Error in Fuzzy control system

[Nikhil Rao]

When i tried to simulate two inputs

input range (-12,12,1) for both the inputs 

output range (-24,24,1)

input crisp values 

input1=-10

input2=0

The following error was obtained

Traceback (most recent call last):

  File "FBB.py", line 44, in <module>

    tipping.compute()

  File "/usr/local/lib/python2.7/dist-packages/scikit_fuzzy-0.2.dev0-py2.7.egg/skfuzzy/control/controlsystem.py", line 296, in compute

    CrispValueCalculator(consequent, self).defuzz()

  File "/usr/local/lib/python2.7/dist-packages/scikit_fuzzy-0.2.dev0-py2.7.egg/skfuzzy/control/controlsystem.py", line 445, in defuzz

    raise ValueError("Crisp output cannot be calculated, likely "

ValueError: Crisp output cannot be calculated, likely because the system is too sparse. Check to make sure this set of input values will activate at least one connected Term in each Antecedent via the current set of Rules.

[JDWarner]

This message appears when your Rules are not sufficient, and a given set of inputs misses all defined Rules. 

These has been some discussion about the possibility of a default firing in such a case, but I'm uncomfortable making assumptions.

Put simply, you need to make sure that your Rules connect Antecedent Terms such that any combination of inputs will fire at least one Rule. If this doesn't happen, the system can't execute because it doesn't have the information needed to proceed. We catch this and tell you with an information message - that's what the exception is saying. Recheck and/or add more Rules.

[JDWarner]

Example script sent to me directly via email:

``` python

import numpy as np

import skfuzzy.control as ctrl

import skfuzzy as fuzz

import commands

error = ctrl.Antecedent(np.arange(-12,12, 0.01), 'error')

derivative = ctrl.Antecedent(np.arange(-12, 12, 0.01), 'derivative')

output = ctrl.Consequent(np.arange(-24, 24, 0.01), 'output')

error['N']=fuzz.trimf(error.universe,[-12,-12,0])

error['P']=fuzz.trimf(error.universe,[0,12,12])

error['Z']=fuzz.trimf(error.universe,[-10,0,10])

derivative['N']=fuzz.trimf(derivative.universe,[-12,-12,0])

derivative['P']=fuzz.trimf(derivative.universe,[0,12,12])

derivative['Z']=fuzz.trimf(derivative.universe,[-10,0,10])

output['N']=fuzz.trimf(output.universe, [-24,-12, 0])

output['P']=fuzz.trimf(output.universe, [0,12,24])

output['Z']=fuzz.trimf(output.universe, [-10,0,10])

rule1 = ctrl.Rule(error['N'] & derivative['N'],output['N'] )

rule2 = ctrl.Rule(error['N'] & derivative['Z'],output['N'] )

rule3 = ctrl.Rule(error['N'] & derivative['P'],output['Z'] )

rule4 = ctrl.Rule(error['P'] & derivative['N'],output['Z'] )

rule5 = ctrl.Rule(error['P'] & derivative['P'],output['P'])

rule6 = ctrl.Rule(error['P'] & derivative['Z'],output['P'])

rule7 = ctrl.Rule(error['Z'] & derivative['N'],output['N'])

rule8 = ctrl.Rule(error['Z'] & derivative['P'],output['P'])

rule9 = ctrl.Rule(error['Z'] & derivative['Z'],output['Z'])

output_ctrl = ctrl.ControlSystem([rule1, rule2, rule3, rule4, rule5, rule6, rule7, rule8, rule9])

outputs = ctrl.ControlSystemSimulation(output_ctrl)

```

Two problems are apparent: 

• The new API has a bug where multiple Rules converging on a single Consequent Term overwrite each other instead of combining. An Issue is raised on GitHub (https://github.com/scikit-fuzzy/scikit-fuzzy/issues/108) and a fix is pending.
• Your Antecedent Terms called 'Z' should be extended to cover the full range of [-12, 12] instead of [-10, 10]. 

With the 'Z' range extended, this should run without problems after the fix for Issue #108 is merged.

[JDWarner]

This was fixed by PR #109 on GitHub, confirmed resolved with regression tests added, prior to 0.2 release.

On Thursday, June 16, 2016 at 3:11:32 AM UTC-5, Nikhil Rao wrote: