Research On Recognition Of Drivers’ Unintentional Lane Departure For Lane Departure Warning System

It is the lane departure of unconscious drivers that has become a main reason for thetraffic accidents. The radar and camera techniques have been widely used in the lanedeparture warning system to perceive the environment changes currently，and furthermore,the lane departure behaviors of unconscious drivers can be detected by steering wheel andturn signals in case that vehicles have a tendency to approach to the traffic line or deviatefrom the lane. While because of a serious fact of the low utilization of turn light and thecurrent deviation decision algorithm does not match the driver’s behavior in reality, there arestill some problems such as a high rate of false and omissions as well as a poor reliability inthe whole system. Thus, in order to enrich the judgment index of the lane departure ofunconscious drivers and avoid an sensitive system which caused by missing vehicle signals,an unintentional lane departure recognition method which combines the drivers’ operation,the vehicles motion and the relative motion between the vehicle and the traffic line has beenproposed.Referring to the researches on the unconscious lane departure of the domestic andforeign scholars’, an experiment about unconscious lane departure has been taken throughthe co-simulation platform based on the CarSim and LabVIEW software. In the experiment,twelve drivers with different genders, proficiency and driving styles are selected, relevantparameters that can reveal the drivers’ characteristics such as operation style, the motion ofvehicles and the relative motion between the vehicle and the traffic line are collected. Relyon these data, the intent time window of normal lane-changing and unconscious lanedeparture can finally be determined. Meanwhile, this paper proposed a method to optimizethe intent time window by ROC (Receiver Operating Character) system which analyzed atvarious times preceding the lane change maneuver. In the paper, a serious of statistical analysis about the characteristics of drivers, vehiclesand road in the unconscious lane departure time window has been taken. It shows that, firstly,the drivers’ ability to control vehicle is reduced and the vehicle’s lateral responses becomeslow when the driver is fatigue. But, in order to ensure driving safety, the driver will adjustthe vehicle frequently on second task which leads to a severe lateral responses.The controlling way of the driver can be speculated by the deviation responses of thevehicle, and then the driver’s different awareness modes can be mapped out. As a result thata behavior prediction model will be established by observing the statements of vehicles anddrivers’ behaviors. A theoretical model based on GM-HMM (Gaussian Mixture-HiddenMarkov Model) which can analyze a series of continuous time data and contain a variety ofenvironmental mutations during driving has been established. At last, the model training andvalidations are accomplished by Matlab.