Probabilities of Mandani FIS output terms

[Unknown]

Hi all, I am currently using jfuzzylite 6.0.1 in my lab project for control problems. I have realized that the probabilities of the output for the terms of an output variable does not always add up to 1 and I would be glad to get any insights into why that is the case. Thanks.

[Juan Rada-Vilela (admin)]

Hi,

thanks for your post.

Firstly, the output of fuzzy logic controllers are not probabilities, and hence do not need to add up to 1.0. However, I understand what you are trying to achieve. It would be easier for me to help you if you post the FLL code of your fuzzy controller. Could you post that?

[Unknown]

Thanks for your feedback on my issue. I am currently using evolutionary processes to tune the knowledge base of my controllers so its FLL may be quite different from the standard jfuzzylite FLL. However, I have modified the demo ObstacleAvoidance FLL according to the basic framework of all my controllers and I am certain if that will help in achieving the goal of making the fuzzy outputs probabilities then I can do same in my project. My plan-b is to take the softmax of the outputs but I was thinking the values could be probs so I do not have to. Thanks.

#File: ObstacleAvoidance.fll
Engine: ObstacleAvoidance
InputVariable: obstacle1
  enabled: true
  range: 0.000 1.000
  lock-range: false
  term: left Triangle -1.000 0.000 1.000
  term: right Triangle 0.000 1.000 2.000
InputVariable: obstacle2
  enabled: true
  range: 0.000 1.000
  lock-range: false
  term: left Triangle -1.000 0.000 1.000
  term: right Triangle 0.000 1.000 2.000
InputVariable: obstacle3
  enabled: true
  range: 0.000 1.000
  lock-range: false
  term: left Triangle -1.000 0.000 1.000
  term: right Triangle 0.000 1.000 2.000
OutputVariable: myOutput
  enabled: true
  range: 0.000 1.000
  lock-range: false
  aggregation: Maximum
  defuzzifier: MeanOfMaximum 100
  default: nan
  lock-previous: false
  term: outputTerm1 Triangle -1.000 0.000 1.000
  term: outputTerm2 Triangle 0.000 1.000 2.000
RuleBlock: mamdani
  enabled: true
  conjunction: Minimum
  disjunction: Maximum
  implication: Minimum
  activation: General
  rule: if obstacle1 is left and obstacle2 is left and obstacle3 is left then myOutput is outputTerm1
  rule: if obstacle1 is left and obstacle2 is left and obstacle3 is right then myOutput is outputTerm1
  rule: if obstacle1 is left and obstacle2 is right and obstacle3 is left then myOutput is outputTerm2
  rule: if obstacle1 is right and obstacle2 is left and obstacle3 is right then myOutput is outputTerm2
  rule: if obstacle1 is right and obstacle2 is left and obstacle3 is left then myOutput is outputTerm1

[Author]

Posts