A comprehensive intention prediction method considering vehicle interaction

Abstract

In this paper, an interactive intention prediction method is proposed. Firstly, the Hidden Markov Model integrated with Gaussian Mixture Model is modeled for current behavior recognition and its parameters are trained through NGSIM dataset. Then, a trajectory prediction method based on Frenet frame is used to predict the future traffic situation, considering which future behavior reasoning is realized by maximum expected utility theory. The final intention prediction result is a combination of historical trajectory recognition and future behavior reasoning. The simulation results show that the proposed method has the ability of reasonably reflecting the interaction process between vehicles and the prediction performance is good.