MIMO order and state-space model identification from interval data

In this paper, a method for MIMO system identification with unknown, but bounded measurement errors is presented, extending a recently published method for systems with only one output. It is based on an interval description of the sampled measurement data of linear, time invariant systems, where the intervals cover the unknown, but bounded errors, and yields a discrete-time model description with interval parameters, where the intervals cover the uncertainty of the parameters due to the measurement errors. The main contributions of the paper are a procedure to determine the model order from specially arranged data matrices and the transformation from the initially identified ARX model to the state-space model, which both cannot be directly extended from the MISO case. The pros and cons of the method are discussed and examples are presented.