🚀 go-pugleaf

Andrzej Kozakiewiecz wrote:
> 
> Klaus Bernhard wrote:
> > Why would one use state feedback ( and therewith Kalman filtering)
> > instead of output feedback, when I can feed my controller with the
> > latter one directly instead of estimating the states with an observer
> > and so on...?

> Hi Klaus,
> 
> In engineering the simplest things are always the better: so if you can
> achieve a decent control scheme just with normal output feedback then do
> it and don't even wonder about it.
> State feedback allows you a much more "sophisticated" approach to
> things:
> - you can position your poles anywhere you want (at the price of high
> control gains) in the LHP not only on your root locus
> - you have access to control tools like the LQR regulator
> 
> Bottom line, it has quite a lot of andvantages but you pay the price of
> having a more complicated control system by including an observer,
> etc...

Actually it isn't necessarily that much more complicated.
In the lab I built up a state space controller for a DC
motor with just op-amp circuits. I also built up a SS
current controller with just a few parts. I guess one
of the other values to state space is that it sets you 
up well to do true feedforward. That gets rid of a lot of
those pesky lags. With the DC motor controller, we managed
to demonstrate trajectory following well in excess of the 
BW of the controller (and surprisingly effective open
loop control too). With the current controller, since I
built it in a state space form, I was able to implement
it with a simple proportional gain (no integrator) without
any of the offset error that would typically be found in
a P controller.

A disadvantage of SS is that many years of experiences and
rules of thumb from PID methods can't readily be used
with State Space methods.

I think that it is worth noting the any PID controller
can be mathmatically unwound into a state space form. The
difference is that the PID gain terms will show up in
differnt SS gains, and the states will not have references.
Fundamently then, if you provided state references into
a PID controller, it would be mathmatically equivalent
to a SS controller.

One comment on observers, if you are going through the
trouble of building an observer, use an observer, not
a filter. Properly developed (and if the right states
are available) an observer can operate without
the phase shifts present in filters.

Kurt Jaeger
Controls and Sensor Technology Group
UW-Madison Graduate SChool of Engineering

Remove the SPAM from my adress to contact me