The invention belongs to the field of vehicle trajectory prediction, and particularly relates to an external knowledge enhanced urban road vehicle trajectory prediction algorithm. Comprising the steps of collecting historical trajectory data of a vehicle and a pedestrian, performing feature extraction on the vehicle trajectory data according to the historical trajectory data of the vehicle and the pedestrian, designing a distance memory network model and correcting a vehicle trajectory. The designed distance memory network model comprises a distance attention mechanism and a distance correction mechanism, and pedestrian trajectory information is fused, so that the precision and reliability of trajectory prediction are remarkably improved. The new activation function design enables the model to be more stable when large-amplitude input is processed, the numerical problem caused by overlarge output is avoided, and the robustness of the model is enhanced. The design of the method depends on high-quality data acquisition, so that the model effect becomes better and better along with continuous enrichment of automatic driving data.
本发明属于车辆轨迹预测领域,尤其涉及一种外部知识增强城市道路车辆轨迹预测算法,包括收集车辆和行人的历史轨迹数据、根据车辆和行人的历史轨迹数据对车辆轨迹数据进行特征提取、设计距离记忆网络模型以及对车辆轨迹进行校正。本发明的设计距离记忆网络模型包括距离关注机制和距离修正机制两部分通过融合行人轨迹信息,显著提高了轨迹预测的精度和可靠性。新的激活函数设计使得模型在处理大幅度输入时更加稳定,避免了输出过大导致的数值问题,增强了模型的鲁棒性。本方法的设计依赖于高质量的数据采集,因此随着自动驾驶数据的不断丰富模型效果会越来越好。
Urban road vehicle trajectory prediction algorithm based on external knowledge enhancement
一种外部知识增强城市道路车辆轨迹预测算法
2024-12-17
Patent
Electronic Resource
Chinese
Vehicle Trajectory Prediction Based on Road Recognition
Trans Tech Publications | 2014
|Vehicle Trajectory Prediction by Knowledge-Driven LSTM Network in Urban Environments
DOAJ | 2020
|Vehicle collision early warning method based on vehicle-road cooperation trajectory prediction
European Patent Office | 2023
|Vehicle Lane-Changes Trajectory Prediction Model Considering External Parameters
BASE | 2021
|