A Learning-Based Approach for Lane Departure Warning Systems with a Personalized Driver Model

摘要

Misunderstanding of driver correction behaviors (DCB) is the primary reasonfor false warnings of lane-departure-prediction systems. We propose alearning-based approach to predicting unintended lane-departure behaviors (LDB)and the chance for drivers to bring the vehicle back to the lane. First, inthis approach, a personalized driver model for lane-departure and lane-keepingbehavior is established by combining the Gaussian mixture model and the hiddenMarkov model. Second, based on this model, we develop an online model-basedprediction algorithm to predict the forthcoming vehicle trajectory and judgewhether the driver will demonstrate an LDB or a DCB. We also develop a warningstrategy based on the model-based prediction algorithm that allows thelane-departure warning system to be acceptable for drivers according to thepredicted trajectory. In addition, the naturalistic driving data of 10 driversis collected through the University of Michigan Safety Pilot Model Deploymentprogram to train the personalized driver model and validate this approach. Wecompare the proposed method with a basic time-to-lane-crossing (TLC) method anda TLC-directional sequence of piecewise lateral slopes (TLC-DSPLS) method. Theresults show that the proposed approach can reduce the false-warning rate to3.07\%.