The invention discloses a deep learning vehicle trajectory prediction method fused with a kinematic model. The method comprises the following steps: acquiring surrounding scene information of an automatic driving vehicle and converting the surrounding scene information to a target vehicle coordinate system; using a deep learning model to establish vehicle historical motion state information and an interaction relationship, and predicting a centroid acceleration and a front wheel turning angle of the target vehicle; vehicle types are divided according to vehicle lengths, predicted values of vehicle centroid acceleration and front wheel turning angles are restrained, and trajectory coordinates of a target vehicle are calculated from the predicted values by using a vehicle two-degree-of-freedom model. According to the method, the vehicle two-degree-of-freedom motion model is fused into the deep learning prediction model, the high-precision long-time-domain prediction performance of deep learning is considered, and the kinematic property of the vehicle is considered, so that the predicted trajectory result better conforms to the real vehicle motion law. According to the method, the feasibility and accuracy of the prediction trajectory can be improved, the situation that main vehicle planning is too conservative due to an infeasible prediction result can be reduced, and the driving safety and smoothness of the automatic driving vehicle are improved.

    本发明公开了一种融合运动学模型的深度学习车辆轨迹预测方法,包括如下步骤:获取自动驾驶车辆周围场景信息并转换至目标车辆坐标系;使用深度学习模型建立车辆历史运动状态信息及交互关系,预测目标车辆的质心加速度和前轮转角;根据车辆长度划分车辆类型并约束车辆质心加速度、前轮转角的预测值,使用车辆二自由度模型由预测值计算出目标车辆的轨迹坐标。本发明将车辆二自由度运动模型融合进深度学习预测模型,兼顾深度学习的高精度长时域预测性能并考虑车辆运动学性质,使预测轨迹结果更符合真实车辆运动规律。本发明能提升预测轨迹的可行性和准确性,能够减少不可行预测结果导致主车规划过于保守的情况,提升自动驾驶汽车行驶安全性、平顺性。


    Access

    Download


    Export, share and cite



    Title :

    Deep learning vehicle trajectory prediction method fused with kinematic model


    Additional title:

    一种融合运动学模型的深度学习车辆轨迹预测方法


    Contributors:
    LI LINHUI (author) / FU YIFAN (author) / WANG XUECHENG (author) / LIAN JING (author) / WANG TING (author) / ZHAO JIAN (author) / WANG YAFEI (author)

    Publication date :

    2023-07-14


    Type of media :

    Patent


    Type of material :

    Electronic Resource


    Language :

    Chinese


    Classification :

    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 / G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen



    Compressing Vehicle Trajectory Data Using Hybrid Coding With Kinematic Motion Prediction

    Xu, Lipeng / Wu, Zhizhou / Wang, Yinhai et al. | IEEE | 2024


    Trajectory optimization method based on vehicle kinematic model predictive control

    WU WENJIAN / PENG YIFAN / LU YISHEN | European Patent Office | 2022

    Free access

    Deep learning-based vehicle trajectory prediction device and method therefor

    LIM SEJOON / OH GEESUNG / KIM HEEJUNG | European Patent Office | 2024

    Free access

    Parallel parking trajectory planning method based on vehicle kinematic model

    LEI CHAO | European Patent Office | 2021

    Free access

    DEEP LEARNING-BASED VEHICLE TRAJECTORY PREDICTION DEVICE AND METHOD THEREFOR

    LIM SEJOON / OH GEESUNG / KIM HEEJUNG | European Patent Office | 2023

    Free access