Adaptive PID

Post Reply
rihardsb
Posts: 14
Joined: Thu Mar 31, 2011 11:47 am
Location: Ireland
Contact:

Adaptive PID

Post by rihardsb »

Hello All,

I wonder, could Adaptive PID be implemented into MultiWii project instead of the current Conventional PID values. This would involve to incorporate Artificial Intelligence (some sort of a Neural Network) into digital control to make PID values adaptive in the closed loop. This question would be very relevant to you who are experts in Digital Signal Processing.

Best Regards,
Rihards

Noctaro
Posts: 280
Joined: Thu Sep 08, 2011 11:15 am
Contact:

Re: Adaptive PID

Post by Noctaro »

hey,
i also thougt of using the genetic algortihm to determine the best values. But at the moment i could not find a way for it to be implemented safe. As GA does create random Chromosomes wich will behave anyhow. In flight there are to much conditional things that could lead to a crash if misinterprated by GA. No problem if we are on a flat surface.(Crawling Bot) But in air its dangerous. Further it would need around 2hours or more to produce some useable settings. This flighttime will not be realistic if you dont have a laserbeam for power supply and a solarpanel on your copter.
Within two hours as human we are also able to produce very usable settings.
Howevery, an auto tuning copter would be great :o

copterrichie
Posts: 2261
Joined: Sat Feb 19, 2011 8:30 pm

Re: Adaptive PID

Post by copterrichie »

IMO, an expert system would be great for a test bench environment. The numbers or values obtained, then can be used to set the actual PID for flight.

mahowik
Posts: 332
Joined: Sun Apr 10, 2011 6:26 pm

Re: Adaptive PID

Post by mahowik »

here is also some thoughts on this...
viewtopic.php?f=7&t=905&start=40#p5956

rihardsb
Posts: 14
Joined: Thu Mar 31, 2011 11:47 am
Location: Ireland
Contact:

Re: Adaptive PID

Post by rihardsb »

There are various algorithms for digital control including Fuzzy Logic and ANN, note that ANN isn't the actual controller. It learns in time for the best constants to feed into PID control algorithm, hence we can start off with the default PID constants.
pid+ann.PNG

Adrian A. Hopgood "Intelligent Systems for Engineers and Scientists" p.448

I'll have more time January after Xmas and New Years, so I'll try to embark on this research area!

Regards,
R

User avatar
Bledi
Posts: 187
Joined: Sat Sep 10, 2011 6:36 pm

Re: Adaptive PID

Post by Bledi »

For beginning why not adding informations to GUI to use the Ziegler-Nichols method to find thé best parameters:
[url]http://en.wikipedia.org/wiki/Ziegler–Nichols_method[/url]

Post Reply