Partitioning identification algorithms

Abstract

In this paper, batch processing partitioning parameter identification agorithms are obtained using the "partitioning" approach to estimation. The algorithms, herein denoted the GPIA's, are applicable to linear as well as nonlinear systems and are derived by a natural applicalton of the generalized partitioned algorithms (GPA's) of Lainiotis; namely, by selecting a natural partitioning of the augmented state vector (the system state and unknown parameters); by linearization of the model equations; and then by using, in an iterative fashion, the GPA algorithms for the augmented state. The relationships between the GPIA's and maximum-likelihood identification methods, which employ gradient based numerical techniques to obtain a solution, are also established. An example of the application of the GPIA to aircraft parameter identification from actual flight test data is presented, as well as a direct comparison with the results obtaining using an iterated extended Kalman filter algorithm.