Signal control intersection electric vehicle trajectory prediction method and system

Abstract

The application discloses a signal control intersection electric vehicle trajectory prediction method and system, relates to the intelligent traffic and automatic driving technical field, and comprises the following steps: collecting signal control intersection vehicle video data; realizing accurate screening of electric vehicle samples based on license plate color difference; constructing a physical information neural network (PINN) fusing speed and acceleration physical constraints; through designing a composite loss function containing data fitting loss, acceleration regularization loss and acceleration change rate regularization loss, the predicted trajectory is forced to comply with the vehicle motion physical law; finally, the trained model is used to realize short-time trajectory prediction of the signal control intersection electric vehicle. The application improves the accuracy and physical consistency of the trajectory prediction, especially optimizes the longitudinal trajectory prediction effect, and can provide reliable support for automatic driving risk avoidance and vehicle-road cooperative control.

Images