A subspace approach to the identification of MIMO piecewise linear systems

摘要

The problem of estimating parameters of a multi-input multi-output (MIMO) piecewise linear system (PWL) is considered. It is assumed that noise is present in both input and output measurements. A subspace approach based on the singular value decomposition (SVD) is proposed, which transforms the problem of associating noisy measurements to submodels to one of identifying a certain permutation structure in the observation subspace. Main advantages of the proposed approach include lowered computation complexity in handling arbitrarily shaped of domain partitions and its ability to cope with measurement noise at the system input.