Lane change trajectory prediction considering driving style uncertainty for autonomous vehicles

摘要

Lane change trajectory prediction is crucial for autonomous vehicles (AVs) to assess their owndriving safety in advance. However, there are significant uncertainties in the implementationof such prediction, including different behaviors caused by agent–agent interaction and drivingstyles of drivers. While prototype trajectories can serve as a means to represent typical motionpatterns and enhance trajectory prediction performance, their utilization in modeling motionpatterns tends to overlook the influence of agent–agent interactions and vehicle dynamics. Thispaper proposes a fusion algorithm that considers driving style and vehicle dynamics to addressthese uncertainties. The algorithm involves a long short-term memory (LSTM) lane changebehavior recognition model that mines key features of agent–agent interaction through theattention mechanism. The Gaussian process (GP) motion modeling trajectory prediction (GPMMTP)algorithm considers the driving style of the prototype trajectory based on the behaviorrecognition results. To further improve short-term and long-term prediction, the interactivemulti-model (IMM) algorithm is used to assign probability weights to the GP model and theExtended Kalman Filter (EKF) model based on prediction accuracy, taking into account thedriving styles and the vehicle dynamics. The proposed algorithm provides a promising approachto improving the accuracy of lane change trajectory prediction for AVs, and its effectiveness isdemonstrated using the HighD dataset.