def generate_mf_group(self, G, x):
mf_group = {}
for (k, v) in G.iteritems():
shp = v['shp']
mf = v['mf']
if mf == 'trap':
mf_group[k] = trapmf(x, shp)
if mf == 'tri':
mf_group[k] = trimf(x, shp)
if mf == 'gbell':
mf_group[k] = gbellmf(x, shp[0], shp[1], shp[2])
if mf == 'gauss':
mf_group[k] = gaussmf(x, shp[0], shp[1])
if mf == 'gauss2':
mf_group[k] = gauss2mf(x, shp[0], shp[1])
if mf == 'sig':
mf_group[k] = sigmf(x, shp[0], shp[1])
if mf == 'psig':
mf_group[k] = psigmf(x, shp[0], shp[1], shp[2], shp[3])
if mf == 'zmf':
mf_group[k] = zmf(x, shp[0], shp[1], shp[2], shp[3])
if mf == 'smf':
mf_group[k] = smf(x, shp[0], shp[1], shp[2], shp[3])
if mf == 'pimf':
mf_group[k] = pimf(x, shp[0], shp[1], shp[2], shp[3])
if mf == 'piecemf':
mf_group[k] = piecemf(x, shp[0], shp[1], shp[2], shp[3])
return mf_group
def _plot_mf_for(x):
'''Given the x-values, plot a series of example membership functions'''
tests = OrderedDict([
# Gaussians:
('gauss', skmemb.gaussmf(x, 50, 10)),
('left gauss', leftgaussmf(x, 50, 10)),
('right gauss', rightgaussmf(x, 50, 10)),
# Triangles:
('triangular', skmemb.trimf(x, [25, 50, 75])),
('left linear', leftlinearmf(x, 25, 75)),
('right linear', rightlinearmf(x, 25, 75)),
# fuzzylite:
('cosine', cosinemf(x, 50, 50)),
('inc concave', concavemf(x, 50, 75)),
('dec concave', concavemf(x, 50, 25)),
('spike', spikemf(x, 50, 50)),
('inc ramp', rampmf(x, 25, 75)),
('dec ramp', rampmf(x, 75, 25)),
# Rectangle-ish
('trapezoid', skmemb.trapmf(x, [20, 40, 60, 80])),
('rectangle', rectanglemf(x, 25, 75)),
('singleton', singletonmf(x, 50)),
])
# Example point sets:
ps_tests = [[(40, 0.5), (60, 1)],
[(10, 0.5), (25, 0.25), (40, 0.75), (80, .5)],
[(0, 1), (40, 0.25), (50, .5), (99, 0)]]
# Now try some interpolation methods on these:
for method in ['linear', 'lagrange', 'spline', 'cubic']:
tests.update([('{} ex{}'.format(method, i), pointsetmf(x, ps, method))
for i, ps in enumerate(ps_tests)])
return tests
def generate_mf_group(self, G, x):
mf_group = {}
for (k, v) in G.iteritems():
shp = v['shp']
mf = v['mf']
if mf == 'trap':
mf_group[k] = trapmf(x, shp)
if mf == 'tri':
mf_group[k] = trimf(x, shp)
if mf == 'gbell':
mf_group[k] = gbellmf(x, shp[0], shp[1], shp[2])
if mf == 'gauss':
mf_group[k] = gaussmf(x, shp[0], shp[1])
if mf == 'gauss2':
mf_group[k] = gauss2mf(x, shp[0], shp[1])
if mf == 'sig':
mf_group[k] = sigmf(x, shp[0], shp[1])
if mf == 'psig':
mf_group[k] = psigmf(x, shp[0], shp[1], shp[2], shp[3])
if mf == 'zmf':
mf_group[k] = zmf(x, shp[0], shp[1], shp[2], shp[3])
if mf == 'smf':
mf_group[k] = smf(x, shp[0], shp[1], shp[2], shp[3])
if mf == 'pimf':
mf_group[k] = pimf(x, shp[0], shp[1], shp[2], shp[3])
if mf == 'piecemf':
mf_group[k] = piecemf(x, shp[0], shp[1], shp[2], shp[3])
return mf_group
def generate_membership(Image_fs):
""" this function accepts Image and its corresponding fuzzy set and generates the membership
values for each pixel based on the input parameters"""
if np.shape(Image_fs)==(512,512):
mean=np.mean(Image_fs) # /|\ figure out hoe to calculate mean using restricted equivaleance
# /_*_\
std=np.std(Image_fs)
#Membership_value=np.zeros((512,512))
#Membership_value=0.582*((np.exp(1-abs(Image_fs-mean)))-1)
#return Membership_value
# k=abs(Image_fs-mean)
# exp=np.exp(1-k)
# val=exp-1
membership_val=membership.gaussmf(Image_fs,mean,std)
if np.amin(membership_val)<0 or np.amax(membership_val)>1:
log.warning('Value Error: membership value out of range')
else:
return membership_val
else:
log.warning('shape of the Image passed to <generate_membership()> is INVALID!')
'above': above,
'any': any_of,
'below': below,
'extremely': extremely,
'intensify': intensify,
'more_or_less': more_or_less,
'norm': norm,
'not': is_not,
'plus': plus,
'seldom': seldom,
'slightly': slightly,
'somewhat': somewhat,
'very': very,
}
def test_all_hedges(y):
results = []
hedgenames = sorted(all_hedges.keys()) # Want them in alphabetical order
for name in hedgenames:
func = all_hedges[name]
results.append(func(y))
return (hedgenames, results)
if __name__ == '__main__':
x = np.arange(0, 100)
y = skmemb.gaussmf(x, 50, 15)
(titles, data) = test_all_hedges(y)
extramf.visualise_all(x, [y] + data, ['original (gaussian)'] + titles)
def __init__(self):
# Create input and output variables
self.lsonar = ctrl.Antecedent(self.sonar_interval, 'lsonar')
self.rsonar = ctrl.Antecedent(self.sonar_interval, 'rsonar')
self.last_theta = ctrl.Antecedent(self.theta_interval, 'last_theta')
self.theta = ctrl.Consequent(self.theta_interval, 'theta')
self.v_x = ctrl.Consequent(self.v_interval, 'v_x')
# we need membership functions!
distance_levels = ['dangerous', 'unsafe', 'safe']
theta_directions = ['left', 'centre', 'right']
velocity_levels = ['slow', 'medium', 'fast']
self.lsonar.automf(names=distance_levels)
self.rsonar.automf(names=distance_levels)
self.last_theta.automf(names=theta_directions)
self.theta.automf(names=theta_directions)
self.v_x.automf(names=velocity_levels)
low_mshp = mshp.trimf(self.sonar_interval, [0.25, 0.25, 0.31])
mid_mshp = mshp.trimf(self.sonar_interval, [0.28, 0.35, 0.45])
high_mshp = mshp.trapmf(self.sonar_interval, [0.4, 0.6, 2.55, 2.55])
self.lsonar['dangerous'] = low_mshp
self.lsonar['unsafe'] = mid_mshp
self.lsonar['safe'] = high_mshp
self.rsonar['dangerous'] = low_mshp
self.rsonar['unsafe'] = mid_mshp
self.rsonar['safe'] = high_mshp
left_angle_mshp = mshp.gaussmf(self.theta_interval, 0.7, 0.2)
centre_angle_mshp = mshp.gaussmf(self.theta_interval, 0, 0.1)
right_angle_mshp = mshp.gaussmf(self.theta_interval, -0.7, 0.2)
self.theta['left'] = left_angle_mshp
self.theta['centre'] = centre_angle_mshp
self.theta['right'] = right_angle_mshp
self.last_theta['left'] = left_angle_mshp
self.last_theta['centre'] = centre_angle_mshp
self.last_theta['right'] = right_angle_mshp
self.v_x['slow'] = mshp.trimf(self.v_interval, [0, 0, 0.32])
self.v_x['medium'] = mshp.gaussmf(self.v_interval, 0.5, 0.36)
self.v_x['fast'] = mshp.trapmf(self.v_interval, [0.75, 0.95, 1.0, 1.0])
# Here the rules are defined!
# look at http://pythonhosted.org/scikit-fuzzy/auto_examples/plot_control_system_advanced.html#set-up-the-fuzzy-control-system
# and http://pythonhosted.org/scikit-fuzzy/auto_examples/plot_tipping_problem_newapi.html
# for good examples
rules = []
# Sonars are NOT dangerous, go fast and forward
rules.append(ctrl.Rule(label="FastAndForward",
antecedent=(~self.lsonar['dangerous'] & ~self.rsonar['dangerous']),
consequent=(self.theta['centre'], self.v_x['fast'])))
# Right sonar reads dangerous, left does not, steer left
rules.append(ctrl.Rule(label="SteerLeftCaution",
antecedent=(~self.lsonar['dangerous'] & self.rsonar['dangerous']),
consequent=(self.theta['left'], self.v_x['slow'])))
# Left sonar reads dangerous, right does not, steer left
rules.append(ctrl.Rule(label="SteerRightCaution",
antecedent=(self.lsonar['dangerous'] & ~self.rsonar['dangerous']),
consequent=(self.theta['right'], self.v_x['slow'])))
# Right sonar is NOT safe, left reads safe, turn left full speed
rules.append(ctrl.Rule(label="SteerLeftReckless",
antecedent=(self.lsonar['safe'] & ~self.rsonar['safe']),
consequent=(self.theta['left'], self.v_x['fast'])))
# Left sonar is NOT safe, right reads safe, turn right full speed
rules.append(ctrl.Rule(label="SteerRightReckless",
antecedent=(~self.lsonar['safe'] & self.rsonar['safe']),
consequent=(self.theta['right'], self.v_x['fast'])))
# Left sonar is NOT dangerous and right is NOT safe, turn left medium speed
rules.append(ctrl.Rule(label="SteerLeftWatchSpeed",
antecedent=(~self.lsonar['dangerous'] & ~self.rsonar['safe']),
consequent=(self.theta['left'], self.v_x['medium'])))
# Right sonar is NOT dangerous and left is NOT safe, turn right medium speed
rules.append(ctrl.Rule(label="SteerRightWatchSpeed",
antecedent=(~self.lsonar['safe'] & ~self.rsonar['dangerous']),
consequent=(self.theta['right'], self.v_x['medium'])))
# Last theta was right, right sonar is dangerous, turn left slow speed
rules.append(ctrl.Rule(label="SteerLeftMedSpeed",
antecedent=(self.rsonar['dangerous'] & self.last_theta['right']),
consequent=(self.theta['left'], self.v_x['medium'])))
# Last theta was left, left sonar is dangerous, turn right slow speed
rules.append(ctrl.Rule(label="SteerRightMedSpeed",
antecedent=(self.lsonar['dangerous'] & self.last_theta['left']),
consequent=(self.theta['right'], self.v_x['medium'])))
# Last theta was left, rsonar reads not dangerous, turn centre, full speed
rules.append(ctrl.Rule(label="OutOfDangerFromLeft",
antecedent=(~self.rsonar['dangerous'] & self.last_theta['left']),
consequent=(self.theta['centre'], self.v_x['fast'])))
# Last theta was right, lsonar reads not dangerous, turn centre, full speed
rules.append(ctrl.Rule(label="OutOfDangerFromRight",
antecedent=(~self.lsonar['dangerous'] & self.last_theta['right']),
consequent=(self.theta['centre'], self.v_x['fast'])))
# Create the control system
self.system = ctrl.ControlSystem(rules=rules)
# System is created, just create a simulation, input values and call compute()
self.sim = ctrl.ControlSystemSimulation(self.system)
def test_gaussmf():
x = np.arange(-4, 5.1, 0.1)
expected = np.exp(-(x - 1.33)**2 / (2 * 0.45**2))
test = gaussmf(x, 1.33, 0.45)
assert_allclose(test, expected)
def test_gaussmf():
x = np.arange(-4, 5.1, 0.1)
expected = np.exp(- (x - 1.33)**2 / (2 * 0.45**2))
test = gaussmf(x, 1.33, 0.45)
assert_allclose(test, expected)
def field(A0,Td,w,t): #potenziale vettore assumendo
return A0 *(((np.cos((t-Td/2)*pi/Td))**2)* np.sin(w*(t-Td/2)) * fuzzy.gaussmf(t-Td/2,0,Td/4) *
fuzzy.gbellmf(t-Td/2,Td/4,7,0)) #inviluppo cos^2
def rightgaussmf(x, mean, sigma):
''' Like Gaussian, but always 1 when >= mean (so, slopes up only)'''
mf = skmemb.gaussmf(x, mean, sigma)
mf[np.nonzero(x >= mean)] = 1
return mf
def leftgaussmf(x, mean, sigma):
''' Like Gaussian, but always 1 when <= mean (so, slopes down only)'''
mf = skmemb.gaussmf(x, mean, sigma)
mf[np.nonzero(x <= mean)] = 1
return mf
