Models For Robust Estimation And Identification

In this paper, estimation and identification theories will be examined with the goal of determining some new methods of adding robustness. The focus will be upon uncertain estimation problems, namely ones in which the uncertainty multiplies the quantities to be estimated. Mathematically the problem can be stated as, for system matrices and data matrices that lie in the sets (A + #A) and (b + #b) respectively, find the value of x that minimizes the cost (b + #b)#. The proposed techniques are compared with currently used methods such as Least Squares (LS), Total Least Squares (TLS), and Tikhonov Regularization (TR). Several results are presented and some future directions are suggested