Visualisation of sFLC3 Fuzzy Sta
|
Home
NEW:
US Markets Daily |
Visualisation of sFLC3 Fuzzy State RepresentationFirst, let's recap for a moment about our sFLC3 DLL. The purpose has always been to facilitate the automatic construction of fuzzy logic controllers so that we could envisage using an optimisation technique to build and test massive amounts of controllers in various configurations. Our answer was to include in the library a number of built-in fuzzy sets and rule sets. We shall at a later stage expand toward fuzzy state machines and optimisation techniques, but for the time being, let's say it is convenience driven i.e. helping the programmer is building decision making systems in the easiest possible way. In the current version, the DLL provides 2 types of FL Controllers, namely 3x3 and 5x5 controllers. Users with full access to the API can design any other model, but they are often too complex to build correctly and are quite difficult to maintain. The 3x3 Shape Model is a simple fuzzy model with 2 inputs and one output, each having 3 possible fuzzy states. It will be suitable for building rules such as for instance: IF (StoK is HIGH) AND (StoK_Gradient is DOWN) THEN Prob_DownMove is HIGH This model has 9 built-in i.e. ready made fuzzy sets, and 5 built-in rule sets, hence a lot of possible combinations to play with already. Again, for simplicity sake, we will here first assume both variables use the same Fuzzy Set model (nota bene: one must keep in mind that sFLC3 uses 2 inputs and 1 output all normalised to the [-1:+1] range). A picture being worth so many words, here is the data output from our sFLC_Sample1 console application exported to Excel then plotted to represent the workings of our FLC.
This example comes from FLC0_3332_3.xls output by our sample console application. All sorts of variations are of course possible to fit the behaviour of virtually any variables. Furthermore, full access to the FLC3 API allows build custom fuzzy sets and rule sets.
|
Page last modified:
May 08, 2008 |