Research On Personalized Lane Keeping Assist Strategy For Human-machine Driving Vehicle

Constrained by technical bottlenecks and equipment costs,fully autonomous driving is difficult to achieve in the short term.The stage of human vehicle co-piloting will continue for a long time in the future.Under the mode of human vehicle co-piloting,lane keeping assistance system can effectively reduce traffic accidents and alleviate the driver’s operating load.But in the complex traffic environment,due to the diverse driving habits of drivers,the conventional lane keeping assistance system always drives along the center line of the lane,which is obviously different from the real human driving experience.For example,in the scene of curve,some drivers will prefer to drive along the inner side of the curve,otherwise it will produce a sense of insecurity and cause the driver’s mental load to increase.Therefore,this paper studies the driving habits of drivers,and constructs a lane keeping assistance strategy that considers human driving experience in real time.It is beneficial to improve driving safety and improve driving comfort,and is the basis of realizing personalized advanced driving assistance.Aiming at the problem that the uniform design of the auxiliary system is difficult to meet the driver’s differentiated needs.This paper studies the real-time personalized lane keeping assistance strategy from the two perspectives of shared driving and switched driving under the human vehicle co-piloting mode.Based on the driving simulator data and smart watch sensor data,the driving habit identification model is established;In order to give full advantages of driver and assistant system in decision-making planning and control execution,a shared lane keeping assistant strategy is studied based on the idea of human-machine division and cooperation.In addition,based on deep learning algorithm,a switching personalized lane keeping assistance strategy is proposed.The research work of this paper mainly includes the following aspects:(1)Based on driving simulator and a smart watch,a driving habit identification method based on multi-source data fusion is studied.The driving habits of drivers are affected by experience,and psychological factors.This paper collects the multi-dimensional information including vehicle motion state,driver operation data and driver physiological parameters during the process of driving.The driving habits of drivers are identified.The drivers are divided into three types: cautious drivers,normal drivers and aggressive drivers,The experimental results show that the method can effectively improve the identification rate of driver’s driving habits.(2)Based on the idea of human-machine division and cooperation,a lane keeping assistance strategy of human-machine sharing is studied.The lane keeping task is divided into two sub tasks.The driver adjusts the lateral position according to the real-time demand,the assistant system tracks the desired position accurately,and designs the lane keeping controller under the shared driving mode.Experiments show that this method can alleviate the driver’s operation load,and improve the safety,and meet the driver’s personalized demand in real time.(3)Based on the deep convolution fuzzy system,a lane keeping assistant strategy of human-machine switching is studied.The driver’s driving behavior is time-varying,and taking the lane center line as the ideal driving trajectory is contrary to the driving habits.In this paper,based on the deep convolution fuzzy system,we establish the simulation trajectory planning model of different driving habits,and update the model online to improve the generalization ability of the model,and realize the research on the switching personalized lane keeping assistance strategy.The experimental results show that the method can ensure the driving safety and meet the personalized needs of drivers in the process of lane keeping.