The invention relates to a pedestrian trajectory prediction method based on a generative adversarial network and an attention mechanism, and the method mainly comprises the steps: collecting pedestrian trajectory data through a device, constructing a data set, and carrying out the data preprocessing, and extracting trajectory features; the method comprises the following steps: modeling a pedestrian historical track and a current position by means of a long short-term memory (LSTM) network, adding a pedestrian interaction pooling module in a generator of a generative adversarial network, considering interaction information between pedestrians, and selecting pedestrian information which is more useful for measured data through the pedestrian interaction pooling module to realize prediction of the pedestrian track; wherein a pedestrian speed attention module is added in the generator, on one hand, a generated pedestrian prediction track can be real, and on the other hand, the defect that pedestrian speed attention is not considered in an encoder is made up; the discriminator extracts input features through a plurality of full connection layers, and memorizes the features of historical tracks through the LSTM network; and finally, performing pedestrian trajectory prediction by using the trained model.

    本发明涉及一种基于生成对抗网络和注意力机制的行人轨迹预测方法,该方法主要包括:使用设备采集行人轨迹数据构建数据集,数据预处理提取轨迹特征;借助长短期记忆网络LSTM对行人历史轨迹和当前位置建模,在生成对抗网络的生成器中加入行人交互池化模块,考虑行人之间的交互信息,并对被测数据更有用的行人信息经过行人交互池化模块进行选择,实现对行人轨迹的预测;其中,在生成器中加入行人速度注意力模块,一方面能够使生成的行人预测轨迹真实,另一方面弥补在编码器中没有考虑行人速度注意力;而鉴别器则通过数个全连接层进行输入特征的提取,同样通过LSTM网络进行历史轨迹的特征记忆;最后利用训练好的模型进行行人轨迹预测。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Pedestrian trajectory prediction method based on generative adversarial network and attention mechanism


    Weitere Titelangaben:

    一种基于生成对抗网络和注意力机制的行人轨迹预测方法


    Beteiligte:
    GAO JIANPING (Autor:in) / DU GUOGUO (Autor:in) / LIU PAN (Autor:in) / CHANG YONGQIANG (Autor:in) / YANG YIMING (Autor:in) / SU ZHIJUN (Autor:in)

    Erscheinungsdatum :

    2023-12-12


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    B60W CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION , Gemeinsame Steuerung oder Regelung von Fahrzeug-Unteraggregaten verschiedenen Typs oder verschiedener Funktion / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / G06V



    DSA-GAN: Driving Style Attention Generative Adversarial Network for Vehicle Trajectory Prediction

    Choi, Seungwon / Kweon, Nahyun / Yang, Chanuk et al. | IEEE | 2021



    SATP-GAN: self-attention based generative adversarial network for traffic flow prediction

    Zhang, Liang / Wu, Jianqing / Shen, Jun et al. | Taylor & Francis Verlag | 2021


    Long-term 4D trajectory prediction using generative adversarial networks

    Wu, Xiping / Yang, Hongyu / Chen, Hu et al. | Elsevier | 2022