A hybrid model for lane change prediction with V2X-based driver assistance

This paper proposes a hybrid model for lane-change prediction, with V2X-based driver assistance (HM4LCP) for improving driving safety. We carried out experiments in the testbed of Chang’an University to investigate driving behaviors on an urban road. 412 lane-changing units and 824 lane-keeping units were obtained from the vehicle trajectories. The HM4LCP integrates the Gaussian Mixture Model and the Continuous Hidden Markov Model to identify and predict lane-changing behaviors within a V2X environment. We collected 412 lane-changing samples and randomly chose 140, in order to test the HM4LCP’s identification accuracy. The results show that the HM4LCP results in an accuracy as high as 93.6% for lane-changing and 90.4% for lane-keeping, respectively. The proposed HM4LCP method can be applied to the driving assistant systems for lane-change warnings, in the initial stage. •Driving experiments are carried out to investigate lane-changing behavior.•We obtained and analyzed the operating characteristics of vehicles during the test.•Three feature vectors are chosen to describe the lane-changing behavior.•A hybrid model is proposed to recognize the lane-changing behaviors.