Lateral State Prediction for Automated Steering using Reliability-Weighted Measurements from Multiple Sensors

This paper presents a time-varying Kalman filter that combines the measurements from four different lateral deviation sensors to produce a reliable prediction of lateral vehicle dynamics state for automated steering. The measurement noises used in the filter vary continuously according to a reliability index, which itself is a function of individual sensor capabilities and the current driving environment. The reliability index is continuously varying and differs from previous work that simply changes the source of the lateral deviation measurement for the filter according to the reliability index. The practicality of the prediction method is shown through the implementation on a test vehicle.