Pith. sign in

REVIEW 3 major objections 5 minor 299 references

Predicting every nearby vehicle's lane-change intent and path together yields more accurate and collision-free traffic forecasts.

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 16:36 UTC pith:YBSIMRPO

load-bearing objection Solid scene-level packaging of known pieces with real multi-dataset gains; the evaluation unit is load-bearing but the core offline results still look useful. the 3 major comments →

arxiv 2607.09740 v1 pith:YBSIMRPO submitted 2026-07-02 cs.AI cs.CV

A Dynamic Scene Interaction Reasoning Framework for Scene-level Lane-Change Intention and Trajectory Prediction of Multiple Interacting Vehicles

Joshua Kofi Asamoah , Blessing Agyei Kyem , Eugene Denteh , Armstrong Aboah This is my paper
classification cs.AI cs.CV
keywords lane-change intention predictiontrajectory forecastingdynamic scene graphgraph attentionmulti-agent predictionautonomous drivingscene-level consistency
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.

Safe planning needs to know how the whole local traffic scene will evolve, not just one target car. Existing lane-change methods often treat neighbors as passive context, while multi-agent trajectory models forecast positions without naming each vehicle's maneuver. This paper claims that representing a local highway scene as a time-varying interaction graph, with vehicles as nodes and spatial-kinematic relations as explicit edges, lets one model jointly predict lane-keeping, left-change, or right-change for every valid vehicle and the matching 4-second trajectory. Temporal graph attention captures evolving pre-maneuver cues; an intention-guided decoder ties each maneuver probability to a maneuver-specific motion path; and a scene consistency loss penalizes futures that place vehicles too close. On NGSIM I-80, US-101, and highD the approach reaches roughly 90–91% intention accuracy, cuts trajectory error by up to about half versus the strongest baseline, and lowers inter-agent collision rate and joint displacement error, producing futures that are both more accurate per vehicle and more coherent as a scene.

Core claim

A dynamic scene graph attention network (DSiGAT) that jointly outputs lane-change intention and future trajectory for every valid vehicle in a local multi-vehicle scene outperforms both target-centered intention models and multi-agent trajectory baselines on standard highway datasets, while also producing lower inter-agent collision rates and joint displacement errors.

What carries the argument

DSiGAT: a time-varying interaction graph whose nodes carry vehicle kinematics and lane state, whose directed edges encode gaps, relative speeds, lane relation, and time-to-collision; temporal graph-attention message passing; intention-guided weighted mixture of three maneuver-specific trajectory decoders; and a scene-level separation consistency loss.

Load-bearing premise

That a local scene of at most six continuously observed vehicles, linked only when they share a lane or adjacent lane and stay within 100 m longitudinally, is enough to capture the interactions that matter for planning.

What would settle it

On held-out highway scenes, check whether joint intention-plus-trajectory prediction of every valid vehicle still beats strong multi-agent baselines on accuracy, RMSE, inter-agent collision rate, and joint displacement error once scenes regularly contain more than six relevant neighbors or when edge construction ignores distant but influential vehicles.

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 proposes DSiGAT, a dynamic scene-graph attention model that jointly predicts lane-change intention (LK/LLC/RLC) and 4 s trajectories for every valid vehicle in a local highway scene (at most 6 continuously observed vehicles). Scenes are built as time-varying directed graphs with node kinematics and explicit edge features (gaps, relative speeds, lane relation, TTC); temporal graph attention produces scene-aware embeddings, an intention head yields per-vehicle maneuver probabilities, and three maneuver-specific decoders are mixed by those probabilities. Training adds an intention–trajectory gating term and a pairwise separation penalty (L_scene). On NGSIM I-80/US-101 and highD, the authors report ~90–91% intention accuracy, large RMSE reductions versus GRIP++ and other baselines, and lower IACR/JDE, with progressive ablations, anticipation-time curves, noise robustness, loss-weight sensitivity, and qualitative cases.

Significance. If the gains hold under a fair multi-agent protocol, the work is a useful step beyond target-centered lane-change models and trajectory-only multi-agent forecasters: it couples explicit maneuver labels to motion for all agents in a local scene and measures joint physical compatibility (IACR/JDE). Strengths include multi-dataset evaluation, trajectory-group splits, progressive ablations (Table 11), anticipation-time analysis, scene-level metrics, and efficiency numbers. The contribution is primarily empirical/architectural rather than theoretical; its lasting value depends on whether scene-level superiority is attributable to the proposed components rather than to the DSiGAT-native N_max=6 filtered neighborhood used for evaluation.

major comments (3)
  1. §5.2.1 and §3.3 / Table 14: The central scene-level claim (lower IACR/JDE and large RMSE gains vs multi-agent baselines) is load-bearing on the evaluation unit. Scenes are DSiGAT-native (≤6 continuously observed nearest vehicles; edges only for same/adjacent lanes and |Δx|<100 m; heavy filtering in Table 2). Baselines keep “architecture-specific feature construction” but are scored on the same retained scenes and synchronized pairs. Without a control that re-runs the strongest multi-agent baselines under their native denser/map-aware neighborhoods (or reports IACR/JDE on a common absolute frame without N_max truncation), part of the reported scene-level superiority may measure the evaluation neighborhood rather than the architecture. Please add that control or clearly bound the claim to the N_max=6 filtered setting.
  2. Tables 9–10 and 13 vs §4.3.4–4.3.6: Trajectory RMSE reductions of up to ~50% vs GRIP++ are unusually large for this literature. The paper should clarify whether baselines receive the same intention-conditioned decoding opportunity (or equivalent multi-task supervision), the same relative-displacement targets, and the same validity masks, and whether GRIP++/Social Conv/MATF were reimplemented or only partially adapted. A short protocol appendix (inputs, outputs, loss, and any missing map/context features) is needed so the RMSE gap can be attributed to DSiGAT rather than unequal task formulation.
  3. §3.4 Eq. (2) and §5.3.1: Maneuver labels are derived solely from persistent future lane-index transitions after aggressive cleaning (Table 2 removes observation-window changes, unstable reversals, boundary tracks). This is reasonable for anticipation, but it systematically excludes hard/ambiguous cases that later appear as the main failure mode (Fig. 8). The paper should quantify how many candidate events are discarded by each filter and report intention metrics on a less-filtered or onset-inclusive hold-out, so that the ~90% accuracy is not overstated for planning-relevant early cues.
minor comments (5)
  1. §4.3.6 Eq. (32): Loss weights (10, 1, 2, 0.5) are fixed after sensitivity (Fig. 9); state whether they were selected only on validation and whether the same weights transfer to highD without retuning.
  2. §4.2.2 Eq. (11): TTC uses a hard saturation T_max=10 s and ε=10^{-3}; a one-line justification or sensitivity note would help, given that TTC is an edge feature feeding attention.
  3. §5.3.3: IACR uses D_safe=2.0 m without vehicle length/width; note that this is a point-mass proxy and may not match physical collision geometry.
  4. Presentation: occasional missing spaces in compound words (e.g., “multi-agentforecasting,” “lane-changeintention”) and dense tables would benefit from a light copy-edit pass.
  5. Related work: briefly contrast L_scene with JFP-style pairwise compatibility (already cited) to clarify what is new versus re-used.

Circularity Check

0 steps flagged

No significant circularity: empirical multi-task model trained and scored on held-out public trajectory splits; self-citations are peripheral and non-load-bearing.

full rationale

DSiGAT is an empirical graph-attention architecture whose claims rest on measured accuracy/RMSE/IACR/JDE against baselines on NGSIM I-80, US-101 and highD under trajectory-group-level held-out splits (Sec. 5.1–5.2). The training objective (Eq. 32: weighted CE + MSE + gate + scene collision penalty) and loss weights (tuned on validation) are ordinary multi-task fitting; they do not redefine the test metrics by construction, nor do any equations reduce a claimed “prediction” to a fitted input. Scene construction (N_max=6, edge rules, filtering in Sec. 3 and Table 2) is a modeling choice that may affect fairness of multi-agent comparisons, but that is an experimental-design concern, not circular derivation. The sole self-citation (Asamoah et al. 2026 on perception) appears only as background and is not used to justify uniqueness, uniqueness theorems, or any central result. No self-definitional loops, no uniqueness imported from the authors, and no renaming of known results as first-principles discoveries. The paper is therefore self-contained against external benchmarks; score 1 only for the minor non-load-bearing self-cite.

Axiom & Free-Parameter Ledger

6 free parameters · 5 axioms · 3 invented entities

The central claim is empirical performance of a designed neural system under fixed data-construction and training choices. Load-bearing free parameters include scene size, edge thresholds, safety distance, and multi-task loss weights. Domain axioms are standard highway trajectory-prediction assumptions (cleaned tracks, lane-index intention labels, fixed 3 s/4 s windows). Invented entities are architectural modules, not new physical objects; they have no independent evidence outside this paper’s experiments.

free parameters (6)
  • loss weights (λ_int, λ_traj, λ_gate, λ_scene)
    Set to 10.0, 1.0, 2.0, 0.5 after sensitivity analysis; balance of the four training terms and thus the reported multi-task trade-off depends on these choices.
  • N_max scene capacity
    Hard cap of 6 vehicles per scene; defines which interactions enter the model and evaluation.
  • edge longitudinal cutoff (100 m) and lane-adjacency rule
    Determines graph connectivity and which relative features are visible to attention.
  • D_safe minimum separation
    Fixed at 2.0 m for scene consistency loss and IACR; directly shapes the scene-level objective and metric.
  • embedding/attention hyperparameters (d_h=128, d_g=64, 8 heads, 3 MP rounds)
    Capacity and message-passing depth chosen by design; affect absolute performance though not claimed as universal constants.
  • TTC saturation T_max and epsilon
    T_max=10 s and ε=1e-3 fix the safety edge feature scale.
axioms (5)
  • domain assumption Lane-change intention ground truth can be recovered from persistent normalized lane-index transitions in the 4 s prediction window after cleaning unstable reversals.
    Sec. 3.4 labeling rule; if labels are noisy or not true intention, classification metrics misstate behavioral prediction quality.
  • domain assumption A local neighborhood of continuously observed vehicles (≤6, same/adjacent lanes, <100 m) is the right prediction unit for planning-relevant scene evolution.
    Sec. 3.3–4.2; excludes longer-range and non-mainline interactions by construction.
  • ad hoc to paper Graph attention over node/edge embeddings plus temporal attention is an adequate inductive bias for pre-maneuver interaction cues on highways.
    Core modeling choice of DSiGAT (Sec. 4.3); justified empirically, not derived.
  • ad hoc to paper Weighted mixture of three maneuver-specific trajectory decoders correctly couples intention probabilities to future motion.
    Eq. (25) intention-guided decoding; alternative multimodal heads could couple differently.
  • domain assumption Standard supervised learning on offline naturalistic trajectories with group-level splits yields transferable estimates of online prediction quality.
    Implicit throughout Sec. 5–6; no closed-loop or onboard perception evaluation.
invented entities (3)
  • DSiGAT dynamic scene interaction graph attention architecture no independent evidence
    purpose: Unify multi-vehicle intention classification, intention-guided trajectory decoding, and scene consistency in one model.
    New composite system defined in Sec. 4; performance evidence is only the paper’s own experiments.
  • Scene-level consistency loss L_scene (pairwise separation penalty) no independent evidence
    purpose: Penalize mutually incompatible multi-vehicle futures during training.
    Eqs. (27)–(28); related to prior joint-future ideas but instantiated here as a simple distance hinge with fixed D_safe.
  • Intention–trajectory gating loss L_gate no independent evidence
    purpose: Align predicted maneuver probabilities with net lateral displacement sign/magnitude.
    Eq. (31); paper-specific multi-task regularizer without external validation beyond ablations.

pith-pipeline@v1.1.0-grok45 · 41371 in / 4100 out tokens · 47912 ms · 2026-07-14T16:36:06.991672+00:00 · methodology

0 comments
read the original abstract

Safe motion planning in advanced driver-assistance systems and autonomous vehicles requires an accurate understanding of how the surrounding traffic scene is likely to evolve. However, many existing lane-change prediction methods remain centered on a single target vehicle, while multi-agent forecasting approaches often describe scene evolution only through future positions and provide limited explicit information about the maneuver associated with each vehicle. This study proposes a dynamic scene graph attention framework that predicts the lane-change intention and future trajectory of every relevant vehicle within a local traffic scene. The scene is represented as a time-varying interaction graph in which vehicles are modeled as nodes and their spatial and kinematic relationships are encoded through explicit edge features. Temporal graph-attention message passing captures evolving inter-vehicle dependencies and pre-maneuver cues, while an intention-guided decoder links each predicted maneuver to its corresponding future motion. A scene-level consistency objective further encourages compatible multi-vehicle futures. Experiments on the NGSIM I-80, NGSIM US-101, and highD datasets demonstrate consistent improvements over competing baselines. DSiGAT achieves intention prediction accuracies of 90.12% and 90.97% on NGSIM I-80 and US-101, respectively, and reduces trajectory RMSE by up to 52.94% relative to the strongest baseline. It also produces lower inter-agent collision rates and joint displacement errors, indicating more coherent scene-level predictions. Ablation, sensitivity, robustness, and qualitative analyses further validate the contribution of the proposed components and the effectiveness of the scene-focused formulation.

Figures

Figures reproduced from arXiv: 2607.09740 by Armstrong Aboah, Blessing Agyei Kyem, Eugene Denteh, Joshua Kofi Asamoah.

Figure 1
Figure 1. Figure 1: Overall architecture of DSiGAT for scene-level joint lane-change intention and trajectory prediction. Dynamic vehicle graphs are encoded through node-edge embeddings and temporal graph attention, followed by per-vehicle intention classification, intention-guided trajectory decoding, and joint loss optimization interaction state through the edge features, and the time￾varying scene structure through the gra… view at source ↗
Figure 2
Figure 2. Figure 2: Confusion matrices of different methods for per-vehicle lane-change intention prediction on the NGSIM I-80 test set [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Confusion matrices of different methods for per-vehicle lane-change intention prediction on the NGSIM US-101 test set [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Accuracy comparison of per-vehicle lane-change intention prediction at different anticipation times on the NGSIM I-80 test set. dynamic scene-level interaction reasoning with intention￾guided trajectory decoding provides more accurate motion forecasts than trajectory-only formulations. The advantage of DSiGAT remains consistent as the pre￾diction horizon increases. This is particularly important for ADAS a… view at source ↗
Figure 5
Figure 5. Figure 5: Accuracy of DSiGAT and competing methods at different anticipation times on the NGSIM US-101 test set [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative comparison of lane-change trajectory prediction under scene context for representative left-lane-change and right-lane-change cases on the NGSIM and highD datasets against competitive baselines. behavior. In sum, the qualitative results show that the pro￾posed architecture improves prediction in two related ways: it captures the timing and shape of lane-change trajectories more accurately, and … view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative examples of joint lane-change intention and trajectory prediction on the NGSIM and highD datasets. Ground-truth and predicted maneuver labels are shown for the key lane-changing vehicle in each scene, while the remaining vehicles in the scene are correctly predicted as lane keeping and therefore are not labeled individually for clarity. vehicles within the same local scene and are therefore eva… view at source ↗
Figure 8
Figure 8. Figure 8: Representative failure cases of DSiGAT on the NGSIM and highD datasets. The examples illustrate situations in which weak or ambiguous early maneuver cues lead to intention misclassification, particularly between lane keeping and lane-change behaviors, while the predicted trajectories remain smooth and physically plausible [PITH_FULL_IMAGE:figures/full_fig_p023_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Sensitivity analysis of the loss weights in DSiGAT. The effects of 𝜆int, 𝜆traj, 𝜆gate, and 𝜆scene are evaluated using Macro F1-score and RMSE@4s [PITH_FULL_IMAGE:figures/full_fig_p025_9.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

299 extracted references · 1 canonical work pages

  1. [1]

    Transportation Research Part B: Methodological , volume=

    A model for the structure of lane-changing decisions , author=. Transportation Research Part B: Methodological , volume=. 1986 , publisher=

    1986

  2. [2]

    Proceedings of the human factors and ergonomics society annual meeting , volume=

    Using support vector machines for lane-change detection , author=. Proceedings of the human factors and ergonomics society annual meeting , volume=. 2005 , organization=

    2005

  3. [3]

    2013 IEEE Intelligent Vehicles Symposium (IV) , pages=

    Learning-based approach for online lane change intention prediction , author=. 2013 IEEE Intelligent Vehicles Symposium (IV) , pages=. 2013 , organization=

    2013

  4. [4]

    IEEE Transactions on Intelligent Transportation Systems , volume=

    On the roles of eye gaze and head dynamics in predicting driver's intent to change lanes , author=. IEEE Transactions on Intelligent Transportation Systems , volume=. 2009 , publisher=

    2009

  5. [5]

    Sensors , volume=

    Prediction of driver’s intention of lane change by augmenting sensor information using machine learning techniques , author=. Sensors , volume=. 2017 , publisher=

    2017

  6. [6]

    Computer Communications , volume=

    A driving intention prediction method based on hidden Markov model for autonomous driving , author=. Computer Communications , volume=. 2020 , publisher=

    2020

  7. [7]

    2017 IEEE 20th International Conference on Intelligent Transportation Systems (ITSC) , pages=

    Convolution neural network-based lane change intention prediction of surrounding vehicles for ACC , author=. 2017 IEEE 20th International Conference on Intelligent Transportation Systems (ITSC) , pages=. 2017 , organization=

    2017

  8. [8]

    Transportation Research Part C: Emerging Technologies , volume=

    An ensemble deep learning approach for driver lane change intention inference , author=. Transportation Research Part C: Emerging Technologies , volume=. 2020 , publisher=

    2020

  9. [9]

    Transportation research part C: emerging technologies , volume=

    An improved learning-based LSTM approach for lane change intention prediction subject to imbalanced data , author=. Transportation research part C: emerging technologies , volume=. 2021 , publisher=

    2021

  10. [10]

    Transportation research part C: emerging technologies , volume=

    A Bayesian inference based adaptive lane change prediction model , author=. Transportation research part C: emerging technologies , volume=. 2021 , publisher=

    2021

  11. [11]

    Scientific Reports , volume=

    Lane-change risk prediction model based on traffic context and driving styles , author=. Scientific Reports , volume=. 2025 , publisher=

    2025

  12. [12]

    IEEE Transactions on Intelligent Transportation Systems , volume=

    Dual transformer based prediction for lane change intentions and trajectories in mixed traffic environment , author=. IEEE Transactions on Intelligent Transportation Systems , volume=. 2023 , publisher=

    2023

  13. [13]

    Accident Analysis & Prevention , volume=

    Lane-change intention recognition considering oncoming traffic: Novel insights revealed by advances in deep learning , author=. Accident Analysis & Prevention , volume=. 2024 , publisher=

    2024

  14. [14]

    Transportation research part C: emerging technologies , volume=

    Driver lane change intention prediction based on topological graph constructed by driver behaviors and traffic context for human-machine co-driving system , author=. Transportation research part C: emerging technologies , volume=. 2024 , publisher=

    2024

  15. [15]

    World Electric Vehicle Journal , volume=

    Research on Vehicle Lane Change Intent Recognition Based on Transformers and Bidirectional Gated Recurrent Units , author=. World Electric Vehicle Journal , volume=. 2025 , publisher=

    2025

  16. [16]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Social lstm: Human trajectory prediction in crowded spaces , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  17. [17]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Social gan: Socially acceptable trajectories with generative adversarial networks , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  18. [18]

    European conference on computer vision , pages=

    Trajectron++: Dynamically-feasible trajectory forecasting with heterogeneous data , author=. European conference on computer vision , pages=. 2020 , organization=

    2020

  19. [19]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Covernet: Multimodal behavior prediction using trajectory sets , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  20. [20]

    arXiv preprint arXiv:1910.05449 , year=

    Multipath: Multiple probabilistic anchor trajectory hypotheses for behavior prediction , author=. arXiv preprint arXiv:1910.05449 , year=

    Pith/arXiv arXiv 1910

  21. [21]

    2022 international conference on robotics and automation (ICRA) , pages=

    Multipath++: Efficient information fusion and trajectory aggregation for behavior prediction , author=. 2022 international conference on robotics and automation (ICRA) , pages=. 2022 , organization=

    2022

  22. [22]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Vectornet: Encoding hd maps and agent dynamics from vectorized representation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  23. [23]

    European Conference on Computer Vision , pages=

    Learning lane graph representations for motion forecasting , author=. European Conference on Computer Vision , pages=. 2020 , organization=

    2020

  24. [24]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Agentformer: Agent-aware transformers for socio-temporal multi-agent forecasting , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  25. [25]

    arXiv preprint arXiv:2106.08417 , year=

    Scene transformer: A unified architecture for predicting multiple agent trajectories , author=. arXiv preprint arXiv:2106.08417 , year=

    Pith/arXiv arXiv

  26. [26]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Hivt: Hierarchical vector transformer for multi-agent motion prediction , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  27. [27]

    arXiv preprint arXiv:2207.05844 , year=

    Wayformer: Motion forecasting via simple & efficient attention networks , author=. arXiv preprint arXiv:2207.05844 , year=

    Pith/arXiv arXiv

  28. [28]

    Advances in Neural Information Processing Systems , volume=

    Motion transformer with global intention localization and local movement refinement , author=. Advances in Neural Information Processing Systems , volume=

  29. [29]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=

    Mtr++: Multi-agent motion prediction with symmetric scene modeling and guided intention querying , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=. 2024 , publisher=

    2024

  30. [30]

    IEEE Transactions on Intelligent Transportation Systems , volume=

    An integrated approach to maneuver-based trajectory prediction and criticality assessment in arbitrary road environments , author=. IEEE Transactions on Intelligent Transportation Systems , volume=. 2016 , publisher=

    2016

  31. [31]

    IEEE Transactions on Intelligent Vehicles , volume=

    How would surround vehicles move? A unified framework for maneuver classification and motion prediction , author=. IEEE Transactions on Intelligent Vehicles , volume=. 2018 , publisher=

    2018

  32. [32]

    Proceedings of the IEEE conference on computer vision and pattern recognition workshops , pages=

    Convolutional social pooling for vehicle trajectory prediction , author=. Proceedings of the IEEE conference on computer vision and pattern recognition workshops , pages=

  33. [33]

    2018 21st International Conference on Intelligent Transportation Systems (ITSC) , pages=

    Intention-aware long horizon trajectory prediction of surrounding vehicles using dual LSTM networks , author=. 2018 21st International Conference on Intelligent Transportation Systems (ITSC) , pages=. 2018 , organization=

    2018

  34. [34]

    IEEE Transactions on Intelligent Vehicles , volume=

    The integration of prediction and planning in deep learning automated driving systems: A review , author=. IEEE Transactions on Intelligent Vehicles , volume=. 2024 , publisher=

    2024

  35. [35]

    2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=

    STFP: Simultaneous traffic scene forecasting and planning for autonomous driving , author=. 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=. 2021 , organization=

    2021

  36. [36]

    2018 IEEE Intelligent Vehicles Symposium (IV) , pages=

    Learning vehicle surrounding-aware lane-changing behavior from observed trajectories , author=. 2018 IEEE Intelligent Vehicles Symposium (IV) , pages=. 2018 , organization=

    2018

  37. [37]

    IEEE Transactions on Intelligent Transportation Systems , year=

    Lane change prediction for autonomous driving with transferred trajectory interaction , author=. IEEE Transactions on Intelligent Transportation Systems , year=

  38. [38]

    arXiv preprint arXiv:2405.03809 , year=

    SocialFormer: Social interaction modeling with edge-enhanced heterogeneous graph transformers for trajectory prediction , author=. arXiv preprint arXiv:2405.03809 , year=

    Pith/arXiv arXiv

  39. [39]

    2026 6th International Conference on Image Processing and Capsule Networks (ICIPCN) , pages=

    Explainable Deep Neural Architecture for Lane-Change Prediction using Vision and Vehicle Interaction Modeling , author=. 2026 6th International Conference on Image Processing and Capsule Networks (ICIPCN) , pages=. 2026 , organization=

    2026

  40. [40]

    Transportation research record , volume=

    Machine learning approach for predicting lane-change maneuvers using the SHRP2 naturalistic driving study data , author=. Transportation research record , volume=. 2021 , publisher=

    2021

  41. [41]

    International Journal of Sensor Networks , volume=

    Introduction to intelligent transportation system: overview, classification based on physical architecture, and challenges , author=. International Journal of Sensor Networks , volume=. 2022 , publisher=

    2022

  42. [42]

    2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=

    Joint intention and trajectory prediction based on transformer , author=. 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=. 2021 , organization=

    2021

  43. [43]

    Conference on robot learning , pages=

    Multimodal trajectory prediction conditioned on lane-graph traversals , author=. Conference on robot learning , pages=. 2022 , organization=

    2022

  44. [44]

    IEEE Transactions on Intelligent Vehicles , volume=

    A temporal multi-gate mixture-of-experts approach for vehicle trajectory and driving intention prediction , author=. IEEE Transactions on Intelligent Vehicles , volume=. 2023 , publisher=

    2023

  45. [45]

    IEEE Transactions on Intelligent Vehicles , volume=

    Lane change--intention inference and trajectory prediction of surrounding vehicles on highways , author=. IEEE Transactions on Intelligent Vehicles , volume=. 2023 , publisher=

    2023

  46. [46]

    Conference on Robot Learning , pages=

    Jfp: Joint future prediction with interactive multi-agent modeling for autonomous driving , author=. Conference on Robot Learning , pages=. 2023 , organization=

    2023

  47. [47]

    2017 IEEE intelligent vehicles symposium (IV) , pages=

    Imitating driver behavior with generative adversarial networks , author=. 2017 IEEE intelligent vehicles symposium (IV) , pages=. 2017 , organization=

    2017

  48. [48]

    2021 , publisher=

    Autonomous driving and advanced driver-assistance systems (ADAS): applications, development, legal issues, and testing , author=. 2021 , publisher=

    2021

  49. [49]

    In 2018 IEEE , author=

    Interaction-aware probabilistic behavior prediction in urban environments. In 2018 IEEE , author=. RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=

    2018

  50. [50]

    Authorea Preprints , year=

    Sparse Attention Graph Convolution Network for Vehicle Trajectory Prediction , author=. Authorea Preprints , year=

  51. [51]

    IEEE Transactions on Intelligent Vehicles , volume=

    A survey on trajectory-prediction methods for autonomous driving , author=. IEEE Transactions on Intelligent Vehicles , volume=. 2022 , publisher=

    2022

  52. [52]

    IEEE Transactions on Intelligent Vehicles , volume=

    Attention based vehicle trajectory prediction , author=. IEEE Transactions on Intelligent Vehicles , volume=. 2020 , publisher=

    2020

  53. [53]

    IEEE Intelligent Transportation Systems Magazine , volume=

    Vehicle trajectory prediction using LSTMs with spatial--temporal attention mechanisms , author=. IEEE Intelligent Transportation Systems Magazine , volume=. 2021 , publisher=

    2021

  54. [54]

    IEEE Transactions on Intelligent Transportation Systems , volume=

    Trajectory prediction-based local spatio-temporal navigation map for autonomous driving in dynamic highway environments , author=. IEEE Transactions on Intelligent Transportation Systems , volume=. 2021 , publisher=

    2021

  55. [55]

    IEEE Transactions on Vehicular Technology , volume=

    Environment-attention network for vehicle trajectory prediction , author=. IEEE Transactions on Vehicular Technology , volume=. 2021 , publisher=

    2021

  56. [56]

    Applied Sciences , volume=

    Vehicle Trajectory Prediction Based on Graph Convolutional Networks in Connected Vehicle Environment , author=. Applied Sciences , volume=. 2023 , publisher=

    2023

  57. [57]

    Authorea Preprints , year=

    A review on intention-aware and interaction-aware trajectory prediction for autonomous vehicles , author=. Authorea Preprints , year=

  58. [58]

    ROBOMECH journal , volume=

    A survey on motion prediction and risk assessment for intelligent vehicles , author=. ROBOMECH journal , volume=. 2014 , publisher=

    2014

  59. [59]

    IEEE Transactions on Intelligent Transportation Systems , volume=

    Looking at intersections: a survey of intersection monitoring, behavior and safety analysis of recent studies , author=. IEEE Transactions on Intelligent Transportation Systems , volume=. 2016 , publisher=

    2016

  60. [60]

    IEEE Transactions on Intelligent Transportation Systems , volume=

    Deep learning-based vehicle behavior prediction for autonomous driving applications: A review , author=. IEEE Transactions on Intelligent Transportation Systems , volume=. 2020 , publisher=

    2020

  61. [61]

    arXiv preprint arXiv:1909.07707 , year=

    A review of tracking, prediction and decision making methods for autonomous driving , author=. arXiv preprint arXiv:1909.07707 , year=

    Pith/arXiv arXiv 1909

  62. [62]

    2021 IEEE International Conference on Robotics and Biomimetics (ROBIO) , pages=

    A survey on deep-learning approaches for vehicle trajectory prediction in autonomous driving , author=. 2021 IEEE International Conference on Robotics and Biomimetics (ROBIO) , pages=. 2021 , organization=

    2021

  63. [63]

    Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering , volume=

    Driver intention prediction using model-added Bayesian network , author=. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering , volume=. 2021 , publisher=

    2021

  64. [64]

    2013 IEEE/RSJ international conference on intelligent robots and systems , pages=

    Vehicle trajectory prediction based on motion model and maneuver recognition , author=. 2013 IEEE/RSJ international conference on intelligent robots and systems , pages=. 2013 , organization=

    2013

  65. [65]

    IEEE Transactions on Control Systems Technology , volume=

    Vehicle dynamics and external disturbance estimation for vehicle path prediction , author=. IEEE Transactions on Control Systems Technology , volume=. 2000 , publisher=

    2000

  66. [66]

    Journal of mechanical science and technology , volume=

    An improved model-based predictive control of vehicle trajectory by using nonlinear function , author=. Journal of mechanical science and technology , volume=. 2009 , publisher=

    2009

  67. [67]

    2008 11th international conference on information fusion , pages=

    Comparison and evaluation of advanced motion models for vehicle tracking , author=. 2008 11th international conference on information fusion , pages=. 2008 , organization=

    2008

  68. [68]

    IEEE Transactions on Intelligent Transportation Systems , volume=

    Model-based threat assessment for avoiding arbitrary vehicle collisions , author=. IEEE Transactions on Intelligent Transportation Systems , volume=. 2010 , publisher=

    2010

  69. [69]

    2022 22nd International Conference on Control, Automation and Systems (ICCAS) , pages=

    Inference of Vehicle Lane Change Intention Using Multiple Model Estimator in Automated Highway Driving , author=. 2022 22nd International Conference on Control, Automation and Systems (ICCAS) , pages=. 2022 , organization=

    2022

  70. [70]

    IET Radar, Sonar & Navigation , volume=

    Switched Kalman filter-interacting multiple model algorithm based on optimal autoregressive model for manoeuvring target tracking , author=. IET Radar, Sonar & Navigation , volume=. 2015 , publisher=

    2015

  71. [71]

    Sensors , volume=

    An improved interacting multiple model filtering algorithm based on the cubature Kalman filter for maneuvering target tracking , author=. Sensors , volume=. 2016 , publisher=

    2016

  72. [72]

    International journal of vehicle safety , volume=

    Trajectory predictions of lane changing vehicles using SVM , author=. International journal of vehicle safety , volume=. 2011 , publisher=

    2011

  73. [73]

    Computational Intelligence and Neuroscience , volume=

    Support vector machine based lane-changing behavior recognition and lateral trajectory prediction , author=. Computational Intelligence and Neuroscience , volume=. 2022 , publisher=

    2022

  74. [74]

    2020 39th Chinese Control Conference (CCC) , pages=

    Support vector machine based model predictive control for vehicle path tracking control , author=. 2020 39th Chinese Control Conference (CCC) , pages=. 2020 , organization=

    2020

  75. [75]

    2019 10th IEEE International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications (IDAACS) , volume=

    Prediction of surrounding vehicles lane change intention using machine learning , author=. 2019 10th IEEE International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications (IDAACS) , volume=. 2019 , organization=

    2019

  76. [76]

    2016 IEEE International Conference on Advanced Intelligent Mechatronics (AIM) , pages=

    Lane changing prediction at highway lane drops using support vector machine and artificial neural network classifiers , author=. 2016 IEEE International Conference on Advanced Intelligent Mechatronics (AIM) , pages=. 2016 , organization=

    2016

  77. [77]

    2017 IEEE 20th International Conference on Intelligent Transportation Systems (ITSC) , pages=

    Vehicle trajectory and lane change prediction using ANN and SVM classifiers , author=. 2017 IEEE 20th International Conference on Intelligent Transportation Systems (ITSC) , pages=. 2017 , organization=

    2017

  78. [78]

    2015 IEEE International Conference on Cyber Technology in Automation, Control, and Intelligent Systems (CYBER) , pages=

    A Hidden Markov Model based driver intention prediction system , author=. 2015 IEEE International Conference on Cyber Technology in Automation, Control, and Intelligent Systems (CYBER) , pages=. 2015 , organization=

    2015

  79. [79]

    2009 IEEE international conference on vehicular electronics and safety (ICVES) , pages=

    Driver intention inference with vehicle onboard sensors , author=. 2009 IEEE international conference on vehicular electronics and safety (ICVES) , pages=. 2009 , organization=

    2009

  80. [80]

    IEEE transactions on vehicular technology , volume=

    Early recognition of driving intention for lane change based on recurrent hidden semi-Markov model , author=. IEEE transactions on vehicular technology , volume=. 2020 , publisher=

    2020

Showing first 80 references.