Feedforward and feedback control of lane keeping by the human driver

A feedforward/feedback control model of drivers' lane keeping behavior is presented in this paper. The model is based on the linearized analysis of the driver's curve negotiation dynamics. In real driving, the driver previews the upcoming road geometry and relies on perceived vehicle states to maintain a desired lateral position. Feedforward and feedback roles are associated with different perceptual cues, and the lane keeping task is formulated into a disturbance rejection problem. Control parameters are determined to reflect natural stable human characteristics. Verification tests in a realistic simulation environment demonstrate the ability of the model to generate driver/vehicle lane keeping responses comparable to those obtained on a simulator. Potentially, the derived control algorithm can also be applied to automatic lane-tracking as long as reliable information regarding vehicle states and upcoming road conditions is accessible.