ABSTRACT
This chapter highlights the importance of the modelling stage and assumptions. It looks at the impact the model structure has on the control law and hence on sensitivity. The use of different prediction models implies the use of different observers and hence gives rise to control laws with different loop sensitivity. An independent model gives equivalent predictions, in that they are based predominantly on input information and a current disturbance estimate. However, an independent model can be constructed as a transfer function or state-space model and hence has fewer parameters and also no truncation errors. The upper bounds on prediction error variance of equations are computed; these are denoted theoretical bounds. Transfer function models realign the state at each sampling instant and hence put far more emphasis on output measurements, that is, on possibly noisy data.
