Novel nonlinear system identification method based on kernel function

A novel algorithm is presented based on kernel function to identify the nonlinear system. This approach requires no priori information about the system inputs and outputs (SIO), and discovers the system's model configuration by estimating the density, clustering the SIO data, and getting the kernels respectively. Then the SIO data are projected into high-dimensional space based on these kernels. The system's parameters are got via the recursive least square method. Several simulation results are presented to support the effectiveness of the proposed adaptive algorithms.