Abstract: To deal with the time-variation,nonlinearity and multivariable couping characteristics in the process of steering system back pressure,the generalized predictive control strategy based on least square support vector machine(LS-SVM) was proposed.The LS-SVM model is established to identify the SSBP.Particle swarm optimization is proposed to get better value of normalizing parameter and kernel parameter.With time-variable inertia,an adaptive direct generalized predictive control method based on online LS-SVM is adopted,which revises the parameters for the model in time.Resultsof the actual application for the control of SSBP demonstrate that the proposed adaptive direct generalized predictive control method based on online LS-SVM is effective,the performance of loading pressure tracking is good,and the robustness to the disturbance is strong.The method can be applied to some control fields like the process of SSBP.