-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathNN_train.py
279 lines (205 loc) · 9.01 KB
/
NN_train.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
import math
import random
import keras
import numpy as np
import tensorflow as tf
from matplotlib import patches as mpatches
from keras.callbacks import TensorBoard
from matplotlib import pyplot as plt
from sklearn import preprocessing
linkLength = 2
Q1 = []
Q2 = []
Q3 = []
posX = []
posY = []
titaEnd = []
samples = 1000
file = open("/home/abin/my_works/github_works/neural-network-learning/dataset/training_data.csv", "w")
def Xe(q1, q2, q3):
return linkLength*math.cos(q1) + linkLength*math.cos(q1+q2) + linkLength*math.cos(q1+q2+q3)
def Ye(q1, q2, q3):
return linkLength*math.sin(q1) + linkLength*math.sin(q1+q2) + linkLength*math.sin(q1+ q2 + q3)
def tita(q1, q2, q3):
return math.degrees(q1) + math.degrees(q2) + math.degrees(q3)
def build_model():
model = keras.Sequential()
model.add(keras.layers.Dense(100,use_bias=True,activation='tanh'))
model.add(keras.layers.Dense(3,use_bias=True,activation='linear'))
model.compile(optimizer=tf.compat.v1.train.AdamOptimizer(0.05), loss=keras.losses.mean_squared_error, metrics=['accuracy'])
return model
for i in range(0,samples):
q1 = round(random.uniform(0, math.pi),2)
q2 = round(random.uniform(-math.pi,0),2)
q3 = round(random.uniform(-math.pi/2, math.pi/2),2)
Q1.append(q1)
file.write(str(q1))
file.write(",")
Q2.append(q2)
file.write(str(q2))
file.write(",")
Q3.append(q3)
file.write(str(q3))
file.write(",")
X = Xe(q1,q2,q3)
posX.append(X)
file.write(str(round(X, 2)))
file.write(",")
Y = Ye(q1,q2,q3)
posY.append(Y)
file.write(str(round(Y, 2)))
file.write(",")
T = tita(q1,q2,q3)
titaEnd.append(T)
file.write(str(round(T,2)))
file.write("\n")
file.close()
for i in range(0, len(posX)):
plt.plot([posX[i], posX[i] + 0.2*math.cos(math.radians(titaEnd[i]))], [posY[i], posY[i]+0.2*math.sin(math.radians(titaEnd[i]))])
plt.scatter(posX,posY)
plt.xlabel("X axis")
plt.ylabel("Y axis")
plt.title("Data set of 1000 possible endeffector positions and orientations")
plt.show()
dataMat = np.c_[Q1,Q2,Q3,posX,posY,titaEnd]
for i in range(0,samples):
check1 = dataMat[i, 3]
check2 = dataMat[i, 4]
check3 = dataMat[i, 5]
for j in range(0, samples):
if i != j:
if(dataMat[j, 3] == check1 and dataMat[j, 4] == check2 and dataMat[j, 5] == check3):
print(i, j, dataMat[j, 3], dataMat[j, 4], dataMat[j, 5])
data = dataMat[:, [3,4,5]] #X,Y,tita
output = dataMat[:, [0,1,2]] #Q1,Q2,Q3
train_input = data[0:int(0.7*samples), :]
train_output = output[0:int(0.7*samples), :]
test_input = data[int(0.7*samples):int(0.85*samples), :]
test_output = output[int(0.7*samples):int(0.85*samples), :]
validate_input = data[int(0.85*samples):int(samples), :]
validate_output = output[int(0.85*samples):int(samples), :]
print("Train Input--------------------")
print(np.shape(train_input))
print(train_input)
print("Output-----------------------")
print(np.shape(output))
print(output)
x_scaler = preprocessing.MinMaxScaler(feature_range=(-1,1))
y_scaler = preprocessing.MinMaxScaler(feature_range=(-1,1))
x_scaler_test = preprocessing.MinMaxScaler(feature_range=(-1,1))
y_scaler_test = preprocessing.MinMaxScaler(feature_range=(-1,1))
x_scaler_eva = preprocessing.MinMaxScaler(feature_range=(-1,1))
y_scaler_eva = preprocessing.MinMaxScaler(feature_range=(-1,1))
dataX = x_scaler.fit_transform((train_input))
dataY = y_scaler.fit_transform(train_output)
dataX_test = x_scaler_test.fit_transform(test_input)
dataY_test = y_scaler_test.fit_transform(test_output)
dataX_eva = x_scaler_eva.fit_transform(validate_input)
dataY_eva = y_scaler_eva.fit_transform(validate_output)
NAME = "Trajectry Tracking"
tensorboard = TensorBoard(log_dir="logs/{}".format(NAME))
model = build_model()
history = model.fit(dataX, dataY, nb_epoch=100, callbacks=[tensorboard]) #train the model
[loss, mae] = model.evaluate(dataX_test,dataY_test, verbose=0) #Evaluation
print("Testing set Mean Abs Error: ${:7.2f}".format(mae))
dataX_input = x_scaler.transform(validate_input)
test_prediction = model.predict(dataX_input)
real_prediction = y_scaler.inverse_transform(test_prediction)
plt.clf()
plt.scatter(validate_output[:, 0], real_prediction[:, 0], c='b')
plt.scatter(validate_output[:, 1], real_prediction[:, 1],c='g')
plt.scatter(validate_output[:, 2], real_prediction[:, 2], c='r')
plt.xlabel('True Values angles in rad')
plt.ylabel('Predictions angles in rad')
plt.title("True Value Vs Prediction")
plt.legend("If all predicted values equal to the desired(true) value, this will be lie on 45 degree line")
plt.show()
print("***************************")
print(validate_input[100,0], " ", validate_input[100,1])
print(Xe(real_prediction[100,0], real_prediction[100,1], real_prediction[100,2]), " "
, Ye(real_prediction[100,0], real_prediction[100, 1], real_prediction[100,2]))
print("***************************")
single_data_1 = np.array([[5,2,60]])
single_data = x_scaler.transform((single_data_1))
single_prediction = model.predict(single_data)
single_real_prediction = y_scaler.inverse_transform(single_prediction)
print(single_data_1[0,0], " ", single_data_1[0,1])
print(Xe(single_real_prediction[0,0], single_real_prediction[0,1], single_real_prediction[0,2]), " "
, Ye(single_real_prediction[0,0], single_real_prediction[0,1], single_real_prediction[0,2]))
print("****************************")
Xc = 0
Yc = 0
r =2
data_points = 100
Input_Circle = np.zeros((data_points, 3), float)
Output_Circle= np.zeros((data_points, 3), float)
Single_input = np.zeros((1,3), float)
titaz = np.linspace(0,6,num=data_points)
tagectory = []
for i in range(0, len(titaz)):
Input_Circle[i][0]=Xc + titaz[i]
Input_Circle[i][1]=Yc + titaz[i]
Input_Circle[i][2]= 90
plt.clf()
plt.scatter(Input_Circle[:,0], Input_Circle[:, 1], c = 'b')
plt.xlabel("X axis")
plt.ylabel("Y axis")
plt.title("Desired Tragectory Coordinates and Predicted ")
inin = np.zeros((1,3), float)
inin[0][0] = Input_Circle[0,0]
inin[0][1] = Input_Circle[0,1]
inin[0][2] = Input_Circle[0,2]
Predicted_coordinates = np.zeros((data_points, 3), float)
print(np.shape(Input_Circle))
print(single_data_1[0,0], " ", single_data_1[0,1])
print(Xe(single_real_prediction[0,0], single_real_prediction[0,1], single_real_prediction[0,2]), " "
, Ye(single_real_prediction[0,0], single_real_prediction[0,1], single_real_prediction[0,2]))
print("********************************")
Joint_angle_predict = np.zeros((data_points,4), float)
Error = np.zeros((data_points, 7), float)
Tita_hat= 0
for q in range(0, data_points):
single_data_1 = np.array([[Input_Circle[q, 0], Input_Circle[q,1], Input_Circle[q,2]]])
single_data = x_scaler.transform(single_data_1)
single_prediction = model.predict(single_data)
single_real_prediction = y_scaler.inverse_transform(single_prediction)
X_hat = Xe(single_real_prediction[0, 0], single_real_prediction[0,1], single_real_prediction[0,2])
Y_hat = Ye(single_real_prediction[0,0], single_real_prediction[0,1], single_real_prediction[0,2])
Tita_hat = tita(single_real_prediction[0,0], single_real_prediction[0,1], single_real_prediction[0,2])
Joint_angle_predict[q][0] = q
Joint_angle_predict[q][1] = single_real_prediction[0,0]
Joint_angle_predict[q][2] = single_real_prediction[0,1]
Joint_angle_predict[q][3] = single_real_prediction[0,2]
Error[q][0] = Input_Circle[q, 0] - X_hat
Error[q][1] = Input_Circle[q, 1] - Y_hat
Error[q][2] = Input_Circle[q, 2] - Tita_hat
Error[q][3] = math.degrees(single_real_prediction[0,0])
Error[q][4] = math.degrees(single_real_prediction[0,1])
Error[q][5] = math.degrees(single_real_prediction[0,2])
Error[q][6] = q
print("X: ", Input_Circle[q, 0], " Y: ", Input_Circle[q, 1], "Tita: ", Input_Circle[q, 2])
print("X^:", X_hat, " Y^:", Y_hat, " Tita^:", Tita_hat)
print(" ")
plt.scatter(X_hat, Y_hat, c='r')
plt.savefig('Desired Tragectory Cordinates and Predicted.png')
Q1_patch = mpatches.Patch(color='red', label='1st joint')
Q2_patch = mpatches.Patch(color='blue', label='2nd joint')
Q3_patch = mpatches.Patch(color='green', label='3rd joint')
plt.clf()
plt.plot(Joint_angle_predict[:, 0], Joint_angle_predict[:, 1], c='r')
plt.plot(Joint_angle_predict[:, 1], Joint_angle_predict[:, 2], c='b')
plt.plot(Joint_angle_predict[:, 2], Joint_angle_predict[:, 3], c='g')
plt.legend(handles = [Q1_patch, Q2_patch, Q3_patch])
plt.title('Joint angle variation over data points')
plt.savefig('Joint angle variation over data points.png')
P1_patch = mpatches.Patch(color='red', label='Error in X coordinates')
P2_patch = mpatches.Patch(color='blue', label='Error in Y coordinates')
P3_patch = mpatches.Patch(color='green', label='Error in Tita ')
plt.clf()
plt.plot(Error[:, 6], Error[:, 0], c='r')
plt.plot(Error[:, 6], Error[:, 1], c='b')
plt.plot(Error[:, 6], Error[:, 2], c='g')
plt.title('Error of X, Y and tita in the evolution')
plt.legend(handles = [P1_patch, P2_patch, P3_patch])
plt.savefig('Error of X,Y and Tita in the evolution')
plt.show()