Road profile classification for vehicle semi-active suspension system based on Adaptive Neuro-Fuzzy Inference System

Abstract

To meet the requirements of excitation information for semi-active suspension control, a new road classification method with application of Adaptive Neuro-Fuzzy Inference System (ANFIS) was presented. Due to distinct system responses for different road levels, the sprung mass acceleration signal was utilized for classification. To analyze the properties of various road inputs from different perspectives, the acceleration signal was first decomposed into 5 categories via wavelet transform, and 11 statistic features were calculated for each category. Then, an improved distance evaluation technique was applied to remove irrelevant features. With the extracted superior features, a new 2-layers ANFIS classifier was implemented to calculate overall road level. Simulation results revealed that the proposed classifier had significantly improved performance compared to all 1-layer ANFIS classifiers for individual category, and can accurately classify road level with negligible time delay.