Decomposition of steering angle measurements inspired by pulse control model for lane keeping

Abstract

Steering control for vehicle lane keeping has attracted great attentions from both automotive industries and vehicle control researchers. Traditionally commonly used linear control model can not adequately represent the intermitted pulse-like qualities in real-world steering angle measurements. Therefore, based on an alternative `pulse control model' of steering control for vehicle lane keeping, steering angle measurements can be decomposed into a combination of ramps (global curvatures) and pulses (local corrections). Measurement trend is extracted through empirical mode decomposition (EMD) technique, where principal component analysis (PCA) is applied as a supplement to quantify the component contribution. Reversal points are found from the extracted pulses by further selections of conditional rules, which then reveals a pattern which is useful to test the overall model validation.