Improving Genetic Algorithm Based Parameter Optimizing LSSVM Model for Tyre Regression

Abstract

According to the requirement of vehicle dynamics' accurate simulation and control, the paper studies the tyre regression algorithm based on the tyre bench test data. Due to the tyre test's characters of few data and big discreteness, the method of least squares support vector regression (LSSVM), which has the superiority of structural risk minimization, was selected to find the nonlinear mapping between input and output variables of tyre test data. Removing data gross error and improving the sparsification measures were taken to increase the calculation real time of standard LSSVM algorithm. An adaptive genetic algorithm (AGA) with global searching ability was chosen to determine the kernel function and regularization parameters of LSSVM. Test data set’s regression root mean square error (RMSE) was taken as the fitness function of AGA. Finally, the tyre test data under steady state sideslip condition was provided to simulate and verify the effectiveness of tyre regression algorithm, according to the precision and real time requirements of vehicle dynamic simulation and control.