A popular approach is using recursive, multivariable least-squares (generally ARARX or ARMAX) to update the system parameters then there are a great many different techniques for updating the control feedback gains depending on your computational power and time horizon.
My problem is the recursive, multivariable least-squares algorithm. It's flipping complicated. In the adaptive control references, it tends to presented in pretty much the same manner as "Step 2" in the following (rather ubiquitous) cartoon:
I'll figure it out... (I hope)...Messages of pity or of disdain are welcome in the comments section.
4 comments:
FIXED!
Note to all:
Extended least-squares algorithms used in feedback control systems are not as forgiving about low sampling rates and aggressive control gains as are standard LQG control systems.
S.W.
I also found a transpose in my Matlab was missing. In Matlab, the (Hermetian) transpose is designated by an apostrophe following the entity you want to transpose. They're easy to overlook.
It's kind of scary that it was actually working pretty well with the missing transpose.
It rocks nads now. That's right. Nads rocking code. NADS ROCKING!
Well, I'm impressed, at least. Although by all that is good and holy, I hope never to have to deal with self-adapting control problems.
Dave, aren't you in mathematics?
What are you doing with any familiarity with something as (very nearly) useful as self-adaptive control?
Post a Comment