In control theory, there is something called "Model reduction". That's means states from a state space model decreasing.
I know that's a small state space model, is much easier to model, that a large state space model.
For exampel. If I want to build a LQGI (Linear Quadratic Gaussian Integral) controller with analog electrical compontens such as several operation amplifiers, resistors and capacitors. It will requires a lot of those if the state space model is large.
So my question is: Can I still have an optimal controller, even if I reducing a lot of states? Or do I lose some optimal properties just by "removing" som states?
I would like to build a LQGI controller with only analog electrical components because that controller will then work in real-time, like a digital controller would not do, due to the sampling time.