Suppose I have a real (physical) dynamical system with some sensors and actuators, and I also have an idealized state-space model of this same system. How, in general, can I adjust the model to match the real system?

My experience is with frequency-domain analysis. My first reaction would be to measure swept-sine transfer functions, and to calculate parameters of my system from the location of resonances or poles in the measured data.

Is there a better, more general technique for system identification in the state-space representation?

  • $\begingroup$ Is the system assumed to be linear? If it is, then a frequency response should be sufficient to characterize it. $\endgroup$ – Jason R Apr 13 '12 at 15:08
  • $\begingroup$ Yes, but how do you put the experimental frequency domain data into the state space form? $\endgroup$ – nibot Apr 13 '12 at 15:34

The experimental frequency domain data gives you information on the input-output behavior of the system. If the system is assumed to be linear and time-invariant (LTI), this information will be in the form of a transfer function. Given a transfer function, there are infinitely many state-space realizations that yield the same input-output behavior and, thus, are I/O-equivalent.

Having said that, you may want to take a look at the Adaptive Control literature.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.