Stabilized least-squares type adaptive identifiers

Least-squares-type adaptive identifiers are proposed which are stabilized in the sense that the covariance equation has a positive definition, asymptotically stable equilibrium. This keeps the adaptive gain matrix bounded and the adaptive abilities of the algorithm always properly in operation. The convergence speed of the stabilized adaptive identifiers is proportional to the signal level when the signal level is low and can be made to achieve (almost) arbitrarily fast convergence for high signal levels. The algorithms require the inversion of positive definite matrices and/or the solution of associated linear equations, which is numerically well behaved. As a result, a slightly increased computational complexity, as compared to known least-squares-type algorithms, is to be paid for advanced stability and adaptive properties of the algorithms