Inverse Dynamic System Identification Using Multiple Support Vector Machines

In order to identify the inverse model for nonlinear dynamic systems, a multiple support vector machines (MSVM) based method was presented. According to their differential orders for the dynamic system, the input and output variables were allocated into multiple calculational subspaces. Taking advantage of its nonlinear regression performance, each subspace was represented by the least squares support vector machine (LS-SVM) to map the influence of the output on the input with a certain differential order. To connect subspaces with differential orders, a synthetic LS-SVM was then delivered to embody the dynamic characteristic of all sub-networks to the inverse model. At last a simulation system was put forward to validate the feasibility of proposed method. The result shows that presented method has clear dynamic structure, which is effective for the inverse dynamic system identification.