Implementation issues of robust state estimation: Nominal system reformation approach

Abstract

This paper considers a robust state estimation (RSE) problem for uncertain standard systems subject to bounded uncertainties, based on the recently proposed signal division method (SDM) and the equivalent nominal system (ENS) reformation approach. Two alternative implementations of the regularized least-squares (RLS) problem in solving the RSE problem are revisited and reexamined in filtering structure differences for a general uncertain discrete-time linear stochastic system with bounded uncertainties in both the system dynamics and the measurement. Conclusions related to the differences of the filtering performances of the two robust filters regarding the two approaches are drawn based on comparisons between their obtained ENSs.