An Extended U-D Algorithm with Multiple Forgetting Factors for RLS Estimation of Model Parameters

Abstract

On-line parameter tracking in near-real time is a good alternative for advanced FDD/FTC systems, because an updated model is often a prerequisite for many indirect adaptive control algorithms. For aircraft, monitoring and reconfiguring is beneficial to impaired airplanes, but also to under-equipped vehicles (UAS, small airplanes). Otherwise, many schemes have been proposed since the 80s to deal with the tricky problem of on-line estimation. However, their computational complexity is often incompatible with the constraints resulting from onboard implementation and A/C certification issues. Accordingly, this paper capitalizes on the reliable but simple enough U-D factorization algorithm, and proposes an extended version that includes multiple forgetting factors to adapt to possibly different parameter rates of change. After a detailed description of the algorithm, the approach is applied to an academic benchmark chosen to enable the comparison of different methods in realistic conditions, with and without excitation.