Estimator and controller design for LaneTrak, a vision-based automatic vehicle steering system

This work focuses on the underlying theory used in the design of a lane estimator and a lane controller to achieve automatic roadway tracking by a vehicle. In concert with a lane sensing algorithm, a comprehensive simulation assesses the roadway tracking performance of a Pontiac 6000 STE. The roadway curvature and the relative positioning of the vehicle within its lane are estimated using Kalman filtering. Inputs to the estimator are vehicle kinematical variables provided by a vehicle directional control (DIRCON) model, and lane boundary information provided by a video camera model. The design of the steering autopilot is formulated as a preview, path tracking and regulation problem. The steering autopilot uses the estimator outputs and the vehicle DIRCON model. The estimator and controller designs were tested, improved, and validated using a comprehensive simulation program which included models for the lane sensing system and the steering actuator. The credibility of the simulation has been further established through successful lane control of a Pontiac 6000 STE car on an unopened section of a limited-access freeway at highway speeds