Method for Improving the Accuracy of Vehicle Behavior Prediction Based on Scenario Information

Abstract

The present invention discloses a method for improving the accuracy of vehicle behavior prediction based on scenario information, including: obtaining a preliminary predicted trajectory of a vehicle in a scenario according to a pre-trained behavior model; generating a vehicle interaction graph based on the spatio-temporal closest distance between vehicles in the preliminary predicted trajectory, which is used to represent whether there is an interaction relationship between different vehicles; at each predicted time step according to the preliminary predicted trajectory, obtaining lane information around the vehicle based on the position of the vehicle, and generating a vehicle-lane interaction graph, which is used to represent whether there is an interaction relationship between the vehicle and the lane; constructing a vehicle hidden space coding sequence with information interaction between vehicles, information interaction between the vehicle and the lane, and vehicle behavior decision interaction according to the vehicle interaction graph and the vehicle-lane interaction graph, and decoding the hidden space coding sequence to obtain a predicted trajectory. The method disclosed by the present invention improves the accuracy of vehicle behavior prediction.

Images