| 49 | 0 | 62 |
| 下载次数 | 被引频次 | 阅读次数 |
【背景】自动驾驶系统在轨迹预测精度方面已取得显著进展，但其决策过程缺乏可解释性，导致算法可信性与公众接受度不高。【目标】以驾驶员的内异质性为切入点，刻画随时空变化的动态驾驶行为模式，对不同驾驶行为模式下个体差异与环境交互的融合过程进行显式建模，在兼顾预测精度的同时提高模型的可解释性。【方法】提出一种融合个体行为模式与多车交互环境的轨迹预测方法，通过驾驶行为波动特征动态识别个体行为模式，并设计个体行为模式–多车交互环境作用分配机制，表征二者对未来轨迹的相对影响。【数据】采用INTERACTION数据集中高速公路入口匝道合流区域的车辆轨迹数据。【结果】同一车辆在驾驶过程中的驾驶行为模式呈现动态演化特征，考虑该特征的轨迹预测模型在各预测时域下的平均位移误差和最终位移误差均优于对比模型，误差降低幅度为3.8% ～ 30.8%。同时，不同驾驶行为模式下个体行为特征与多车交互特征的作用分配结果呈现出显著差异，反映出车辆运动过程中对两类信息依赖程度的变化特征。【结论】动态驾驶行为模式的引入，使模型具备更强的轨迹差异化表征能力，从而提升轨迹预测精度。随着驾驶行为模式由保守型向激进型转变，个体行为特征的贡献逐渐增加，多车交互特征的贡献逐渐降低，为理解模型预测过程提供可解释依据。
Abstract：[Background] Significant progress has been made in trajectory prediction accuracy for autonomous driving systems， but the lack of interpretability in their decision-making processes results in low algorithm trustworthiness and public acceptance. [Objective] Taking intra-driver heterogeneity as the starting point， dynamic driving behavior patterns varying across time and space are characterized， and the fusion process between individual differences and environmental interactions under different driving behavior patterns is explicitly modeled， so as to improve model interpretability while maintaining prediction accuracy. [Method] A trajectory prediction method integrating individual driving behavior patterns and multi-vehicle interaction environments is proposed. Individual driving behavior patterns are dynamically identified through driving behavior fluctuation features， and an individual behavior pattern-multi-vehicle interaction allocation mechanism is designed to characterize their relative influences on future trajectories. [Data] Vehicle trajectory data from the highway entrance-ramp merging area of the INTERACTION dataset are used. [Result] The driving behavior patterns of the same vehicle exhibit dynamic evolution during the driving process， and the trajectory prediction model that considers this characteristic achieves lower Average Displacement Error (ADE) and Final Displacement Error (FDE) than the baseline models across all prediction horizons， with error reductions ranging from 3.8% to 30.8%. Meanwhile， the allocation results of individual behavioral features and multi-vehicle interaction features exhibit significant differences under different driving behavior patterns， reflecting the varying dependence of vehicle motion on these two types of information. [Conclusion] Considering the dynamic variation of driving behavior patterns can effectively improve trajectory prediction accuracy and enhance the model's ability to characterize differences in vehicle motion. As driving behavior patterns evolve from conservative to aggressive， the contribution of individual behavioral features gradually increases， whereas the contribution of multi-vehicle interaction features gradually decreases， thereby providing an interpretable basis for understanding the trajectory prediction process.
[1] Yan Y， Zhang P， Du C， et al. Vehicle trajectory prediction for autonomous driving applications: state-of-the-art review， research challenges， and future directions[J]. Automotive Innovation， 2026， 9(2): 263-295.
[2] 章程， 赵靖， 杨晓光， 等. 城市快速路入口匝道控制研究综述[J]. 上海理工大学学报， 2023， 45(4): 332-344.
[3] Gomes I P， Wolf D F. A comprehensive review of deep learning techniques for interaction-aware trajectory prediction in urban autonomous driving[J]. Neurocomputing， 2025， 651: 131014.
[4] Rasouli A， Tsotsos J K. Autonomous vehicles that interact with pedestrians: a survey of theory and practice[J]. IEEE Transactions on Intelligent Transportation Systems， 2020， 21(3): 900-918.
[5] Heiden E， Palmieri L， Bruns L， et al. Bench-MR: a motion planning benchmark for wheeled mobile robots[J]. IEEE Robotics and Automation Letters， 2021， 6(3): 4536-4543.
[6] Gupta A， Johnson J， Li F F， et al. Social GAN: socially acceptable trajectories with generative adversarial networks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City， USA: IEEE， 2018: 2255-2264.
[7] Xu R， Li J， Zhang S， et al. Interactive trajectory prediction for autonomous driving based on Transformer[J]. Mechanical Sciences， 2025， 16(1): 87-97.
[8] 曹越， 上官伟， VISSER Arnoud， 等. 基于模式匹配注意力机制的车辆轨迹预测[J]. 中国公路学报， 2025， 38(10): 386-401.
[9] Mohamed A， Qian K， Elhoseiny M， et al. Social-STGCNN: a social spatio-temporal graph convolutional neural network for human trajectory prediction[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle：IEEE， 2020: 14412-14420.
[10] 徐东伟， 程千兵， 顾同成， 等. 基于时空交互特征融合的车辆轨迹预测方法研究[J]. 中国公路学报， 2025， 38(11): 447-458.
[11] Wang W， Xi J， Zhao D. Driving style analysis using primitive driving patterns with Bayesian nonparametric approaches[J]. IEEE Transactions on Intelligent Transportation Systems， 2019， 20(8): 2986-2998.
[12] Su J， Li M， Zhu L， et al. Vehicle trajectory-prediction method based on driver behavior-classification and informer models[J]. World Electric Vehicle Journal， 2024， 15(6): 237.
[13] Li X， Zhang M， Tavares A J， et al. Personalized vehicle trajectory prediction method based on driving style classification[J]. Scientific Reports， 2025， 15: 31587.
[14] 殷宇婷， 赵靖. 面向交叉口通行轨迹离散性的交通流模型参数标定方法[J]. 铁道科学与工程学报， 2022， 19(9): 2563-2573.
[15] Mobini Seraji M H， Shaffiee Haghshenas S， Shaffiee Haghshenas S， et al. A state-of-the-art review on machine learning techniques for driving behavior analysis: clustering and classification approaches[J]. Complex & Intelligent Systems， 2025， 11(9): 386.
[16] 吴红兰， 胡德富， 郭旭周. 基于Informer的车辆多意图运动轨迹预测[J]. 交通运输工程与信息学报， 2024， 22(3): 68-79.
[17] 周亦威， 劳逸男， 赵靖， 等. 考虑多类型交通参与者的图遍历多模态车辆轨迹预测[J]. 公路交通科技， 2025， 42(3): 1-10.
[18] Zhang K P， Li L. Explainable multimodal trajectory prediction using attention models[J]. Transportation Research Part C: Emerging Technologies， 2022， 143: 103829.
[19] Zhao J W， Qu T， Gong X， et al. Interaction-aware personalized trajectory prediction for traffic participant based on interactive multiple model[J]. IEEE Transactions on Intelligent Vehicles， 2023， 8(3): 2184-2196.
[20] Wang C M， Jia D Y， Zheng Z D， et al. Segment-then-refine: a general calibration framework incorporating intra-driver heterogeneity into Car-Following Models[J]. Transportation Research Part C: Emerging Technologies， 2025， 176: 105144.
[21] Zhan W， Sun L T， Wang D， et al. INTERACTION dataset: an INTERnational， Adversarial and Cooperative moTION dataset in interactive driving scenarios with semantic maps [PP/OL]. [PP/OL]. V1(2019-09-30)[2026-10-12]. https://doi.org/10.48550/arXiv.1910.03088.
[22] Mohammadnazar A， Arvin R， Khattak A J. Classifying travelers' driving style using basic safety messages generated by connected vehicles: Application of unsupervised machine learning[J]. Transportation Research Part C: Emerging Technologies， 2021， 122: 102917.
[23] Zhang Y， Zou Y J， Xie Y C， et al. Identifying dynamic interaction patterns in mandatory and discretionary lane changes using graph structure[J]. Computer-Aided Civil and Infrastructure Engineering， 2024， 39(5): 638-655.
基本信息:
DOI：10.19961/j.cnki.1672-4747.2026.07.019
中图分类号:U463.6
引用信息:
[1]赵靖,胡浩,张越,等.驾驶行为模式引导下的可解释多车交互轨迹预测[J].交通运输工程与信息学报().DOI:10.19961/j.cnki.1672-4747.2026.07.019.
基金信息:
国家重点研发计划项目(2025YFE0124200); 国家自然科学基金项目(52372304); 上海市白玉兰人才计划浦江项目(24PJC062); 上海市科技攻关计划项目(25GA3200200)
2026-10-16
2026-10-16
2026-10-16