching overall scores of 83.63 (nonreactive) and 78.29 (reactive) while remaining deployable in real time.

What carries the argument

The residual-based differentiable optimizer: a Levenberg–Marquardt solver that refines the highest-confidence ego trajectory under soft penalties for speed, reference-line consistency, comfort, selective safety buffers, and bicycle kinematics, and that back-propagates residual gradients into the upstream planner and cost-weight network so generation and refinement are learned jointly.

Load-bearing premise

That a hand-tuned rule scoring traffic density, conflicts, time-to-collision, intersections, navigation changes, and short-term variation is a good enough proxy for when language-model semantics must be refreshed, so reusing the last feature for many frames still keeps planning quality under a real-time budget.

What would settle it

On the same Hard20 reactive split, replace the fixed thresholds and reuse lengths with a learned or oracle refresh policy (or force full synchronous language-model updates) and check whether overall score or safety metrics rise enough to erase the claimed advantage while runtime stays inside the online envelope.

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

If this is right

  • Existing logged multi-agent episodes can be turned into denser long-tail planning supervision simply by re-indexing surrounding vehicles as ego, without new collection.
  • High-level language semantics can be injected into real-time planners if a lightweight complexity gate keeps invocation frequency low in simple traffic.
  • Training through a differentiable residual optimizer aligns the network’s proposals with the same constraints used at execution time, reducing the usual train–refine mismatch.
  • Closed-loop Hard20 leadership plus a working CARLA-ROS loop implies the stack is already a practical candidate for software-in-the-loop IoT vehicle testing.

Where Pith is reading between the lines

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

  • The same agent-reindexing idea could be applied to other multi-agent logs (pedestrians, cyclists, or mixed fleets) wherever rare interactions dominate failure modes.
  • A learned complexity policy trained to maximize planning score per unit of language-model latency would test how much headroom remains in the hand-tuned scheduler.
  • Because gradients flow through the residual solver, the same pattern could be used to co-train prediction and planning under richer multi-directional safety costs without changing the online inference graph.

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 / 4 minor

Summary. The paper proposes an imitation-learning trajectory planner for IoT-enabled autonomous driving that combines three components: (i) surrounding-agent-centric data augmentation that reindexes filtered non-ego trajectories as additional planning supervision, (ii) a complexity-aware asynchronous LLM module that injects scene-associated semantic features with adaptive reuse lengths, and (iii) a residual-based differentiable nonlinear optimizer (Theseus LM) that refines the selected ego trajectory and backpropagates optimization gradients to the upstream planner. On nuPlan closed-loop Hard20, the full system reports best overall scores of 83.63 (nonreactive) and 78.29 (reactive) among the listed baselines, with component ablations, scheduler/residual studies, qualitative closed-loop comparisons, and CARLA-ROS SIL tests supporting online 5 Hz deployment.

Significance. If the reported closed-loop gains hold under fair comparison, the work is a solid systems contribution at the intersection of IL planning, LLM semantic guidance, and differentiable trajectory refinement. Strengths include full nuPlan Hard20 closed-loop tables (II–III), stepwise component ablations (IV), augmentation design (VI), LLM schedule and scheduler-factor studies (V, VII), residual-category ablations (VIII), and a separate CARLA-ROS stack that demonstrates real-time closed-loop execution rather than offline metrics alone. The residual optimizer with gradient flow through LM iterations and the agent-centric reuse of logged trajectories are practically useful ideas even if the LLM scheduler is secondary. The IoT/real-time framing is relevant to IEEE IoT Journal, but the significance of the complexity-aware LLM design specifically depends on cleaner isolation from the larger optimizer/augmentation lift.

major comments (3)
  1. [§V-A, Tables II–V] The strongest claim attributes best Hard20 scores under real-time budgets to the full stack, especially complexity-aware AsyncLLM. Table IV shows most reactive gain already at M2 (aug + Diff. Opt.: 75.19 vs M0 60.43); full M3 adds only +3.1 to 78.29. Table V shows complexity-aware AsyncLLM (78.29 / 172 ms) is below fixed-3-frame (79.61 / 237 ms) and synchronous (80.46 / 477 ms). Without a fixed-schedule, matched-budget re-run of the full model that still ranks above PLUTO (76.88 reactive), the IoT/real-time framing over-attributes the headline ranking to the hand-tuned scheduler rather than residual optimization and augmentation.
  2. [§III-C, Eq. (7), Tables I, V, VII] Eq. (7) defines C_t with equal α_i and fixed thresholds τ_l=0.35, τ_h=0.65 and K∈{3,9,29} (Table I). Table VII shows threshold/factor changes move score by a few points, but the scheduler remains a free-parameter rule without a learned policy or validation that these proxies are necessary for the claimed ranking under a fixed compute budget. Please either (a) report matched-budget fixed-K full-model results, or (b) soften claims that complexity-aware scheduling is what enables best overall real-time performance.
  3. [§III-D, Eq. (15), Tables II–III, VIII] Safety residual (Eq. 15) uses a selective local corridor with a fixed +5.0 m buffer and explicitly defers rear-end risk to upstream prediction/closed-loop feedback. Given that reactive Hard20 emphasizes interaction, please quantify how often rear-end or multi-agent conflicts fall outside the selected corridor, and whether residual design choices (not only LLM scheduling) drive Coll./TTC differences vs PLUTO in Tables II–III.
minor comments (4)
  1. [Fig. 1, §I, §III-B] Fig. 1 scenario taxonomy (SST/CIST/CST/RST) is used to motivate long-tail imbalance, but the manuscript does not state how these labels are assigned or whether augmentation preferentially samples RST/CST.
  2. [§V-B, Fig. 5] CARLA-ROS reports 80% success, <5% collision, <30 s traversal at 5 Hz, but scenario count, route set, and comparison against a non-LLM or non-optimizer baseline on the same SIL stack are not specified.
  3. [§III-E, §IV-B, Table I] Several loss weights and LM solver settings are listed as free parameters; a short sensitivity note (beyond residual-category removal) would help reproducibility.
  4. [Abstract, §III-C] Minor presentation: spacing/hyphenation issues in the abstract (“sur rounding”, “asyn chronous”) and occasional notation overload (K^sem_t vs Δt) should be cleaned.

Circularity Check

0 steps flagged

Empirical systems paper with external closed-loop benchmarks; no derivation reduces to its inputs by construction.

full rationale

The paper is an IL-based planning systems contribution (agent-centric data augmentation, complexity-aware async LLM features, residual differentiable refinement). Its load-bearing claims are empirical closed-loop scores on the official nuPlan Hard20 protocol and CARLA-ROS SIL runs, not first-principles predictions. Training losses (L_plan, L_cls, L_pred, L_cost, L_align) supervise optimized trajectories and predictions against expert/ground-truth labels from the dataset; residual forms and physical thresholds are hand-specified priors (nuPlan criteria, bicycle model), while weights are learned—standard supervised learning, not a fitted parameter renamed as a prediction of the same quantity. The complexity score C_t and thresholds (τ_l, τ_h, K) are rule-based scheduling hyperparameters for LLM reuse; they do not define the reported Score/Coll./TTC metrics. Citations to AsyncDriver, PLUTO, Theseus, and related work supply architectural priors from non-overlapping authors and are not uniqueness theorems that force the result. No self-definitional loop, fitted-input-as-prediction, or self-citation chain makes the headline ranking true by construction. Ablation incompleteness (scheduler vs. residual optimizer attribution) is a support/correctness concern, not circularity.

Axiom & Free-Parameter Ledger

8 free parameters · 5 axioms · 3 invented entities

The central SOTA-style claim rests on standard IL/closed-loop AD assumptions plus many hand-chosen scheduler and residual constants, not on new physical entities. Load-bearing modeling choices include treating reindexed surrounding trajectories as valid ego supervision, using soft residual penalties (with a selective safety corridor) as a stand-in for hard constraints, freezing a LoRA-adapted LLM as an async semantic feature source, and scheduling LLM updates with a fixed equal-weight complexity formula. Free parameters dominate the online efficiency–quality tradeoff; invented ‘entities’ are engineering modules validated only inside this paper’s sims.

free parameters (8)
  • Complexity thresholds τ_l, τ_h
    Set to 0.35 and 0.65 from validation statistics and latency budget; directly control when LLM features refresh and thus the reported runtime/score tradeoff.
  • Semantic reuse lengths K_min, K_mid, K_max
    Hand-set to 3, 9, 29 frames; define the asynchronous schedule that underpins the real-time claim.
  • Complexity factor weights α_1…α_6
    Default equal weights with no learned policy; define the scalar C_t that gates LLM calls.
  • Target-vehicle screening radius
    Default 50 m chosen after radius sweep; controls which surrounding trajectories become extra labels.
  • Safety residual buffer (+5.0 m) and corridor selection rules
    Fixed conservative urban buffer and selective nearest/highest-risk target in Frenet corridor; shapes collision/TTC behavior without being learned from first principles.
  • Comfort residual limits a_max, a_min, j_max
    Taken from nuPlan criteria (2.40, −4.05 m/s², 4.13 m/s³) and used as soft penalties inside the optimizer.
  • Loss weights λ_align, λ_pred, λ_imi, λ_cost
    Main task weights set to 1.0 and optimization auxiliary to 1e−2; scale how strongly residual costs train the upstream planner.
  • Levenberg–Marquardt solver hyperparameters
    Max 5 iterations, step 0.2, abs tol 1e−3, initial damping 1e−1, stability 1e−6; affect both refined trajectories and backpropagated gradients.
axioms (5)
  • domain assumption Expert and reindexed surrounding-agent trajectories are valid planning supervision for imitation learning after local-frame reindexing and simple finite-difference state completion.
    §III-B treats filtered non-ego logs as additional ego labels; if non-ego behavior is non-expert or incomplete, augmentation injects biased supervision.
  • domain assumption Soft residual penalties on efficiency, comfort, selective local safety, and discrete bicycle kinematics adequately approximate executable closed-loop constraints when optimized with a few LM steps.
    §III-D formulates u* via weighted residual least squares; hard collision-free guarantees and full multi-agent rear-end coverage are not claimed inside the solver.
  • domain assumption A frozen general-purpose LLM, after lightweight alignment losses on nuPlan-derived instruction tasks, yields useful scene-associated instruction features for trajectory decoding.
    §III-C and §III-E freeze the LLM backbone after L_align; planning quality depends on that transfer.
  • ad hoc to paper Scene complexity for LLM scheduling is well captured by a linear combination of density, conflicts, inverse min TTC, topology, navigation change, and short-term variation.
    Eq. (7)–(9) define the paper-specific scheduler; not derived from a validated complexity theory beyond cited indicators.
  • domain assumption nuPlan closed-loop Hard20 metrics and CARLA-ROS SIL are sufficient proxies for IoT-enabled urban deployment performance.
    §IV evaluation design; conclusions themselves note missing real perception noise and HIL/real-vehicle validation.
invented entities (3)
  • Surrounding agent-centric data augmentation pipeline no independent evidence
    purpose: Reindex filtered nearby vehicles as planning subjects to enrich long-tail interaction labels without new raw collection.
    Methodological construct introduced in §III-B; evidence is internal ablation score gains only.
  • Complexity-aware asynchronous LLM semantic feature module with adaptive gate no independent evidence
    purpose: Inject high-level navigation/scene semantics into the planning decoder while controlling online LLM cost.
    Builds on AsyncDriver but adds paper-specific C_t schedule and gate; validated only in this paper’s tables/runtime.
  • Residual-based differentiable trajectory optimizer coupled to the IL planner no independent evidence
    purpose: Refine the top ego proposal under explicit soft costs and backpropagate optimization gradients to trajectory generation.
    Composition of known differentiable optimization with paper-specific residual set and cost-weight network; no external formal certificate.

pith-pipeline@v1.1.0-grok45 · 23629 in / 4382 out tokens · 52321 ms · 2026-07-14T11:43:24.510975+00:00 · methodology

0 comments
read the original abstract

Autonomous driving planning is a key component of IoT-enabled intelligent transportation systems, requiring vehicles to generate safe, efficient, and executable trajectories in complex urban environments from multi-source contextual information. While imitation learning (IL) has shown promise on large-scale datasets, IL-based planners still suffer from limited coverage of complex long-tail interactions, weak consistency with downstream constrained refinement, and insufficient use of high level scene semantics under real time constraints. To address these issues, this paper proposes a large language model (LLM) enhanced differentiable trajectory planning framework for IoT-enabled autonomous driving. Specifically, we introduce a surrounding agent centric data augmentation strategy to reorganize sur rounding agent trajectories as additional planning supervision, thereby improving the training distribution without collecting additional raw data. We further design a complexity-aware asyn chronous LLM-based semantic enhancement module to extract scene-related high-level semantic features with controlled online overhead. In addition, a differentiable optimization module is incorporated to refine generated trajectories with explicit residual penalties while backpropagating optimization gradients to the upstream planner. Experiments show that the proposed method achieves the best overall scores of 83.63 and 78.29 on the nuPlan closed-loop nonreactive and reactive Hard20 benchmarks, respectively, and CARLA-ROS tests further verify its online deployment and real time closed-loop execution capability.

Figures

Figures reproduced from arXiv: 2607.10438 by Ashok Kumar Das, Haitao Ding, Hemant Ghayvat, Jing Yang, Lip Yee Por, Shihao Zhang, Sunil Prajapat, Zhaochen Xia, Zheng Lin, Ziyu Song.

Figure 1
Figure 1. Figure 1: Scenario distribution statistics of the nuPlan dataset. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed planning framework. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: SIL platform communication architecture. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison of closed-loop planning results on representative scenarios from the nuPlan Hard20 split. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative closed-loop results on the CARLA-ROS platform, shown from both bounding box and RGB camera views. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] 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

53 extracted references · 6 linked inside Pith

  1. [1]

    Cooperative motion planning and decision making for cavs at roundabouts: A data-efficient learning-based iterative optimization method,

    X. Gong, P. Lyu, and B. Wang, “Cooperative motion planning and decision making for cavs at roundabouts: A data-efficient learning-based iterative optimization method,”IEEE Internet of Things Journal, vol. 11, no. 19, pp. 32 205–32 220, 2024

    2024

  2. [2]

    Explainable deep reinforcement learning for anomaly detection in iot-enabled metaverse healthcare: Toward trustworthy cyber threat intelligence,

    J. Yang, X. Xu, M. A. Khan, G. B. Brahim, J. Baili, L. Y . Por, and C. Li, “Explainable deep reinforcement learning for anomaly detection in iot-enabled metaverse healthcare: Toward trustworthy cyber threat intelligence,”Research, vol. 9, p. 1245, 2026

    2026

  3. [3]

    Gapsl: A gradient-aligned parallel split learning on heterogeneous data,

    Z. Lin, O. Aouedi, W. Ni, S. Chatzinotas, and X. Chen, “Gapsl: A gradient-aligned parallel split learning on heterogeneous data,”arXiv preprint arXiv:2603.18540, 2026

    arXiv 2026

  4. [4]

    IC3M: In-Car Multimodal Multi-Object Monitoring for Abnormal Status of Both Driver and Passengers,

    Z. Fang, Z. Lin, S. Hu, H. Cao, Y . Deng, X. Chen, and Y . Fang, “IC3M: In-Car Multimodal Multi-Object Monitoring for Abnormal Status of Both Driver and Passengers,”arXiv preprint arXiv:2410.02592, 2024

    Pith/arXiv arXiv 2024

  5. [5]

    Enhanced control for handling and stability of 4wd electric vehicles with uncertain stability margins,

    Z. Song, W. Li, B. Yang, H. Ding, and Y . Huang, “Enhanced control for handling and stability of 4wd electric vehicles with uncertain stability margins,”IEEE Transactions on Transportation Electrification, 2024

    2024

  6. [6]

    Safety-critical scenarios for autonomous driving: A survey of methods, benchmarks, and verification pipelines,

    Z. Song, Z. Lin, Y . Hu, Y . Deng, J. Yang, S. Prajapat, Z. Fang, Y . Zhang, L. Y . Por, H. Dinget al., “Safety-critical scenarios for autonomous driving: A survey of methods, benchmarks, and verification pipelines,” 2018

    2018

  7. [7]

    Reliable lane-level localization for autonomous vehicles with enhanced hmm-based map matching and lane-line assistance,

    S. Zhang, J. Yang, Z. Song, Z. Xia, N. Kumar, L. Y . Por, T. R. Gadekallu, and H. Ding, “Reliable lane-level localization for autonomous vehicles with enhanced hmm-based map matching and lane-line assistance,” IEEE Transactions on Vehicular Technology, 2026

    2026

  8. [8]

    Smart-city spatiotemporal data-driven trajec- tory prediction for autonomous vehicles via attention mechanisms and self-supervised learning,

    Z. Song, H. Ding, L. Jamel, J. Yang, M. A. Khan, J. M. Gorriz, J. Baili, and L. Y . Por, “Smart-city spatiotemporal data-driven trajec- tory prediction for autonomous vehicles via attention mechanisms and self-supervised learning,”IEEE Transactions on Consumer Electronics, 2025

    2025

  9. [9]

    Qugrid-fed: Quantum-assisted federated iov energy and security management for sustainable urban transportation systems,

    J. Yang, L. Fang, Z. Liu, Z. A. Shaikh, A. Ksibi, Z. Song, S. Prajapat, and L. Y . Por, “Qugrid-fed: Quantum-assisted federated iov energy and security management for sustainable urban transportation systems,” IEEE Communications Magazine, 2026

    2026

  10. [10]

    Leed: A highly efficient and scalable llm-empowered expert demonstrations framework for multi-agent rein- forcement learning,

    T. Duan, Z. Zhang, S. Guo, D. Huang, Y . Zhao, Z. Lin, Z. Fang, D. Luan, H. Cui, and Y . Cui, “Leed: A highly efficient and scalable llm-empowered expert demonstrations framework for multi-agent rein- forcement learning,”arXiv preprint arXiv:2509.14680, 2025

    arXiv 2025

  11. [11]

    Robust deep reinforcement learning in robotics via adaptive gradient-masked adversarial attacks,

    Z. Zhang, T. Duan, Z. Lin, D. Huang, Z. Fang, Z. Sun, L. Xiong, H. Liang, H. Cui, Y . Cuiet al., “Robust deep reinforcement learning in robotics via adaptive gradient-masked adversarial attacks,”Proc. IROS, 2025

    2025

  12. [12]

    Episodic future thinking with offline reinforce- ment learning for autonomous driving,

    D. Lee and M. Kwon, “Episodic future thinking with offline reinforce- ment learning for autonomous driving,”IEEE Internet of Things Journal, vol. 12, no. 11, pp. 17 012–17 023, 2025

    2025

  13. [13]

    Plant: Explainable planning transformers via object-level representa- tions,

    K. Renz, K. Chitta, O.-B. Mercea, A. Koepke, Z. Akata, and A. Geiger, “Plant: Explainable planning transformers via object-level representa- tions,”arXiv preprint arXiv:2210.14222, 2022

    Pith/arXiv arXiv 2022

  14. [14]

    Urban driver: Learning to drive from real-world demonstrations using policy gradients,

    O. Scheel, L. Bergamini, M. Wolczyk, B. Osi ´nski, and P. Ondruska, “Urban driver: Learning to drive from real-world demonstrations using policy gradients,” inConference on Robot Learning. PMLR, 2022, pp. 718–728

    2022

  15. [15]

    Rethinking imitation-based planners for autonomous driving,

    J. Cheng, Y . Chen, X. Mei, B. Yang, B. Li, and M. Liu, “Rethinking imitation-based planners for autonomous driving,” in2024 IEEE Inter- national Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 14 123–14 130

    2024

  16. [16]

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

    H. Caesar, J. Kabzan, K. S. Tan, W. K. Fong, E. Wolff, A. Lang, L. Fletcher, O. Beijbom, and S. Omari, “nuplan: A closed-loop ml- based planning benchmark for autonomous vehicles,”arXiv preprint arXiv:2106.11810, 2021. IEEE INTERNET OF THINGS JOURNAL 13

    Pith/arXiv arXiv 2021

  17. [17]

    Neuroagent-x: A self-evolving cognitive agent for securing consumer iot systems against ai-enabled anomalies and adversarial threats,

    J. Yang, V . Govindarajan, M. Y . H. Al-Shamri, H. Aldossary, A. Ksibi, Z. A. Shaikh, L. Y . Por, and K. Qi, “Neuroagent-x: A self-evolving cognitive agent for securing consumer iot systems against ai-enabled anomalies and adversarial threats,”IEEE Transactions on Consumer Electronics, vol. 71, no. 4, pp. 12 226–12 235, 2025

    2025

  18. [18]

    Scalability in perception for autonomous driving: Waymo open dataset,

    P. Sun, H. Kretzschmar, X. Dotiwalla, A. Chouard, V . Patnaik, P. Tsui, J. Guo, Y . Zhou, Y . Chai, B. Caineet al., “Scalability in perception for autonomous driving: Waymo open dataset,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 2446–2454

    2020

  19. [19]

    Embedding synthetic off-policy experience for au- tonomous driving via zero-shot curricula,

    E. Bronstein, S. Srinivasan, S. Paul, A. Sinha, M. O’Kelly, P. Nikdel, and S. Whiteson, “Embedding synthetic off-policy experience for au- tonomous driving via zero-shot curricula,” inConference on Robot Learning. PMLR, 2023, pp. 188–198

    2023

  20. [20]

    Multi-modal motion prediction with transformer-based neural network for autonomous driving,

    Z. Huang, X. Mo, and C. Lv, “Multi-modal motion prediction with transformer-based neural network for autonomous driving,” in2022 International Conference on Robotics and Automation (ICRA). IEEE, 2022, pp. 2605–2611

    2022

  21. [21]

    An interacting multiple model for trajectory prediction of intelligent vehicles in typical road traffic scenario,

    H. Gao, Y . Qin, C. Hu, Y . Liu, and K. Li, “An interacting multiple model for trajectory prediction of intelligent vehicles in typical road traffic scenario,”IEEE transactions on neural networks and learning systems, vol. 34, no. 9, pp. 6468–6479, 2021

    2021

  22. [22]

    Multi-agent trajectory prediction with heterogeneous edge-enhanced graph attention network,

    X. Mo, Z. Huang, Y . Xing, and C. Lv, “Multi-agent trajectory prediction with heterogeneous edge-enhanced graph attention network,”IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 7, pp. 9554–9567, 2022

    2022

  23. [23]

    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

  24. [24]

    Pluto: Pushing the limit of imita- tion 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

    Pith/arXiv arXiv 2024

  25. [25]

    Differentiable integrated motion prediction and planning with learnable cost function for autonomous driving,

    Z. Huang, H. Liu, J. Wu, and C. Lv, “Differentiable integrated motion prediction and planning with learnable cost function for autonomous driving,”IEEE transactions on neural networks and learning systems, vol. 35, no. 11, pp. 15 222–15 236, 2023

    2023

  26. [26]

    Pushing large language models to the 6g edge: Vision, challenges, and opportunities,

    Z. Lin, G. Qu, Q. Chen, X. Chen, Z. Chen, and K. Huang, “Pushing large language models to the 6g edge: Vision, challenges, and opportunities,” IEEE Communications Magazine, vol. 63, no. 9, pp. 52–59, 2025

    2025

  27. [27]

    Automated Federated Pipeline for Parameter-Efficient Fine-Tuning of Large Lan- guage Models,

    Z. Fang, Z. Lin, Z. Chen, X. Chen, Y . Gao, and Y . Fang, “Automated Federated Pipeline for Parameter-Efficient Fine-Tuning of Large Lan- guage Models,”IEEE Trans. Mobile Comput., 2025

    2025

  28. [28]

    Hsplitlora: A heterogeneous split parameter- efficient fine-tuning framework for large language models,

    Z. Lin, Y . Zhang, Z. Chen, Z. Fang, X. Chen, P. Vepakomma, W. Ni, J. Luo, and Y . Gao, “Hsplitlora: A heterogeneous split parameter- efficient fine-tuning framework for large language models,”IEEE Trans- actions on Mobile Computing, 2026

    2026

  29. [29]

    LLM-Driven Stationarity-Aware Expert Demonstrations for Multi-Agent Reinforcement Learning in Mobile Systems,

    T. Duan, Z. Zhang, Z. Lin, S. Guo, X. Guan, G. Wu, Z. Fang, H. Meng, X. Du, J.-Z. Zhouet al., “LLM-Driven Stationarity-Aware Expert Demonstrations for Multi-Agent Reinforcement Learning in Mobile Systems,”arXiv preprint arXiv:2511.19368, 2025

    arXiv 2025

  30. [30]

    Hfedmoe: Resource-aware heterogeneous federated learning with mixture-of-experts,

    Z. Fang, Z. Lin, S. Hu, Y . Ma, Y . Tao, Y . Deng, X. Chen, and Y . Fang, “Hfedmoe: Resource-aware heterogeneous federated learning with mixture-of-experts,”arXiv preprint arXiv:2601.00583, 2026

    arXiv 2026

  31. [31]

    A survey of data augmentation techniques for traffic visual elements,

    M. Yang, L. S. Ewe, W. K. Yew, S. Deng, and S. K. Tiong, “A survey of data augmentation techniques for traffic visual elements,”Sensors, vol. 25, no. 21, p. 6672, 2025

    2025

  32. [32]

    Chatgpt-based sce- nario engineer: A new framework on scenario generation for trajectory prediction,

    X. Li, E. Liu, T. Shen, J. Huang, and F.-Y . Wang, “Chatgpt-based sce- nario engineer: A new framework on scenario generation for trajectory prediction,”IEEE Transactions on Intelligent Vehicles, vol. 9, no. 3, pp. 4422–4431, 2024

    2024

  33. [33]

    Augmenting safety-critical driving scenarios while preserving similarity to expert trajectories,

    H. Mirkhani, B. Khamidehi, and K. Rezaee, “Augmenting safety-critical driving scenarios while preserving similarity to expert trajectories,” in 2024 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2024, pp. 2085– 2090

    2024

  34. [34]

    Gameformer: Game-theoretic modeling and learning of transformer-based interactive prediction and planning for autonomous driving,

    Z. Huang, H. Liu, and C. Lv, “Gameformer: Game-theoretic modeling and learning of transformer-based interactive prediction and planning for autonomous driving,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 3903–3913

    2023

  35. [35]

    Optnet: Differentiable optimization as a layer in neural networks,

    B. Amos and J. Z. Kolter, “Optnet: Differentiable optimization as a layer in neural networks,” inInternational conference on machine learning. PMLR, 2017, pp. 136–145

    2017

  36. [36]

    Differentiable mpc for end-to-end planning and control,

    B. Amos, I. Jimenez, J. Sacks, B. Boots, and J. Z. Kolter, “Differentiable mpc for end-to-end planning and control,” inAdvances in Neural Information Processing Systems, vol. 31. Curran Associates, Inc., 2018, pp. 8289–8300

    2018

  37. [37]

    Lmdrive: Closed-loop end-to-end driving with large language models,

    H. Shao, Y . Hu, L. Wang, G. Song, S. L. Waslander, Y . Liu, and H. Li, “Lmdrive: Closed-loop end-to-end driving with large language models,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2024, pp. 15 120–15 130

    2024

  38. [38]

    Lego-drive: Language-enhanced goal-oriented closed-loop end-to-end autonomous driving,

    P. Paul, A. Garg, T. Choudhary, A. K. Singh, and K. M. Krishna, “Lego-drive: Language-enhanced goal-oriented closed-loop end-to-end autonomous driving,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 10 020–10 026

    2024

  39. [39]

    Drivelm: Driving with graph visual ques- tion answering,

    C. Sima, K. Renz, K. Chitta, L. Chen, H. Zhang, C. Xie, J. Beißwenger, P. Luo, A. Geiger, and H. Li, “Drivelm: Driving with graph visual ques- tion answering,” inEuropean conference on computer vision. Springer, 2024, pp. 256–274

    2024

  40. [40]

    Reason2drive: Towards interpretable and chain-based reasoning for autonomous driving,

    M. Nie, R. Peng, C. Wang, X. Cai, J. Han, H. Xu, and L. Zhang, “Reason2drive: Towards interpretable and chain-based reasoning for autonomous driving,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 292–308

    2024

  41. [41]

    Llm4ad: Large language models for au- tonomous driving—concept, review, benchmark, experiments, and future trends,

    C. Cui, Y . Ma, S.-Y . Park, Z. Yang, Y . Zhou, P. Liu, J. Lu, J. Peng, J. Zhang, R. Zhanget al., “Llm4ad: Large language models for au- tonomous driving—concept, review, benchmark, experiments, and future trends,”Proceedings of the IEEE, 2026

    2026

  42. [42]

    SplitLoRA: A Split Parameter-Efficient Fine-Tuning Framework for Large Language Models,

    Z. Lin, X. Hu, Y . Zhang, Z. Chen, Z. Fang, X. Chen, A. Li, P. Vepakomma, and Y . Gao, “SplitLoRA: A Split Parameter-Efficient Fine-Tuning Framework for Large Language Models,”arXiv preprint arXiv:2407.00952, 2024

    Pith/arXiv arXiv 2024

  43. [43]

    Mobile edge intelligence for large language models: A contemporary survey,

    G. Qu, Q. Chen, W. Wei, Z. Lin, X. Chen, and K. Huang, “Mobile edge intelligence for large language models: A contemporary survey,”IEEE Communications Surveys & Tutorials, 2025

    2025

  44. [44]

    Hi- erarchical Split Federated Learning: Convergence Analysis and System Optimization,

    Z. Lin, W. Wei, Z. Chen, C.-T. Lam, X. Chen, Y . Gao, and J. Luo, “Hi- erarchical Split Federated Learning: Convergence Analysis and System Optimization,”IEEE Trans. Mobile Comput., 2025

    2025

  45. [45]

    Dynamic path- speed planning algorithm for autonomous driving on structured roads,

    L. Fang, Z. Xiaowen, S. Weixing, and W. Yonggang, “Dynamic path- speed planning algorithm for autonomous driving on structured roads,” Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, vol. 238, no. 10-11, pp. 3172–3193, 2024

    2024

  46. [46]

    Optimal resource allocation for u-shaped parallel split learning,

    S. Lyu, Z. Lin, G. Qu, X. Chen, X. Huang, and P. Li, “Optimal resource allocation for u-shaped parallel split learning,” in2023 IEEE Globecom Workshops (GC Wkshps), 2023, pp. 197–202

    2023

  47. [47]

    Asynchronous large language model enhanced planner for autonomous driving,

    Y . Chen, Z.-h. Ding, Z. Wang, Y . Wang, L. Zhang, and S. Liu, “Asynchronous large language model enhanced planner for autonomous driving,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 22–38

    2024

  48. [48]

    Complexity evaluation for urban intersection scenarios in autonomous driving tests: Method and validation,

    J. Li, R. Zong, Y . Wang, and W. Deng, “Complexity evaluation for urban intersection scenarios in autonomous driving tests: Method and validation,”Applied Sciences, vol. 14, no. 22, p. 10451, 2024

    2024

  49. [49]

    Event-triggered model predictive control for autonomous vehicle path tracking: Validation using carla simulator,

    Z. Zhou, C. Rother, and J. Chen, “Event-triggered model predictive control for autonomous vehicle path tracking: Validation using carla simulator,”IEEE Transactions on Intelligent Vehicles, vol. 8, no. 6, pp. 3547–3555, 2023

    2023

  50. [50]

    Theseus: A library for differentiable nonlinear optimization,

    L. Pineda, T. Fan, M. Monge, S. Venkataraman, P. Sodhi, R. T. Chen, J. Ortiz, D. DeTone, A. Wang, S. Andersonet al., “Theseus: A library for differentiable nonlinear optimization,”Advances in Neural Information Processing Systems, vol. 35, pp. 3801–3818, 2022

    2022

  51. [51]

    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, vol. 62, no. 2, pp. 1805–1824, 2000

    2000

  52. [52]

    From prediction to planning with goal conditioned lane graph traversals,

    M. Hallgarten, M. Stoll, and A. Zell, “From prediction to planning with goal conditioned lane graph traversals,” in2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC). IEEE, 2023, pp. 951–958

    2023

  53. [53]

    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. Zhanet al., “Diffusion-based planning for autonomous driving with flexible guidance,”arXiv preprint arXiv:2501.15564, 2025

    Pith/arXiv arXiv 2025