Research On Lane Keep Assistant System Based On Genetic Algorithms Optimizing BP Neural Network

With the rapid development of the automotive industry,LKAS has been attracted more and more attention as an active safety technology for automobiles.The lane keeping assistant system based on BP neural network optimized by genetic algorithm is studied,and studies the related algorithms such as lane departure warning.The BP neural network lane keeping prediction controller based on genetic algorithm optimization is designed.Simulation test study was finished including departure warning.Firstly,based on two-point driver direction control model is used to estimate the lateral offset of the vehicle center of mass and the lane centerline through the near vision point.The far vision point is used to estimate the curvature of the road,and the relative position model of the vehicle and the road is established.At the same time,EPS is selected as the actuator for the active control of the lane,and the simulation model of EPS is built.Considering that frequent warning may cause driver discomfort,a driver operating state judgment model is established to ensure that the driver’s operation is the highest priority.For the lane departure warning decision,based on the TLC algorithm,the driver types are classified by studying the driving habits of different drivers.According to the driver type,the alarm threshold is designed,and the earliest warning boundary line is designed based on the vehicle deviation speed.Finally,a dynamic early warning boundary decision algorithm based on different driver types is designed.In order to realize the free switching of the vehicle in the active control mode and the coordinated control mode,the EPS-based coordinated control strategy and active control strategy are studied.Finally,a lane-preserving predictive controller based on genetic algorithm is proposed to optimize BP neural network,and the improved algorithm combining the additional momentum,adaptive learning rate,momentum and adaptive learning rate and the effect of genetic algorithm on network performance optimization are compared.Based on Carsim and Matlab/Simulink,the joint simulation test of lane departure warning and lane keeping is carried out respectively.The simulation conditions include straight road,small curvature radius road and large curvature radius road.The simulation test results show that when the vehicle deviates from the lane,the lane departure warning algorithm designed in this papercan issue alarms on both straight roads and corners.At the same time,under the driver’s inoperative awareness,the designed lane keeping predictive controller can control the vehicle to return to the vicinity of the lane centerline when the vehicle deviates.