Lecture 6

Question 1

the output error method

Answer 1

maximum likelihood estimator for data containing only measurement noise and no process noise

1. choose suitable starting values
2. compute system responses by solving the IVP
3. compute residuals and residual covariance
4. compute the output sensitivities
5. compute parameter update with non-linear optimisation algorithm
6. iterate on step 2 until convergence

assumptions:
1. no process noise
2. only additive, independent, Gaussian measurement noise
3. inputs are measurable without error
4. inputs are independent of system outputs
5. inputs are sufficiently and adequately varied to excite the various modes of the dynamic system