Research On Lane-changing Trajectory Prediction Based On Driver's Driving Intention

In advanced driver assistance systems(ADAS),accurate vehicle lane change trajectory prediction is a key issue for risk assessment and danger warning.Lane-changing behavior is one of the most important and most general driving behavior in vehicle driving,and it is closely related to safety traffic.However,almost all advanced driver assistance systems focus on vehicles and ignore drivers themselves lane change intention.In view of these phenomena,this paper puts forward an approach for vehicle lane change trajectory prediction based on driver's lane change intention,in order to Increase the safety of vehicle driving.The main research contents are as follows:(1)Establish the database of lane change intention.A total of 1200 sets of lane changing data,including 900 sets of training samples,300 sets of test data.Build the Driver in loop simulation experiments,under the simulated road environment,the driver performs corresponding lane change behavior by handling accelerator pedal,brake pedal and steering wheel and other components.The driver's operation data is measured by various sensors and passed to the simulation experimental platform terminal.The abnormal experiment data is eliminated by t-test algorithm,and then the database of lane change intention is established.(2)Set up a hybrid model which combining the hidden Markov model(HMM)with the BP neural network(BP).This paper summarizes the advantages and disadvantages of various pattern recognition algorithms on the basis of research about driver's lane change intention identification.Finally,a HMM-BP hybrid model with time series and classification characteristics is proposed.Three HMM model parameters of left lane change,lane keeping and right lane change are obtained through parameter training by using the HMM module in Matlab simulation software.Then,the maximum likelihood estimation,which is outputted by each HMM model,is taken as the input of BP neural network.Then train the BP neural network to achieve the purpose of identifying the driver's lane change intention.The experimental results that the recognition rate of hybrid model is higher compared to the single model with HMM or BP neural network,and the recognition rate is up to 97.33%.(3)Predict lane change trajectory based on the driver's lane change intention.The model includes lane keeping trajectory prediction model and vehicle lane change trajectory prediction model.This paper mainly studies th e prediction of the lateral trajectory of vehicles,that is,the trajectory prediction of lane change.This paper proposed a lane change trajectory model based on trigonometric function,and the original model is improved according to the training sample d ata.The predicted trajectory according to the test sample data is compared with the actual trajectory,and it is found that the prediction trajectory is close to the actual trajectory.The result error is acceptable,the lateral error is not greater than 0.2m,and the maximum longitudinal error is not more than 2m,which verifies the validity of the proposed model.Finally,the safety analysis of the lane change trajectory is carried out to ensure the safety of the lane change process from trajectory curvature,lane change trajectory with current lane and lane change trajectory with target lane.