-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmodel_learning_preprocessing_yumi_big_data.py
291 lines (248 loc) · 8.77 KB
/
model_learning_preprocessing_yumi_big_data.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
279
280
281
282
283
284
285
286
287
288
289
290
291
import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import axes3d
import pickle
from sklearn.preprocessing import StandardScaler
from YumiKinematics import YumiKinematics
from mjc_exp_policy import kin_params, exp_params_rob
# from pykdl_utils.kdl_kinematics import *
from model_leraning_utils import obtian_joint_space_policy, get_ee_points
# data from yumi exp
logfile_ip = "./Results/yumi_peg_exp_new_raw_data_train.p"
p2_logfile_ip = "./Results/yumi_peg_exp_new_raw_data_test_p2.p"
p5_logfile_ip = "./Results/yumi_peg_exp_new_raw_data_test_p5.p"
p10_logfile_ip = "./Results/yumi_peg_exp_new_raw_data_test_p10.p"
m2_logfile_ip = "./Results/yumi_peg_exp_new_raw_data_test_m2.p"
m5_logfile_ip = "./Results/yumi_peg_exp_new_raw_data_test_m5.p"
m10_logfile_ip = "./Results/yumi_peg_exp_new_raw_data_test_m10.p"
# logfile_op = "./Results/Final/yumi_peg_exp_new_preprocessed_data_train_big_data_wom10_fixed.p"
logfile_op = "./Results/Final/yumi_peg_exp_new_preprocessed_data_train_small_data_wom10_fixed.p"
test_flag = False
# pol_per_factor = -0.02
pol_per_factor = -0.1
# data from mjc exp
# logfile_ip = "./Results/mjc_exp_2_sec_raw.p"
# logfile_op = "./Results/mjc_exp_2_sec_raw_preprocessed.p"
yumiKin = YumiKinematics(kin_params)
# Overwrite wrist data with no action
overwrite_j7_data = False
p_7 = -2.1547
v_7 = 0.0
u_7 = 0.0
n_train = 40
# big data train data
exp_data = pickle.load(open(logfile_ip, "rb"))
Xs_ = exp_data['X'] # state
Us_ = exp_data['U']
# Xs_0 = Xs_[1:11]
# Us_0 = Us_[1:11]
Xs_0 = Xs_[1:]
Us_0 = Us_[1:]
p2_exp_data = pickle.load(open(p2_logfile_ip, "rb"))
Xs_p2 = p2_exp_data['X'] # state
Us_p2 = p2_exp_data['U']
Xs_p2 = Xs_p2[1:]
Us_p2 = Us_p2[1:]
p5_exp_data = pickle.load(open(p5_logfile_ip, "rb"))
Xs_p5 = p5_exp_data['X'] # state
Us_p5 = p5_exp_data['U']
Xs_p5 = Xs_p5[1:]
Us_p5 = Us_p5[1:]
p10_exp_data = pickle.load(open(p10_logfile_ip, "rb"))
Xs_p10 = p10_exp_data['X'] # state
Us_p10 = p10_exp_data['U']
Xs_p10 = Xs_p10[1:]
Us_p10 = Us_p10[1:]
m10_exp_data = pickle.load(open(m10_logfile_ip, "rb"))
Xs_m10 = m10_exp_data['X'] # state
Us_m10 = m10_exp_data['U']
Xs_m10 = Xs_m10[1:]
Us_m10 = Us_m10[1:]
m5_exp_data = pickle.load(open(m5_logfile_ip, "rb"))
Xs_m5 = m5_exp_data['X'] # state
Us_m5 = m5_exp_data['U']
Xs_m5 = Xs_m5[1:]
Us_m5 = Us_m5[1:]
m2_exp_data = pickle.load(open(m2_logfile_ip, "rb"))
Xs_m2 = m2_exp_data['X'] # state
Us_m2 = m2_exp_data['U']
Xs_m2 = Xs_m2[1:]
Us_m2 = Us_m2[1:]
# Xs_train = np.concatenate((Xs_0, Xs_p2, Xs_p5, Xs_p10, Xs_m2, Xs_m5, Xs_m10), axis=0)
# Us_train = np.concatenate((Us_0, Us_p2, Us_p5, Us_p10, Us_m2, Us_m5, Us_m10), axis=0)
# Xs_test = Xs_[11:]
# Us_test = Us_[11:]
# # p10
# Xs_train = np.concatenate((Xs_0, Xs_p2, Xs_p5, Xs_m2, Xs_m5, Xs_m10), axis=0)
# Us_train = np.concatenate((Us_0, Us_p2, Us_p5, Us_m2, Us_m5, Us_m10), axis=0)
# Xs_test = Xs_p10
# Us_test = Us_p10
# m10
Xs_train = np.concatenate((Xs_0, Xs_p2, Xs_p5, Xs_m2, Xs_m5, Xs_p10), axis=0)
Us_train = np.concatenate((Us_0, Us_p2, Us_p5, Us_m2, Us_m5, Us_p10), axis=0)
Xs_test = Xs_m10
Us_test = Us_m10
# EXs = exp_data['EX'] # state
# Fs = exp_data['F']
exp_params = exp_data['exp_params']
dP = exp_params['dP']
dV = exp_params['dV']
dU = exp_params['dU']
Xs = np.concatenate((Xs_train, Xs_test), axis=0)
Us = np.concatenate((Us_train, Us_test), axis=0)
N, T, dX = Xs.shape
Qts = Xs[:, :, :dP]
Qts_d = Xs[:, :, dP:dP+dV]
Uts = Us
Ets = np.zeros((N, T, 6))
Ets_d = np.zeros((N, T, 6))
Ets_ee = np.zeros((N, T, 9))
Ets_d_ee = np.zeros((N, T, 9))
Fts = np.zeros((N, T, 6))
for n in range(N):
for i in range(T):
Ets[n, i] = yumiKin.fwd_pose(Qts[n,i])
J_A = yumiKin.get_analytical_jacobian(Qts[n,i])
Ets_d[n,i] = J_A.dot(Qts_d[n,i])
Fts[n,i] = np.linalg.pinv(J_A.T).dot(Uts[n,i])
ee_offsets = kin_params['ee_offsets']
epos, erot = yumiKin.get_fwd_mat(Qts[n, i])
ee_points = get_ee_points(ee_offsets, epos.reshape(1, -1), erot)
ee_points = ee_points.T
Ets_ee[n, i] = ee_points.reshape(-1)
J_ee_1 = yumiKin.kdl_kin.jacobian(Qts[n, i], ee_points[0])
J_ee_2 = yumiKin.kdl_kin.jacobian(Qts[n, i], ee_points[1])
J_ee_3 = yumiKin.kdl_kin.jacobian(Qts[n, i], ee_points[2])
J_ee_1 = J_ee_1[:3, :]
J_ee_2 = J_ee_2[:3, :]
J_ee_3 = J_ee_3[:3, :]
ee_vel_1 = J_ee_1.dot(Qts_d[n, i])
ee_vel_2 = J_ee_2.dot(Qts_d[n, i])
ee_vel_3 = J_ee_3.dot(Qts_d[n, i])
Ets_d_ee[n, i] = np.concatenate((ee_vel_1, ee_vel_2, ee_vel_3), axis=1)
EXs = np.concatenate((Ets,Ets_d),axis=2)
Fs = Fts
EXs_ee = np.concatenate((Ets_ee,Ets_d_ee),axis=2)
dt = exp_params['dt']
n_trials, n_time_steps, dim_state = Xs.shape
N = n_trials
_, _, dim_action = Us.shape
assert(dX==dim_state)
assert(n_time_steps==T)
assert(dU==dim_action)
plt.figure()
tm = np.linspace(0, T * dt, T)
# jPos
for j in range(7):
plt.subplot(3, 7, 1 + j)
plt.title('j%dPos' % (j + 1))
plt.plot(tm, Xs[:,:, j].T, color='g', alpha=0.3)
plt.plot(tm, np.mean(Xs[:, :, j], axis=0), color='g', linestyle='--')
# jVel
for j in range(7):
plt.subplot(3, 7, 8 + j)
plt.title('j%dVel' % (j + 1))
plt.plot(tm, Xs[:, :, dP+j].T, color='b', alpha=0.3)
plt.plot(tm, np.mean(Xs[:, :, dP + j], axis=0), color='b', linestyle='--')
# jTrq
for j in range(7):
plt.subplot(3, 7, 15 + j)
plt.title('j%dTrq' % (j + 1))
plt.plot(tm, Us[:, :, j].T, color='r', alpha=0.3)
plt.plot(tm, np.mean(Us[:, :, j], axis=0), color='r', linestyle='--')
plt.show(block=False)
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1, projection='3d')
ax.set_xlabel('X')
ax.set_ylabel('Y')
ax.set_zlabel('Z')
for i in range(N):
# ax.scatter3D(EXs[i,:,0], EXs[i,:,1], EXs[i,:,2], marker='$'+str(i)+'$', s=20)
# ax.scatter3D(EXs[i, :, 0], EXs[i, :, 1], EXs[i, :, 2], color='r')
ax.scatter3D(EXs_ee[i, :, 0], EXs_ee[i, :, 1], EXs_ee[i, :, 2], color='b')
ax.scatter3D(EXs_ee[i, :, 3], EXs_ee[i, :, 4], EXs_ee[i, :, 5], color='g')
ax.scatter3D(EXs_ee[i, :, 6], EXs_ee[i, :, 7], EXs_ee[i, :, 8], color='m')
#plt.show()
# plt.show()
n_train = 40
n_test = 5
exp_data['n_train'] = n_train
exp_data['n_test'] = n_test
XUs = np.concatenate((Xs, Us), axis=2)
XUs_train = XUs[:n_train, :, :]
exp_data['XUs_train'] = XUs_train
XUs_t_train = XUs_train[:,:-1,:]
exp_data['XUs_t_train'] = XUs_t_train
XU_t_train_avg = np.mean(XUs_t_train, axis=0)
exp_data['XU_t_train_avg'] = XU_t_train_avg
Xs_t_train = XUs_train[:,:-1,:dX]
exp_data['Xs_t_train'] = Xs_t_train
Xs_t1_train = XUs_train[:,1:,:dX]
exp_data['Xs_t1_train'] = Xs_t1_train
Xs_t1_train_avg = np.mean(Xs_t1_train, axis=0)
exp_data['Xs_t1_train_avg'] = Xs_t1_train_avg
Us_t_train = XUs_train[:,:-1,dX:dX+dU]
exp_data['Us_t_train'] = Us_t_train
params = {
'kp': exp_params_rob['Kp'],
'kd': exp_params_rob['Kd'],
'dX': dX,
'dP': dP,
'dV': dV,
'dU': dU,
'dt': dt,
}
x_init = exp_params_rob['x0']
Xrs_t_train = obtian_joint_space_policy(params, XUs_t_train[:10], x_init)
exp_data['Xrs_t_train'] = Xrs_t_train
EXFs = np.concatenate((EXs, Fs), axis=2)
EXFs_train = EXFs[:n_train, :, :]
exp_data['EXFs_train'] = EXFs_train
EXFs_t_train = EXFs_train[:,:-1,:]
exp_data['EXFs_t_train'] = EXFs_t_train
EXs_t_train = EXFs_train[:,:-1,:12]
exp_data['EXs_t_train'] = EXs_t_train
EXs_t1_train = EXFs_train[:,1:,:12]
exp_data['EXs_t1_train'] = EXs_t1_train
Fs_t_train = EXFs_train[:,:-1,12:]
exp_data['Fs_t_train'] = Fs_t_train
EXs_ee_t_train = EXs_ee[:n_train, :-1, :]
exp_data['EXs_ee_t_train'] = EXs_ee_t_train
EXs_ee_t1_train = EXs_ee[:n_train, 1:, :]
exp_data['EXs_ee_t1_train'] = EXs_ee_t1_train
XUs_test = XUs[n_train:, :, :]
XUs_t_test = XUs_test[:,:-1,:]
exp_data['XUs_t_test'] = XUs_t_test
Xs_t1_test = XUs_test[:,1:,:dX]
exp_data['Xs_t1_test'] = Xs_t1_test
Xs_t_test = XUs_test[:,:-1,:dX]
exp_data['Xs_t_test'] = Xs_t_test
Us_t_test = XUs_test[:,:-1,dX:dX+dU]
exp_data['Us_t_test'] = Us_t_test
x_init = exp_params_rob['x0']
kp = params['kp']
params['kp'] = kp + kp*pol_per_factor
Xrs_t_test = obtian_joint_space_policy(params, XUs_t_test, x_init)
exp_data['Xrs_t_test'] = Xrs_t_test
X0s = XUs_train[:, 0, :dX]
X0_mu = np.mean(X0s, axis=0)
exp_data['X0_mu'] = X0_mu
X0_var = np.var(X0s, axis=0)
exp_data['X0_var'] = X0_var
EXFs_test = EXFs[n_train:, :, :]
EXFs_t_test = EXFs_test[:,:-1,:]
exp_data['EXFs_t_test'] = EXFs_t_test
EXs_t1_test = EXFs_test[:,1:,:12]
exp_data['EXs_t1_test'] = EXs_t1_test
EXs_t_test = EXFs_test[:,:-1,:12]
exp_data['EXs_t_test'] = EXs_t_test
EXs_ee_t_test = EXs_ee[n_train:, :-1, :]
exp_data['EXs_ee_t_test'] = EXs_ee_t_test
EXs_ee_t1_test = EXs_ee[n_train:, 1:, :]
exp_data['EXs_ee_t1_test'] = EXs_ee_t1_test
EX0s = EXFs_train[:, 0, :12]
EX0_mu = np.mean(EX0s, axis=0)
exp_data['EX0_mu'] = EX0_mu
EX0_var = np.var(EX0s, axis=0)
exp_data['EX0_var'] = EX0_var
pickle.dump(exp_data, open(logfile_op, "wb"))