Two level Differential Evolution algorithms for ARMA parameters estimation

The problem of determining simultaneously the model order and coefficient of an Autoregressive Moving Average (ARMA) model is examined in this paper. An Evolutionary Algorithm (EA) comprising two-level Differential Evolution (DE) optimization scheme is proposed. The first level searches for the appropriate model order while the second level computes the optimal/sub-optimal corresponding parameters. The performance of the algorithm is evaluated using both simulated ARMA models and practical rotary motion system. The results of both examples show the effectiveness of the proposed algorithm over a well known conventional technique.