• classic control
• IMC
• MPC
• observer
• optimal control
• stabilizing feedback

Internal Model Control (IMC) is a well-known control strategy provided with simple
tuning rules requiring a model in order to control a given single-input single-output
plant; furthermore, it allows an easy and straightforward closed-loop analysis.
However, it has some limitations. For instance, it cannot be applied to open-loop
unstable plants, it does not cope easily with constraints, and disturbance rejection may
be sluggish for disturbances other than output steps.
On the other hand, Model Predictive Control (MPC), that still requires the definition of
a model, has not limitation from the point of view of the nature of the plant, but it does
not give allows simple CL analysis.
IMC and MPC have many common features but, at the same time, they are also quite
different control strategies: the goal we want to achieve in this work is to find a
compromise between them that should have advantages of both control structures.
In this work a Disturbance Observer Based Internal Model Control (DOB-IMC) is
proposed: it works with an augmented model, classical IMC controller design is left
unchanged, while the block standing for the model has been replaced by an observer
block, where predicted states are ”filtered” through a Luenberger observer, known to
deal better with dynamic disturbances rather than classic IMC deadbeat observer.
Afterwards, this structure has been extended to open-loop unstable plants through
application of the Q parametrization, and to integrating plants as well.
The effectiveness of this control scheme has been validated through several
simulations: first, different kind of Single-Input Single-Output linear systems have
been tested; then, as a pratical application, the multivariable ”Shell oil fractionator”
case study has been simulated with unmeasured disturbance and with saturated
inputs.