According to the heterogeneous traffic participant trajectory prediction method, in the construction process of a graph, an edge selection mechanism based on a fixed distance is abandoned for edge construction, more domain knowledge including endpoint types, interaction relations and relative orientation relations of targets is considered, and information contained in the graph is enriched; meanwhile, factors, including motion features and environment features, influencing a trajectory prediction result are converted into the same data form and feature type, so that heterogeneous points and edges are converted into a unified form; finally, the motion features, the interaction features among the various traffic participants and the interaction features of the road are fused together and transmitted to a multi-layer perceptron decoder, and multi-modal trajectory prediction is generated. Compared with an existing common graph neural network model, the method has the advantages that the prediction accuracy is greatly improved, and particularly for a multi-target scene in a complex and dense interaction scene, the accurate and stable prediction capability can be shown.

    一种异质交通参与者轨迹预测的方法,该方法在图的构建过程中,对于边的构建摒弃了基于固定距离的边选取机制,考虑了更多的“领域知识”,包括端点类型、交互关系、目标的相对方位关系,丰富了图包含的信息;同时,将影响轨迹预测结果的因素,包括运动特征和环境特征,转化成相同的数据形式和特征类型,从而将异质的点和边转化成统一的形式;最终,各类交通参与者的运动特征、相互间的交互特征、与道路的交互特征被融合在一起,传送到多层感知机解码器中,生成多模态的轨迹预测。和现有的普通图神经网络模型相比,在预测准确率上得到很大提升,尤其对于复杂密集交互场景中多目标场景,能够表现出准确且稳定的预测能力。


    Access

    Download


    Export, share and cite



    Title :

    Method for predicting tracks of heterogeneous traffic participants


    Additional title:

    一种异质交通参与者轨迹预测的方法


    Contributors:
    FU MENGYIN (author) / ZHANG TING (author) / SONG WENJIE (author) / YANG YI (author) / WANG MEILING (author)

    Publication date :

    2023-09-15


    Type of media :

    Patent


    Type of material :

    Electronic Resource


    Language :

    Chinese


    Classification :

    IPC:    G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS



    Predicting Parameters for Modeling Traffic Participants

    Moradipari, Ahmadreza / Bae, Sangjae / Alizadeh, Mahnoosh et al. | IEEE | 2022




    Method for identifying traffic participants

    PFITZER MARTIN / HEINRICH STEPHAN / DUBOIS MATHIEU et al. | European Patent Office | 2022

    Free access