Nuclear norm minimization in subspace based continuous-time Hammerstein system identification

A novel method combining the nuclear norm minimization (NNM) and continuous-time (CT) subspace identification method (CSIM) is proposed to identify the CT Hammerstein model with little priori information. The nuclear norm minimization, which is the heuristic convex relaxation of the minimum rank constraint, is applied to the CT subspace identification method, for the purpose of improving the robustness and accuracy of identification. The proposed method can perform the identification well without the priori information about the Hammerstein model, which not only reduces the complexity of the identification problem but also broaden its applications. The nonlinear block of Hammerstein model is approximated with the pseudospectral method, which replaces the nonlinear function with Lagrange basis functions. A typical numerical example is presented to verify the NNMCSI method and the identification results are compared with the refined instrumental variable method.