Avoiding Competing Objectives in Observability-Oriented Nonlinear Model Predictive Control

Abstract

This paper describes a method to avoid the presence of a competition between control error and observability in the objective function of NMPC implementations that are concerned with enhancing state estimation accuracy. This method involves solving a preliminary optimization to find suitable hard constraints that are included in the computation of the next control action to prompt the system to get closer to the reference even if observability terms are used instead of control error terms in the objective function. The proposed controller is illustrated and compared with an NMPC that allows the trade-off between both objectives to take place, with the former displaying a far better performance in state estimation accuracy than the latter.