Row Alignment via Hidden Markov Model based Learning Control

Understanding human driver behaviors is crucial for an advanced driver assist system in the path control of a vehicle. In this study, an uncertain parameter in a human driver internal model is estimated using a Hidden Markov Model-based learning process. Based on the identified parameter, a Linear Quadratic Gaussian controller is designed for the vehicle to follow an online planned, optimal path, while reducing the row alignment error caused by the deviation of the driver internal model from the actual vehicle model and sensor/actuator noise. Simulation is used to show the effectiveness of the controller.