Recursive direct weight optimization in nonlinear system identification: A minimal probability approach

In this paper, a direct weight optimization method is proposed for nonlinear system identification based on a minimal probability idea. The approach has several quite attractive features and is very different from existing ones. It is optimal for any given number of finite data points and at the same time possesses asymptotic convergence. The estimator admits a closed form and no numerical optimization is needed. Theoretical analysis and numerical simulations show that the approach is a very competitive alternative to existing nonlinear identification methods.