Stochastic Road Condition Identification for Electromagnetic Active Suspension Based on Support Vector Regression

Abstract

Accurate road condition identification is conducive to improving the accuracy of vehicle performance. Aiming at electromagnetic active suspension, a novel method is proposed to realize accurate road condition identification using finite unknown samples. Because actual road condition is changeable, it is not exactly consistent with the standard grade road. Therefore, this paper adopts the power spectral density value Gq(n0) as the identification object to identify the non-standard road condition. Accordingly, back propagation neural network (BPNN) and support vector regression (SVR) are employed to identify road conditions respectively. The results suggest that these two methods have high accuracy for the identification of standard grade roads. However, the random oscillation of road conditions increases the sample uncertainty, which seriously affects the identification accuracy of BPNN. This also causes that the accuracy of road condition identification obtained by SVR with finite sample data is significantly higher than that obtained by BPNN.