The invention relates to a deep learning car-following prediction method considering driver fuzzy perception. The method comprises the following steps of S1, training a deep learning car-following model used for predicting the speed of a vehicle according to car-following feature data; S2, acquiring real car-following feature time series data, and inputting the real car-following feature time series data into the deep learning car-following model to obtain vehicle speed prediction time series data; S3, extracting the real car-following feature time series data of N moments before the moment tobe predicted and the vehicle speed prediction time series data as a first data set and a second data set; S4, determining a fuzzy sensing time window, and replacing the vehicle speed in the fuzzy sensing time window in the first data set with the vehicle speed prediction value at the corresponding moment in the second data set to form prediction input data; and S5, inputting the prediction inputdata into the deep learning car-following model to obtain a vehicle speed prediction value at the to-be-predicted moment. Compared with the prior art, different fuzzy sensing time windows are set, andheterogeneous following behaviors can be generated for the same following scene.
本发明涉及一种考虑驾驶员模糊感知的深度学习跟驰预测方法,包括如下步骤:S1、训练用于根据跟驰特征数据预测本车速度深度学习跟驰模型;S2、获取真实跟驰特征时序数据,输入至深度学习跟驰模型得到本车速度预测时序数据;S3、提取待预测时刻前N个时刻的真实跟驰特征时序数据以及本车速度预测时序数据作为第一数据集和第二数据集;S4、确定模糊感知时间窗,将第一数据集中位于模糊感知时间窗内的本车速度替换为第二数据集中对应时刻的本车速度预测值形成预测输入数据;S5、将预测输入数据输入至深度学习跟驰模型得到待预测时刻的本车速度预测值。与现有技术相比,本发明设置不同的模糊感知时间窗可对同一跟驰场景生成异质跟驰行为。
Deep learning car-following prediction method considering driver fuzzy perception
一种考虑驾驶员模糊感知的深度学习跟驰预测方法
2021-01-08
Patent
Electronic Resource
Chinese
Vehicle following model stability control method considering driver perception error
European Patent Office | 2020
|Vehicle following system stability control method considering driver characteristics
European Patent Office | 2024
|