January 30, 2014 at 08:13 #925
Thank you for very generously sharing your impressive fuzzy logic skills and knowledge.
I am almost finished converting my Arduino based solar charge controller from PID to fuzzy. I am using the centre of maximums (plural) defuzzifier as described by Microchip in their AN600 fuzzy logic controller document. This defuzzifier seems to be a good choice for very minimalist microcontrollers. Given it requires very little computational resources, could it be easily incorporated into fuzzylite, please.
While I expect I could remove the qtfuzzylite initial nag screen by recompiling qtfuzzylite, it would be nice to have an easier way of doing so. I also use the very useful freeware image manipulation program Irfanview, which does not incorporate a nag screen. When you make a donation to the Irfanview developer he provides a registration code that lets you show the program as being individually registered. I suspect implementing such a feature for qtfuzzylite nag screen removal based on public key cryptography or some other appropriate method after receipt of a nominal donation to your project would not be too difficult. I think this would be a good feature to incorporate into the fuzzylite project.
BrianJanuary 30, 2014 at 09:45 #926
thank you for your kind words and good ideas.
I believe the defuzzifier you are after is the Weighted Average in fuzzylite, and you have to model the output terms as Constant values. In the example from the figure, you would model the OutputVariable using the following terms:
NegativeLarge Constant -30
NegativeMedium Constant -15
NegativeSmall Constant -5
Zero Constant 0
PositiveSmall Constant 5
PositiveMedium Constant 15
PositiveLarge Constant 30
As for the initial screen, I will consider your suggestions for version 5.0. At the moment, you can skip it by pressing Escape, or surely modifying the source and building it again.
Juan.January 30, 2014 at 11:00 #927
Thanks for your reply.
I was not certain if the weighted average defuzzifier only applied to Takagi-Sugeno FLCs.
BrianJanuary 30, 2014 at 13:50 #928
Yes, your fuzzy logic controller would become a Takagi-Sugeno model, which matches your description of little computational complexity. However, have in mind that you need to model the output terms as Constant values. Otherwise, if you use Triangles or any other membership function, you would end up with an inverse tsukamoto controller, which is hardly any useful.
You must be logged in to reply to this topic.