>-- Original Message -- >Date: Tue, 13 Jul 2004 18:57:13 +1200 >From: Birger Brunswiek >To: FlightGear developers discussions <[EMAIL PROTECTED]> >Subject: Re: [Flightgear-devel] Autopilot >Reply-To: FlightGear developers discussio s <[EMAIL PROTECTED]> > > >rvovesen wrote:
> Isn't it more apropriate to have a multidimentional PID c > ntroller rather > than multipl > >>1 dimentional PID controllers? Or is that just the same? > > > I'm not sure what you ean by multidimentional PID c >ntroller, but I believe > that that it would be the same as multiple one dimentional PID controllers. The one described is one dimentional because all variables are scalars. In a multidimentional version the target, input and output would be vect >rs and gains matrices. With the matrices you should be able to express interaction between the components. e.g. as you said rudder and ailerons have overlapping effects. You mean a state-space form?!: x' = x + Bu y = Cx + Du I think that would make the autopilot-configuration a bit more complicated. One of the things that was requested on this list when we redone the autopilot system was that it had to be simple to configure. For the state-sp ce form one would require a matematical model of the aircraft, the A matrix. With PID controllers you don't need matematical models. Since every aircraft have different matematical models, that would add to the complexity of configuring the autopilo . I'd say with two 1-dimenational controllers you ignore that fact and get imper >ect results. Yes, since we don't have a matematical model to help us out, we ignore that fact. But we still know, in our head, that rudder and aileron interact and consequently we tune the two PID controllers in a slightly different way. So in theory we ignore it, but in pactice we consider it. could you not take the heading as input and aileron controll as output directly, since the effect of the ilerons (rol >) is directly connected to change of heading. Pleasy try it out! ;-) I guess that you will learn that as the controller demands full aileron deflection, the aircraft will roll, roll and roll until it's inverted, and it won't stop t ere, it will continue to roll. And the aircraft will return to almost the same heading as it had befor the roll started. :-( Back to the state-space form. There is nothing stopping you from implementing a state-space form to the existing autopi ot code. The source files to look for are "xmlauto.*" in the "autopilot" subdir. If you think that a multidimentional state-space model would give better performance than the PID controllers, go ahead! I also think that it would give better performa ce, and I would gladly help you in implementing. -- Roy Vegard Ovesen -- Roy Vegard Ovesen _______________________________________________ Flightgear-devel mailing list [EMAIL PROTECTED] http://mail.flightgear.org/mailman/listinfo/flightgear-devel
