Research On Driver-Adaptive Lane Keeping Assitance Control

The lane keeping assist system is essential to improve the lateral stability of the vehicle and reduce the burden of the driver,but the lane keeping assist system with fixed parameters is difficult to apply to different drivers with different driving habits,and it cannot be closed in time when the driver initiates a lane change because of the single lane change intention criterion.To address these problems,this paper proposes a lane keeping assistance control system that adapts to driver characteristics in terms of both behavioral habits and driving intentions.The lane departure decision model is designed based on the analysis of driver’s lane keeping behavior,and the lane change intention is identified by integrating driver’s operation behavior and vehicle motion parameters to provide support for system decision,and the lane departure intervention lane keeping master controller is established by using reinforcement learning theory.The details of the study are as follows.First,a driving simulation scenario is built by VI-grade,and driving data are collected by using the driving simulation system.On the basis of pre-processing the data,the parameters that reflect the driver’s perception of lane departure risk are analyzed and selected,a safe driving zone is established based on the driver’s lateral offset during lane keeping,and the latest warning timing is determined based on the driver’s cross-lane time at the steering return point and the vehicle dynamics limit,and a lane departure decision model based on the safe driving zone and the latest warning timing is established.Second,to avoid the conflict between control decision and driver intention,a driver intention recognition module based on LSTM is established.Based on the analysis of the feature variables in the driver lane change process,the variables among them that can generally reflect the driver lane change pattern are selected as the input of the LSTM network.The results show that this algorithm improves the accuracy by4% over the traditional SVM’s intention recognition algorithm.In order to further verify the effectiveness of the algorithm in the actual lane change process,the right lane change condition and continuous lane change condition were selected to verify the performance of the algorithm for continuous recognition,and the causes of misclassification were analyzed.Third,a lane keeping controller based on Deep Deterministic Policy Gradient(DDPG)is designed.The vehicle lateral position deviation and heading angle deviation,which can reflect the vehicle deviation state,are selected to generate the state space of the intelligent body,and the front wheel rotation angle is used as the intelligent body output to control the lateral motion of the vehicle.A reward function that takes into account safety,comfort and lateral stability is designed,and the training termination conditions are set according to the vehicle dynamics limits and deviation states,allowing the intelligent body to perform 20,000 rounds of learning in the training environment.The lane keeping capability of the DDPG intelligentsia is validated under straight,curved and double-shifted lane conditions.The results are compared with the model predictive control and show that the DDPG smartbody has better lateral stability while accomplishing the lane keeping task.Finally,the above modules are integrated to establish a deviation-intervention lane keeping assisted control system,and the driver-in-the-loop simulation experiments are conducted through a driving simulation system with the help of the virtual scenario and dynamics module of VI-grade and a system model based on Matlab/Simulink,focusing on verifying whether the lane keeping controller can control the vehicle in time when the heading is deviated.The results show that the system rarely misjudges and interferes with the driver’s behavior due to the lane departure decision model and the driver intention recognition module,and can take over the control of the vehicle in time when the driver is unconscious of the lane departure.