def load(rfl, testper = 0.08, verbose = 0, traindataname = 'traindata', testdataname = 'testdata'):
a = []
for rfn in rfl:
b = readraytracerout(rfn, verbose = verbose, indice = [2,3,4,5,6])
if a == []:
a = b
else:
a = np.append(a,b,axis=0)
# inputs
ty = a[:,1]
ty1 = ty + 15
ty2 = 15 - ty
tz = a[:,2]
rx = a[:,3]
ry = a[:,4]
ry1 = ry + 15
ry2 = 15 - ry
rz = a[:,5]
# +2 relationship
y = a[:,10] # result in dB
muB = a[:,12]
muS = a[:,13]
stdB = a[:,14]
stdS = a[:,15]
width = a[:,16]
dg = np.sqrt( (ty - ry) ** 2 + (tz-rz) ** 2 + rx ** 2)
dl = np.sqrt( (ty1 - ry1) ** 2 + (tz-rz) ** 2 + rx ** 2 )
dr = np.sqrt( (ty2 - ry2) ** 2 + (tz-rz) ** 2 + rx ** 2 )
lty1, lty2, lry1, lry2 = np.log10([ty1,ty2,ry1,ry2])
ltz, lrz, lrx, ldg, ldl, ldr = np.log10([tz,rz,rx,dg,dl,dr])
muB = np.log10(muB)
muS = np.log10(muS)
width = np.log10(width)
#[lty1,lty2,lrx,ltz,lry1,lry2,lrz,ldg,ldl,ldr, muB,muS,stdB,stdS]
#normalize([lty1,lty2,lrx,ltz,lry1,lry2,lrz,ldg,ldl,ldr, muB,muS,stdB,stdS])
x = np.stack((lty1,lty2,lrx,ltz,lry1,lry2,lrz,ldg,ldl,ldr, muB,muS,stdB,stdS))
x = x.T
y = -y
xtrain, ytrain, xtest, ytest = seperate_test_set(x,y,testper)
traindata = mydata(xtrain,ytrain,traindataname)
testdata = mydata(xtest,ytest,testdataname)
print("LOAD: Training data size is: ", ytrain.size)
print("LOAD: Testing data size is: ", ytest.size)
print("LOAD: One sample of the data is:\n")
print(str(xtrain[10]) + ' ' + str(ytrain[10]))
traindata.save()
testdata.save()
print("LOAD: Saved traindata to ./data/" + traindataname)
print("LOAD: Saved testdata to ./data/" + testdataname)
return 0
def load5in(rfl,
testper=0.08,
verbose=0,
traindataname='traindata',
testdataname='testdata'):
a = []
for rfn in rfl:
a = readraytracerout(rfn, verbose=verbose)
h1r = a[:, 2] # height of the source
d = a[:, 3] # distance
h2r = a[:, 5] # height of the receiver
y = a[:, 8] # result in dB
#d = np.sqrt(d*d+(h1-h2)*(h1-h2))
d = np.log10(d)
h1 = np.log10(h1r)
h2 = np.log10(h2r)
#normalize everything
dmean = np.mean(d)
h1mean = np.mean(h1)
h2mean = np.mean(h2)
h1rmean = np.mean(h1r)
h2rmean = np.mean(h2r)
dstd = np.std(d)
h1std = np.std(h1)
h2std = np.std(h2)
h1rstd = np.std(h1r)
h2rstd = np.std(h2r)
# correct variance:
# d = (d - dmean) / dstd
# h1 = (h1 - h1mean) / h1std
# h2 = (h2 - h2mean) / h2std
# h1r = (h1r - h1rmean) / h1rstd
# h2r = (h2r - h2rmean) / h2rstd
#x = np.stack((h1,h2,d))
x = np.stack((h1, h2, d, h1r, h2r))
exp1 = 40 * d - 20 * h1 - 20 * h2
x = x.T
y = -y
xtrain, ytrain, xtest, ytest = seperate_test_set(x, y, testper)
traindata = mydata(xtrain, ytrain, traindataname)
testdata = mydata(xtest, ytest, testdataname)
print("LOAD: Training data size is: ", ytrain.size)
print("LOAD: Testing data size is: ", ytest.size)
print("LOAD: One sample of the data is:\n")
print(str(xtrain[1]) + ' ' + str(ytrain[1]))
traindata.save()
testdata.save()
print("LOAD: Saved traindata to ./data/" + traindataname)
print("LOAD: Saved testdata to ./data/" + testdataname)
return 0
def load(rfl,
testper=0.08,
verbose=0,
traindataname='traindata',
testdataname='testdata'):
a = []
for rfn in rfl:
a = readraytracerout(rfn, verbose=verbose)
h1 = a[:, 2] # height of the source
d = a[:, 3] # distance
h2 = a[:, 5] # height of the receiver
y = a[:, 8] # result in dB
if verbose:
freq = a[:, -3]
epsilon = a[:, -2]
sigma = a[:, -1]
d = np.sqrt(d * d + (h1 - h2) * (h1 - h2))
d = np.log10(d)
h1 = np.log10(h1)
h2 = np.log10(h2)
dstd = np.std(d)
h1std = np.std(h1)
h2std = np.std(h2)
print("Std. of d is: ", dstd)
print("Std. of h1 is: ", h1std)
print("Std. of h2 is: ", h2std)
# correct variance:
h1cor = dstd / h1std
h2cor = dstd / h2std
#h1 = h1*h1cor
#h2 = h2*h2cor
exp1 = 40 * d - 20 * h1 - 20 * h2
x = np.stack((h1, h2, d))
#x = np.stack((h1,h2,d))
x = x.T
y = -y
xtrain, ytrain, xtest, ytest = seperate_test_set(x, y, testper)
traindata = mydata(xtrain, ytrain, traindataname)
testdata = mydata(xtest, ytest, testdataname)
print("LOAD: Training data size is: ", ytrain.size)
print("LOAD: Testing data size is: ", ytest.size)
print("LOAD: One sample of the data is:\n")
print(str(xtrain[1:10]) + ' ' + str(ytrain[1]))
traindata.save()
testdata.save()
print("LOAD: Saved traindata to ./data/" + traindataname)
print("LOAD: Saved testdata to ./data/" + testdataname)
return 0
def load(rfl,
testper=0.08,
verbose=0,
traindataname='traindata',
testdataname='testdata'):
a = []
for rfn in rfl:
a = readraytracerout(rfn, verbose=verbose)
# inputs
ty = a[:, 1]
ty1 = ty + 15
ty2 = 15 - ty
tz = a[:, 2]
rx = a[:, 3]
ry = a[:, 4]
ry1 = ry + 15
ry2 = 15 - ry
rz = a[:, 5]
dg = np.sqrt((ty - ry)**2 + (tz - rz)**2 + rx**2)
dl = np.sqrt((ty1 - ry1)**2 + (tz - rz)**2 + rx**2)
dr = np.sqrt((ty2 - ry2)**2 + (tz - rz)**2 + rx**2)
lty1, lty2, lry1, lry2 = np.log10([ty1, ty2, ry1, ry2])
ltz, lrz, lrx, ldg, ldl, ldr = np.log10([tz, rz, rx, dg, dl, dr])
# outputs
y = a[:, 10]
# normalize
#means, stds = normalize([lty1,lty2,lrx,ltz,lry1,lry2,lrz,ldg,ldl,ldr])
x = np.stack((lty1, lty2, lrx, ltz, lry1, lry2, lrz, ldg, ldl, ldr))
#x = np.stack((h1,h2,d))
x = x.T
y = -y
xtrain, ytrain, xtest, ytest = seperate_test_set(x, y, testper)
traindata = mydata(xtrain, ytrain, traindataname)
testdata = mydata(xtest, ytest, testdataname)
print("LOAD: Training data size is: ", ytrain.size)
print("LOAD: Testing data size is: ", ytest.size)
print("LOAD: One sample of the data is:\n")
print(str(xtrain[1]) + ' ' + str(ytrain[1]))
traindata.save()
testdata.save()
print("LOAD: Saved traindata to ./data/" + traindataname)
print("LOAD: Saved testdata to ./data/" + testdataname)
return 0
result = mlp3L('model3L0',testmode = 1, mlp_ninput = 10, epochs = 10000, breaklim = 18, plot = 0,
mlp_learning_rate = 0.06, traindata = 'uctrainsetr1', testdata = dataname,
plotrangel = 150, plotrangeh = 200, mlp_nbatchs = 5,
mlp_nhidden1 = 50, mlp_nhidden2=30, mlp_nhidden3=20, plotTitle = '900MHz urban canyon model', normalize = True);
# linear fit
#print(result[0])
combined = list(zip(xlist, result[0]))
combined.sort()
xlist, result[0] = zip(*combined)
# get raytracer result
xray = dump_main(rxfn = 'ucpl_rx_pts.txt', txfn = 'ucpl_tx_pts.txt',
envfn = 'env_urbancanyon.txt', ty = ty, tz = tz, ry = ry, rz = rz)
execute_main()
a = readraytracerout('plresult', verbose=0)
yray = -a[:,11]
combined = list(zip(xray, yray))
combined.sort()
xray, yray = zip(*combined)
regr = linear_model.LinearRegression()
regr.fit(xlist, result[0])
yfit = regr.predict(xlist)
plt.suptitle('Urban canyon path loss curve')
plt.xlabel('log10(x)')
plt.ylabel('negative signal strength in db')
plt.plot(xlist,result[0],'b',xlist,yfit,'y',xray,yray,'r')
plt.legend(['Predicted','Linear fit','Raytracer'])