Model order selection for continuous time instrumental variable methods using regularization

The aim of this paper is to propose a new method to select the model order in continuous time system identification, instrumental variable methods. The idea is to over-parameterize the model and utilize regularization based on the l1 norm to obtain a sparse estimate. The model order of the identified system is then determined by the rank of the Hankel matrix of the estimated parameter. Simulation results show that the proposed method works very effectively. For low signal to noise ratio (SNR), it offers a significant improvement to existing model order selection methods with the performance at high SNR comparable to the existing methods.