Robust estimation for signal reconstruction of nonlinear multifunctional sensor

Training sample set inevitably contains gross error in input signal reconstruction of nonlinear multifunctional sensor. In order to achieve the stronger robustness and the higher efficiency, Robust Least Squares (RLS) method is applied for parameter estimation under the disturbance of gross error. Based on the principle of equivalent weights, RLS method is combined with robust estimation theory and the Weighted Least Squares' form, which can resist a certain degree of gross error and maintain the advantage of Least Squares regression method. As shown in the calculation results of gross error restraint in nonlinear signal reconstruction, the Robust Least Square method has favorable outlier resistance, convergence reliability and high calculating speed.