Driver Lane Changing Intention Recognition Based on Multi-class Information

In this paper, we will combine the driving behavior, road environment and self-vehicle status information, select some of the information as the observation states to establish the driver lane changing intention recognition model. Wherein the driving behavior includes steeling, braking, and acceleration information. The road environment information contain the relative safety distance and the relative velocity, the vehicle status information includes vehicle lateral position, vehicle speed, and vehicle lateral acceleration. Here we proposed the constraint effect of the vehicle lateral position on the lane changing directions, and take it as one of the observing states of the model. By analyzing the relationship between steering action and vehicle speed in lane changing process, realize the driver's steering state transformation based on self-driving speed. It is effective to avoid the phenomenon of low recognition rate at the roundabout when only consider the lateral acceleration and the steering as the lateral observation states. We will using the Caretaker simulation platform to design simulated driving test, complete the data acquisition and model validation experimental design work.