TABLE OF CONTENTS The previous chapter presented a parametric modeling approach using transfer-function identification, with primary application to the characterization of single-input/single-output decoupled responses. Transfer-function modeling is generally accomplished in quick fashion and provides important insight into the appropriate model structure and dynamic characteristics (i.e., dominant modes, time delay, and key derivatives) required to track the primary responses. In the many applications illustrated in Chapter 11, these transfer-function models provide the needed endproduct of system identification. However, the transfer-function model structure is not easily extended to generalized MIMO systems, where the state-space representation is better suited. Also, state-space models are often the needed final product of system identification for applications such as simulation model development and control system design.
The frequency-response method for state-space model identification is a direct extension of the transfer-function identification, using the same cost function based on frequency-response error. Here we seek a state-space model that produces a frequency-response matrix T that best fits the (measured) MIMO composite frequency-response estimate matrix 
obtained from Chapter 10. State-space model identification is covered in two chapters (Chapters 12 and 13). This chapter presents the basic concepts of the frequency-response method for state-space model identification and applications to canonical representation of SISO systems. Chapter 13 extends the discussion to the identification of physical structure representations of MIMO systems.
There is a wealth of reference material on state-space model identification of aircraft and helicopter dynamics in the time and frequency domains. An excellent overview and reference list of the maximum likelihood...
