Modification of recursive least squares algorithm for linear time-varying systems

Modification of recursive least squares algorithm is presented for identification of linear time-varying systems. The basic idea is to divide the weighting vector into two part. One represents the main variation tendency of the time-varying parameters, while the other is a slowly varying vector. First the former is determined by a procedure based on the dependency of data, and then the remaining part is identified using recursive least squares algorithm with forgetting factor. The modified algorithm avoids the problems appeared in rapidly varying situations, with good tracking ability.