Nonstationary noise identification with the interacting multiple model algorithm

The interacting multiple-model state estimation algorithm has been shown to be one of the most cost-effective schemes for estimating the state of hybrid systems. Such systems, which have continuous and discrete uncertainties, are represented by a finite set of noisy state equations, each pertaining to a certain mode. The system can switch from one mode to another according to an assumed underlying Markov chain. This framework is described and used here to estimate the time-varying intensity of the noise processes in a dynamic system