The invention provides a traffic participant trajectory prediction method. The traffic participant trajectory prediction method comprises the steps of preprocessing a historical trajectory sequence to obtain a space-time category diagram; encoding the time sequence of the to-be-predicted pedestrian to obtain a time sequence hidden state of the pedestrian; learning the interaction influence of different types of agents on pedestrians from the space-time type graph; learning interaction weight information of different agents in the surrounding field of the to-be-predicted pedestrian; splicing and mapping the above three feature vectors into a hidden vector of a specific dimension; and coding the hidden vector, splicing the obtained coding vector and Gaussian noise, and decoding the obtained spliced vector into a Gaussian distribution parameter vector of a future trajectory as the output of trajectory prediction. According to the method, not only time and space interaction characteristics of the traffic intersection intelligent agent are considered, but also state category information of the traffic intersection intelligent agent is fully utilized, so that the prediction accuracy on an experimental data set is greatly improved, and good stability is shown.

    本发明提供了一种交通参与者的轨迹预测方法,包括:对历史轨迹序列进行预处理得到时空类别图;对待预测行人的时间序列进行编码得到行人的时间序列隐藏状态;从时空类别图中学习不同类别智能体对行人的交互影响力;学习待预测行人周边领域的不同智能体的交互权重信息;将以上三种特征向量进行拼接映射成特定维度的隐藏向量;将隐藏向量进行编码,并将得到的编码向量与高斯噪声进行拼接,并将得到的拼接向量解码成未来轨迹的高斯分布参数向量,作为轨迹预测的输出。本发明不仅仅考虑了交通路口智能体时间和空间交互特征,还充分利用了交通路口智能体的状态类别信息,从而在实验数据集上的预测准确度得到较大的提升,表现出了良好的稳定性。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Trajectory prediction method and system for traffic participants and storage medium


    Weitere Titelangaben:

    一种交通参与者的轨迹预测方法、系统及存储介质


    Beteiligte:
    LUO DENING (Autor:in) / QUAN XUEFENG (Autor:in) / ZHANG GEXIANG (Autor:in) / YANG QIANG (Autor:in)

    Erscheinungsdatum :

    01.03.2024


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    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 / G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung



    METHOD FOR LONG-TERM TRAJECTORY PREDICTION OF TRAFFIC PARTICIPANTS

    VELKEY GEZA / KIS KORNEL ISTVAN / KOROSI-SZABO PETER | Europäisches Patentamt | 2021

    Freier Zugriff

    Social force model for trajectory prediction of other traffic participants

    THIEM CHRISTIAN / WEBER NORBERT / EBERLE ULRICH | Europäisches Patentamt | 2024

    Freier Zugriff

    COMPUTATIONALLY EFFICIENT TRAJECTORY REPRESENTATION FOR TRAFFIC PARTICIPANTS

    REICHARDT JÖRG | Europäisches Patentamt | 2023

    Freier Zugriff

    COMPUTATIONALLY EFFICIENT TRAJECTORY REPRESENTATION FOR TRAFFIC PARTICIPANTS

    REICHARDT JÖRG | Europäisches Patentamt | 2023

    Freier Zugriff

    Computationally Efficient Trajectory Representation for Traffic Participants

    REICHARDT JOERG | Europäisches Patentamt | 2023

    Freier Zugriff