I would like to know which tool (software) I can use to calibrate a fuzzy model by using ANFIS. I have seen that MATLAB includes the ANFIS tool box but there are some limitations.
You can write a script and use any optimization algorithms to calibrate Mamdani type FIS. In one publication we obtained rules from the experts and tuned other parameters using Matlab. I have not used any other software for developing FIS.
ANFIS can be written easily in Matlab. I agree that ANFIS command and toolbox are not enough for some applications.
I recommend you to study in detail ANFIS papers of Jang, clustering techniques, LMS and backpropagation algorithm to write a simply code to combine them to reach a better ANFIS.
Remember that jang´s ANFIS is a MISO sugeno type, but GENFIS3 Matlab option allows you to compose a mamdani ANFIS and use it as idea to your code.
For instance, Jang´s paper here: link 1, and summarizing some concepts here: link 2
Clustering techniques: attached
In my courses, I use to start teching fuzzy sets, fuzzy logic and fuzzy inference systems, later, neural networks concepts and backpropagation algorithm are introduced until student learn to write in a program the neural net.
After that, clustering techniques for unsupervised learning concepts relating them to fuzzy logic and membership functions..
Finally ANFIS concepts become easier to understand.
Btw, I have found already a paper on the application of ANFIS concepts to investigate Human Behaviour in Fire Evacuation: http://link.springer.com/article/10.1007/s10694-008-0064-6
I do hope to develop another interesting application in the future using new data-sets.