👋 你好!我是胡俊杰,一名 中南大学交通运输工程学院的硕士学生。我的导师是 Jaeyoung Jay Lee (李载宁) 教授,他是道路安全领域全球排名前2%的顶尖科学家。我致力于通过分析交通时空数据来揭示其内在规律,为构建更安全、更智能和更可持续的交通系统铺平道路。
Hu, J., Hu, C., Yang, J., Bai, J., & Lee, J. J. (2024). Do traffic flow states follow Markov properties? A high-order spatiotemporal traffic state reconstruction approach for traffic prediction and imputation. Chaos, Solitons, and Fractals, 183, 114965. [链接]
设计了一种高阶交通状态重构方法,证明了交通流的马尔可夫特性。利用马尔可夫矩阵的先验知识,在自编码器框架下完成了高精度的交通插补任务。
Hu, J., Bai, J., Yang, J., & Lee, J. (2024). Crash Risk Prediction Using Sparse Collision Data: Causal Inference and Graph Convolutional Networks Approaches. Expert Systems with Applications, 248, 125315. [链接]
创建了一种利用历史先验时空知识的碰撞数据增强策略。开发了一种基于因果推断和图卷积网络的新方法(PST-CGCN),用于预测碰撞风险。
Hu, J., Zhang, J., Bai, J., & Lee, J. Dynamic Correlation Analysis of Urban Crashes Using Tucker-Net Based SIRS Model: A Case Study in New York City. Journal of the Franklin Institute, 107946. [链接]
设计了一种新的数据驱动且可迁移的分析模型——基于Tucker网络的SIRS模型(TNBSM),该模型利用张量分解和SIRS模型来探索碰撞区域内的动态相关性。
Hu, J., & Lee, J. (2025). Car following dynamics in mixed traffic flow of autonomous and human-driven vehicles: complex networks approach. Physica A: Statistical Mechanics and its Applications, 665, 130519. [链接]
应用粗粒化相空间算法,将复杂网络技术引入车辆跟驰行为分析,揭示了自动驾驶车辆(AV)与人类驾驶车辆(HV)之间的显著特征和差异。
Hu, J., Gao, D., Lee, J., & Wang, L. Vehicle dynamics analytics based on complex network techniques: a trajectory-based visibility graph approach. Chaos, Solitons, and Fractals, 200, 117161. [链接]
引入了一种自适应的基于轨迹的视图(TVG)框架,这是一种通过将平面轨迹数据转换为复杂网络来剖析车辆动力学的新方法。该框架具有一个可调的可见性容差系数,通过横向位移动态缩放,使TVG能够捕捉几何遮挡和特定机动的空间尺度。
Hu, J., Bai, J., & Lee, J. Simplified and Efficient KNN-Based Method for High-Resolution Traffic Time-Space Diagram Imputation. (审稿中).
提出了一个改进的K近邻(KNN)框架,通过匹配由后邻接空间规则定义的相似时空邻域特征,从稀疏数据中重建和插补时空图(TSD)。
Hu, J., Gao, D., Hu, C., Zhou, H., & Lee, J. Rethinking driving style recognition: A prediction error-based driving behavior modeling. (审稿中).
通过探索轨迹预测模型的误差来研究驾驶员行为的异质性,重新审视了端到端的驾驶风格识别方法。利用空间注意力和卷积社交池化来学习车辆运动的相互依赖关系,并引入了基于驾驶风格的未来轨迹多模态分布。
Hu, J., Lee, J., & Wang, L. Re-examining the Explanatory Boundaries of Car-Following Models: From a Systematic Decomposition of Fitting Errors to the Revelation of Adaptive Feedback Mechanisms. (审稿中).
提出车辆跟驰模型的残差是结构误差和随机误差的复合体,并利用非观测成分模型对三个已标定跟驰模型的残差序列进行分解。
Hu, J., Bai, J., &Lee, J. Meta-Learning Guided by Physics-Based Residual Priors: A Plug-and-Play Training Strategy for Lightweight Car following Models. (审稿中).
设计一种名为残差感知元学习重加权(RAMR)的跟车模型即插即用框架,该框架将对应训练问题重新定义为双层优化任务,并在4个数据集,3个物理学模型以及3个预测组件上进行全面验证。
Wang, L., Lee, J.& Hu, J *. The Resilience Benefit of Disorder: A Global Typological Analysis of Road Network Morphology and Urban Performance. (审稿中).
使用包含70个城市的全球数据集,协助研究了由方向熵量化的道路网络形态如何影响城市可持续性与韧性的关键指标,包括二氧化碳排放、通勤时间和拥堵状况。
Yang, J., Lee, J., Mao, S., & Hu, J. (2024). Dynamic safety estimation of airport pick-up area based on video trajectory data. IEEE Transactions on Intelligent Transportation Systems, 25(2), 1774–1786. [链接]
协助设计了一种改进的CUSBoost算法,用于不平衡轨迹数据的分类,该算法利用了四种风险指标和空间分布分析。
Wang, L., Lee, J., Hu, J., Yang, Y., & Mao, S. Analysis of injury severity of single-vehicle and two-vehicle crashes with lightweight vehicles (kei cars) in Japan: A random parameters approach with heterogeneity in means. (被Traffic Injury Prevention接受).
使用均值异质性的随机参数Probit模型,协助研究了日本K-car(轻型汽车)的碰撞伤害严重程度。设计了一种样本外预测方法,以阐明各种碰撞严重性机制之间的异质性。
Wang, L., Lee, J.,Hu, J., & Mao, S. Contributing Factors to the Severity of Crash Injury and Vehicle Damage Involving Japanese Lightweight K-cars: Considering Unobserved Heterogeneity. (审稿中).
协助模型设计。
Hu, C., Tang, J.,Hu, J., Wang, Y.,Li, Z., Zeng, J., &Han, C. Dynamic partitioning of heterogeneously loaded road networks: A two-level regionalization scheme with Monte Carlo tree search. Transportation Research Part C: Emerging Technologies, 180, 105341. [链接]
协助模型设计。