-
Notifications
You must be signed in to change notification settings - Fork 10
/
COMMAG_FIG_4_6.py
1486 lines (1016 loc) · 51.6 KB
/
COMMAG_FIG_4_6.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
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
from __future__ import absolute_import, division, print_function
import tensorflow as tf
import numpy as np
from tensorflow import keras
from tensorflow.keras import layers
from itertools import product
np.set_printoptions(precision=2)
import time
import itertools
'''
Find the feasible location of users
This function finds the feasible location of INITIALIZATION
In this function, the location of RX is randomly chosen such that its distance with
TX is within Dist_TX_RX
'''
def Feasible_Loc_Init(Cur_loc, Size_area, Dist_TX_RX):
temp_dist = Dist_TX_RX * (np.random.rand(1, 2) - 0.5)
temp_chan = Cur_loc + temp_dist
while (np.max(abs(temp_chan)) > Size_area / 2) | (np.linalg.norm(temp_dist) > Dist_TX_RX):
temp_dist = Dist_TX_RX * (np.random.rand(1, 2) - 0.5)
temp_chan = Cur_loc + temp_dist
return temp_chan
'''
Find the feasible location of users
This function finds the feasible location of LOCATION UPDATE
In this function, the location of RX is randomly chosen such that its distance with
TX is within Dist_TX_RX AND the updated location is Delta_mov away from current RX location
'''
def Feasible_Loc_Update(Cur_RX_loc, Cur_TX_loc, Size_area, Dist_TX_RX, Delta_mov):
temp_chan = 0
temp_dist = 2 * Dist_TX_RX
while (np.max(abs(temp_chan)) > Size_area / 2) | (np.linalg.norm(temp_dist) > Dist_TX_RX):
temp_dir = np.random.rand()
temp_dist_delta = [Delta_mov * np.cos(2 * np.pi * temp_dir), Delta_mov * np.sin(2 * np.pi * temp_dir)]
temp_chan = Cur_RX_loc + temp_dist_delta
temp_dist = Cur_TX_loc - temp_chan
return temp_chan
'''
Initialization of location information.
It will return the location of one PU and SUs.
The Users will be allocated to 2D area whose range is -Size_area/2 ~ Size_area/2.
The distance between RX and TX for the same transmit pair is limited to "Dist_TX_RX"
Input: Size_area, Dist_TX_RX, Num_D2D, Num_Ch
-> Number of channel is same with the number of CUE
'''
def loc_init(Size_area, Dist_TX_RX, Num_D2D, Num_Ch):
tx_loc = Size_area * (np.random.rand(Num_D2D, 2) - 0.5)
rx_loc = np.zeros((Num_D2D + 1, 2))
for i in range(Num_D2D):
temp_chan = Feasible_Loc_Init(tx_loc[i, :], Size_area, Dist_TX_RX)
rx_loc[i, :] = temp_chan
tx_loc_CUE = Size_area * (np.random.rand(Num_Ch, 2) - 0.5)
return rx_loc, tx_loc, tx_loc_CUE
'''
Update location of users.
Update is conducted by considering the Delta_mov which is the amount of distance that users moves
Input: Size_area, Dist_TX_RX, Num_D2D, Num_Ch
-> Number of channel is same with the number of CUE
'''
def loc_update(Size_area, Dist_TX_RX, rx_loc, tx_loc, tx_loc_CUE, Delta_mov):
tx_loc_update = tx_loc
rx_loc_update = rx_loc
tx_loc_CUE_update = tx_loc_CUE
Num_D2D = np.shape(tx_loc)[0]
Num_CH = np.shape(tx_loc_CUE)[0]
## Determine the location of D2D users
for i in range(Num_D2D):
## Use 2*size_area to deactivate the second condition
tx_loc_update[i, :] = Feasible_Loc_Update(tx_loc[i, :], tx_loc_update[i, :], Size_area, 2 * Size_area,
Delta_mov)
rx_loc_update[i, :] = Feasible_Loc_Update(rx_loc[i, :], tx_loc_update[i, :], Size_area, Dist_TX_RX, Delta_mov)
## Determine the location of CUE Users
for i in range(Num_CH):
tx_loc_CUE_update[i, :] = Feasible_Loc_Update(tx_loc_CUE[i, :], tx_loc_CUE[i, :], Size_area, 2 * Size_area,
Delta_mov)
return rx_loc_update, tx_loc_update, tx_loc_CUE_update
'''
Determine the channel gain (UPLINK)
--------------------------------------------
** The basic setting
: Pathloss exponent -> 3.8
: Pathloss constant -> 34.5
: Moving speed -> 3km/h
: Power update -> 100ms
: Reset location -> every 100 samples
--------------------------------------------
Each channel is time varying according to speed
Location of all users are initialized every 1000 samples.
The location of CUE for each band is different
The output looks like as follows:
output[Sample][Channel][Users][Users]
Example for 2 D2D and 1 CUE channel :
[
h_{D2D_1 -> D2D_1} h_{D2D_1->D2D_2} h_{D2D_1->BS}
h_{D2D_2 -> D2D_1} h_{D2D_2->D2D_2} h_{D2D_2->CUE}
h_{CUE -> D2D_1} h_{CUE->D2D_2} h_{CUE->BS}
]
Accordingly, output[0] returns the first sample
output[0][0] return the (N+1) X (N+1) channel gain for first band
'''
def ch_gen(Size_area, D2D_dist, Num_D2D, Num_Ch, Num_samples, Delta_mov=0.0833, PL_alpha=38., PL_const=34.5):
ch_w_fading = []
## Perform initialization just once and the rest channel is generated by moving users
rx_loc, tx_loc, tx_loc_CUE = loc_init(Size_area, D2D_dist, Num_D2D, Num_Ch)
## Calculate the
for i in range(Num_samples):
rx_loc, tx_loc, tx_loc_CUE = loc_init(Size_area, D2D_dist, Num_D2D, Num_Ch)
ch_w_temp_band = []
for j in range(Num_Ch):
tx_loc_with_CUE = np.vstack((tx_loc, tx_loc_CUE[j]))
## generate distance_vector
dist_vec = rx_loc.reshape(Num_D2D + 1, 1, 2) - tx_loc_with_CUE
dist_vec = np.linalg.norm(dist_vec, axis=2)
dist_vec = np.maximum(dist_vec, 5)
# find path loss // shadowing is not considered
pu_ch_gain_db = - PL_const - PL_alpha * np.log10(dist_vec)
pu_ch_gain = 10 ** (pu_ch_gain_db / 10)
multi_fading = 0.5 * np.random.randn(Num_D2D + 1, Num_D2D + 1) ** 2 + 0.5 * np.random.randn(Num_D2D + 1,
Num_D2D + 1) ** 2
final_ch = np.maximum(pu_ch_gain * multi_fading, np.exp(-30))
ch_w_temp_band.append(np.transpose(final_ch))
ch_w_fading.append(ch_w_temp_band)
return ch_w_fading
'''
Find the optimal value for one sample with one channel
The shape of input channel is as follows:
-> channel[users][users]
Note that i
Example for 2 D2D and 1 CUE channel :
[
h_{D2D_1 -> D2D_1} h_{D2D_1->D2D_2} h_{D2D_1->BS}
h_{D2D_2 -> D2D_1} h_{D2D_2->D2D_2} h_{D2D_2->BS}
h_{CUE -> D2D_1} h_{CUE->D2D_2} h_{CUE->BS}
]
'''
def cal_SINR_one_sample_one_channel(channel, tx_power, noise):
## Note that we transpose the channel to
diag_ch = np.diag(channel)
inter_ch = channel-np.diag(diag_ch)
tot_ch = np.multiply(channel, np.expand_dims(tx_power, -1))
int_ch = np.multiply(inter_ch, np.expand_dims(tx_power, -1))
sig_ch = np.sum(tot_ch-int_ch, axis=1)
int_ch = np.sum(int_ch, axis=1)
SINR_val = np.divide(sig_ch, int_ch+noise)
cap_val = np.log(1.0+SINR_val)
return cap_val
def cal_CUE_INTER_one_sample_one_channel(channel, tx_power):
## Note that we transpose the channel to
diag_ch = np.diag(channel)
inter_ch = channel-np.diag(diag_ch)
int_ch = np.multiply(inter_ch, np.expand_dims(tx_power, -1))
int_ch = np.sum(int_ch, axis=1)
return int_ch
'''
Find all possible combination
How to use it?
>> optimal_power(channel, granuity of transmit power, noise, CUE_thr)
CUE_thr: Determine the minimum data rate required for CUE
DUE_thr: Determine the minimum data rate required for D2D user
Returned value is the sum of all D2D users
'''
'''
Find all possible combination
How to use it?
>> optimal_power(channel, granuity of transmit power, noise, CUE_thr)
CUE_thr: Determine the minimum data rate required for CUE
DUE_thr: Determine the minimum data rate required for D2D user
Returned value is the sum of all D2D users
'''
def cal_rate_NP(channel, tx_power_in, tx_max, noise, DUE_thr, I_thr, P_c):
num_sample = channel.shape[0]
num_channel = channel.shape[1]
## Note that the num_user i
num_D2D_user = channel.shape[2] - 1
tot_SE = 0
tot_EE = 0
tot_PW = 0
tot_cap_CUE_vio = 0
tot_cap_DUE_vio = 0
tot_cap_CUE_vio_num = 0.001
tot_cap_DUE_vio_num = 0.001
DUE_violation = 0
CUE_violation = 0
# tot_success_prob counts the number of successful samples
tot_success_prob = 0
tx_power = np.hstack((tx_power_in, tx_max * np.ones((tx_power_in.shape[0], 1, num_channel))))
for i in range(num_sample):
cur_cap = 0
DUE_mask = 1
CUE_mask = 1
for j in range(num_channel):
cur_ch = channel[i][j]
cur_power = tx_power[i, :, j]
cur_power = np.array([cur_power])
cur_ch_cap = cal_SINR_one_sample_one_channel(cur_ch, cur_power, noise)
inter = cal_CUE_INTER_one_sample_one_channel(cur_ch, cur_power)
cur_cap = cur_cap + cur_ch_cap[0]
CUE_mask = CUE_mask * (inter[0, num_D2D_user] > I_thr)
if inter[0, num_D2D_user] > I_thr:
tot_cap_CUE_vio = tot_cap_CUE_vio + cur_ch_cap[0, num_D2D_user]
tot_cap_CUE_vio_num = tot_cap_CUE_vio_num + 1
for j in range(num_D2D_user):
DUE_mask = DUE_mask * (cur_cap[j] > DUE_thr)
if cur_cap[j] < DUE_thr:
tot_cap_DUE_vio = tot_cap_DUE_vio + cur_cap[j]
tot_cap_DUE_vio_num = tot_cap_DUE_vio_num + 1
summed_tx_power = np.sum(tx_power[i, :, :], axis=1) + P_c
D2D_SE_sum = np.sum(cur_cap[:-1])
D2D_EE_sum = np.sum(cur_cap[:-1]/summed_tx_power[:-1])
D2D_PW_sum = np.sum(np.sum(tx_power[i, :, :], axis=1)[:-1])
if np.sum(CUE_mask) == 0:
CUE_violation = CUE_violation + 1
if np.sum(DUE_mask) == 0:
DUE_violation = DUE_violation + 1
tot_SE = tot_SE + D2D_SE_sum
tot_EE = tot_EE + D2D_EE_sum
tot_PW = tot_PW + D2D_PW_sum
tot_SE = tot_SE / num_D2D_user / num_sample
tot_EE = tot_EE / num_D2D_user / num_sample
tot_PW = tot_PW / num_D2D_user / num_sample
PRO_DUE_vio = tot_cap_DUE_vio_num / (num_sample * num_D2D_user)
PRO_CUE_vio = tot_cap_CUE_vio_num / (num_sample * num_channel)
return tot_SE, tot_EE, tot_PW, PRO_CUE_vio, PRO_DUE_vio
'''
Find all possible combination
Use product to find combination
'''
def all_possible_tx_power(num_channel, num_user, granuty):
items = [np.arange(granuty)] * num_user * num_channel
temp_power = list(itertools.product(*items))
temp_power = np.reshape(temp_power, (-1, num_user, num_channel))
power_check = np.sum(temp_power, axis=2)
flag = (power_check / (granuty - 1) <= 1).astype(int)
flag = (np.sum(flag, axis=1) / num_user == 1).astype(int)
flag = np.reshape(flag, (-1, 1))
temp_power_1 = np.reshape(temp_power, (-1, num_user * num_channel))
temp_power = temp_power_1 * flag
power = np.reshape(temp_power, (-1, num_user, num_channel)) / (granuty - 1)
power_mat = []
for i in range(power.shape[0]):
sum_val = np.sum(power[i])
if sum_val != 0:
power_mat.append(power[i])
return np.array(power_mat)
def WMMSE_sum_rate(p_int, H, Pmax, int_cell):
K = np.size(p_int)
vnew = 0
b = np.sqrt(p_int)
f = np.zeros(K)
w = np.zeros(K)
for i in range(K):
f[i] = H[i, i] * b[i] / ( np.matmul(np.square(H[i, :]), np.square(b)) + int_cell[i])
w[i] = 1 / (1 - f[i] * b[i] * H[i, i])
vnew = vnew + np.log(w[i])
VV = np.zeros(100)
for iter in range(100):
vold = vnew
for i in range(K):
btmp = w[i] * f[i] * H[i, i] / sum(w * np.square(f) * np.square(H[:, i]))
b[i] = np.minimum(btmp, np.sqrt(Pmax)) + np.maximum(btmp, 0) - btmp
vnew = 0
for i in range(K):
f[i] = H[i, i] * b[i] / ( np.matmul(np.square(H[i, :]), np.square(b) ) + int_cell[i])
w[i] = 1 / (1 - f[i] * b[i] * H[i, i] + 1e-12)
vnew = vnew + np.log(w[i])
VV[iter] = vnew
if vnew - vold <= 1e-3:
break
p_opt = np.square(b)
return p_opt
'''
Find all possible combination
How to use it?
>> optimal_power(channel, granuity of transmit power, noise, CUE_thr)
CUE_thr: Determine the minimum data rate required for CUE
DUE_thr: Determine the minimum data rate required for D2D user
Returned value is the sum of all D2D users AND POWER val
'''
def optimal_power(channel, tx_max, granuty, noise, DUE_thr, I_thr, P_c, tx_power_set):
num_channel = channel.shape[1]
## Note that the num_user i
num_D2D_user = channel.shape[2] - 1
num_samples = channel.shape[0]
num_success = 0
tot_SE = 0
tot_EE = 0
tot_PW = 0
power_mat_SE = []
power_mat_EE = []
power_mat_PW = []
feasible_channel_mat = []
infeasible_channel_mat = []
# tot_success_prob counts the number of successful samples
tot_success_prob = 0
### IMPORTANT: 2019-02-17
### Granuity should be change to include minus 1
tx_power = tx_power_set
tx_power = tx_max * np.hstack((tx_power, np.ones((tx_power.shape[0], 1, num_channel))))
for i in range(num_samples):
cur_cap = 0
DUE_mask = 1
CUE_mask = 1
if i%10000 == 0:
print(i)
for j in range(num_channel):
cur_ch = channel[i][j]
cur_ch_cap = cal_SINR_one_sample_one_channel(cur_ch, tx_power[:, :, j], noise)
inter = cal_CUE_INTER_one_sample_one_channel(cur_ch, tx_power[:, :, j])
cur_cap = cur_cap + cur_ch_cap
CUE_mask = CUE_mask * (inter[:, num_D2D_user] < I_thr)
for j in range(num_D2D_user):
DUE_mask = DUE_mask * (cur_cap[:, j] > DUE_thr)
summed_tx_power = np.sum(tx_power, axis=2) + P_c
SE_sum = np.sum(cur_cap[:, :-1], axis=1)
EE_sum = np.sum(cur_cap[:, :-1] / summed_tx_power[:, :-1], axis=1)
PW_sum = np.sum(np.sum(tx_power, axis=2)[:, :-1], axis=1)
SE_sum = SE_sum * CUE_mask * DUE_mask
EE_sum = EE_sum * CUE_mask * DUE_mask
PW_sum = PW_sum * (1/(CUE_mask+1e-10)) * (1/(DUE_mask+1e-10))
D2D_SE_arg = np.argmax(SE_sum)
D2D_EE_arg = np.argmax(EE_sum)
D2D_PW_arg = np.argmin(PW_sum[1:])
max_SE = np.max(SE_sum)
max_EE = np.max(EE_sum)
min_PW = np.min(PW_sum[1:])
if (max_SE) > 0 & (max_EE >0) & (min_PW > 0):
feasible_channel_mat.append(channel[i])
power_mat_EE.append(tx_power[D2D_EE_arg][:-1])
power_mat_SE.append(tx_power[D2D_SE_arg][:-1])
power_mat_PW.append(tx_power[D2D_PW_arg][:-1])
num_success = num_success + 1
tot_SE = tot_SE + max_SE
tot_EE = tot_EE + max_EE
tot_PW = tot_PW + min_PW
else:
infeasible_channel_mat.append(channel[i])
tot_SE = tot_SE / num_success / num_D2D_user
tot_EE = tot_EE / num_success / num_D2D_user
tot_PW = tot_PW / num_success / num_D2D_user
return tot_SE, tot_EE, tot_PW, np.array(power_mat_SE), np.array(power_mat_EE), np.array(power_mat_PW), np.array(feasible_channel_mat)
'''
This function is used to divide feasible channels and infeasible channels
'''
def optimal_power_check_valid(channel, tx_max, granuty, noise, DUE_thr, I_thr, tx_power_set):
num_channel = channel.shape[1]
## Note that the num_user i
num_D2D_user = channel.shape[2] - 1
num_samples = channel.shape[0]
## Feasible
feasible_channel_mat = []
infeasible_channel_mat = []
# tot_success_prob counts the number of successful samples
tx_power = tx_power_set
tx_power = tx_max * np.hstack((tx_power, np.ones((tx_power.shape[0], 1, num_channel))))
for i in range(num_samples):
if i%10 == 0:
print(i)
cur_cap = 0
DUE_mask = 1
CUE_mask = 1
for j in range(num_channel):
cur_ch = channel[i][j]
cur_ch_cap = cal_SINR_one_sample_one_channel(cur_ch, tx_power[:, :, j], noise)
inter = cal_CUE_INTER_one_sample_one_channel(cur_ch, tx_power[:, :, j])
cur_cap = cur_cap + cur_ch_cap
CUE_mask = CUE_mask * (inter[:, num_D2D_user] > I_thr)
for j in range(num_D2D_user):
DUE_mask = DUE_mask * (cur_cap[:, j] > DUE_thr)
D2D_sum = np.sum(cur_cap[:, :-1], axis=1)
D2D_sum = D2D_sum * CUE_mask * DUE_mask
max_cap = np.max(D2D_sum)
if max_cap > 0:
feasible_channel_mat.append(channel[i])
else:
infeasible_channel_mat.append(channel[i])
return np.array(feasible_channel_mat), np.array(infeasible_channel_mat)
'''
This function is used to divide feasible channels and infeasible channels
'''
def convert_optimal_power(tx_power_mat, tx_max, Num_power_level, Num_channel):
num_samples = tx_power_mat.shape[0]
num_user = tx_power_mat.shape[1]
resource_alloc = []
for i in range(num_samples):
resource_alloc_inner = []
for j in range(num_user):
channel_select = np.argmax(tx_power_mat[i, j])
power_select = np.round(tx_power_mat[i, j, channel_select] / tx_max * (Num_power_level - 1))
resource_alloc_mat = np.zeros((Num_power_level + Num_channel,))
resource_alloc_mat[int(power_select)] = 1
resource_alloc_mat[int(Num_power_level + channel_select)] = 1
resource_alloc_inner.append(resource_alloc_mat)
resource_alloc.append(np.array(resource_alloc_inner))
return np.array(resource_alloc)
'''
Find random power
'''
def random_power(channel, tx_max, granuty, noise, DUE_thr, I_thr, P_c):
num_sample = channel.shape[0]
num_channel = channel.shape[1]
## Note that the num_user i
num_D2D_user = channel.shape[2] - 1
tot_SE = 0
tot_EE = 0
tot_PW = 0
tot_cap_CUE_vio = 0
tot_cap_DUE_vio = 0
tot_cap_CUE_vio_num = 0.001
tot_cap_DUE_vio_num = 0.001
DUE_violation = 0
CUE_violation = 0
for i in range(num_sample):
cur_cap = 0
DUE_mask = 1
CUE_mask = 1
tx_power = np.random.random((num_D2D_user+1, num_channel))
tx_power = tx_power / np.reshape(np.sum(tx_power, axis=1), (-1, 1))
tx_power = tx_max * tx_power * np.random.random((num_D2D_user+1, 1))
tx_power[num_D2D_user, :] = tx_max
for j in range(num_channel):
cur_ch = channel[i][j]
tx_power_select = np.reshape(tx_power[:, j], (1, -1))
cur_ch_cap = cal_SINR_one_sample_one_channel(cur_ch, tx_power_select, noise)
inter = cal_CUE_INTER_one_sample_one_channel(cur_ch, tx_power_select)
cur_cap = cur_cap + cur_ch_cap[0]
CUE_mask = CUE_mask * (inter[0, num_D2D_user] > I_thr)
if inter[0, num_D2D_user] > I_thr:
tot_cap_CUE_vio = tot_cap_CUE_vio + cur_ch_cap[0, num_D2D_user]
tot_cap_CUE_vio_num = tot_cap_CUE_vio_num + 1
for j in range(num_D2D_user):
DUE_mask = DUE_mask * (cur_cap[j] > DUE_thr)
if cur_cap[j] < DUE_thr:
tot_cap_DUE_vio = tot_cap_DUE_vio + cur_cap[j]
tot_cap_DUE_vio_num = tot_cap_DUE_vio_num + 1
D2D_sum = np.sum(cur_cap[:-1])
CUE_sum = np.sum(cur_cap[-1])
D2D_sum_filter = D2D_sum * CUE_mask * DUE_mask
summed_tx_power = np.sum(tx_power, axis=1) + P_c
SE_sum = np.sum(cur_cap[:-1])
EE_sum = np.sum(cur_cap[:-1]/summed_tx_power[:-1])
PW_sum = np.sum(np.sum(tx_power, axis=1)[:-1])
if np.sum(CUE_mask) == 0:
CUE_violation = CUE_violation + 1
if np.sum(DUE_mask) == 0:
DUE_violation = DUE_violation + 1
tot_SE = tot_SE + SE_sum
tot_EE = tot_EE + EE_sum
tot_PW = tot_PW + PW_sum
tot_SE = tot_SE / num_D2D_user / num_sample
tot_EE = tot_EE / num_D2D_user / num_sample
tot_PW = tot_PW / num_D2D_user / num_sample
PRO_DUE_vio = tot_cap_DUE_vio_num / (num_sample * num_D2D_user)
PRO_CUE_vio = tot_cap_CUE_vio_num / (num_sample * num_channel)
return tot_SE, tot_EE, tot_PW, PRO_CUE_vio, PRO_DUE_vio
'''
LOSS MODELS
'''
'''
This function calculates the capacity of D2D users and CUE user
The return of this function is D2D capacity and CUE capacity
Capacity for each channel is acculmulated.
The shape of return is as follows:
[Num_sample, Num_users]
The output will be capacity of D2D and capacity of CUE
'''
def cal_RATE_tf(channel, tx_power, tx_max, noise, num_samples, log_data_mean, log_data_std):
chan_num = channel.shape[1]
user_num = channel.shape[2]
cap_val = tf.constant(0.0)
CUE_cap = []
channel_rev = tf.exp(channel * log_data_std + log_data_mean)
for i in range(chan_num):
tx_power_w_CUE = tf.concat([tx_power[:, :, i], tx_max * tf.ones((num_samples, 1))], axis=1)
tot_ch = tf.multiply(channel_rev[:, i], tf.expand_dims(tx_power_w_CUE, -1))
sig_ch = tf.linalg.diag_part(tot_ch)
inter_ch = tot_ch - tf.linalg.diag(sig_ch)
inter_ch = tf.reduce_sum(inter_ch, axis=1)
SINR_val = tf.div(sig_ch, inter_ch + noise)
cap_val = cap_val + tf.log(tf.constant(1.0) + SINR_val)
CUE_cap.append(tf.log(tf.constant(1.0) + SINR_val)[:, -1])
cap_val_D2D = cap_val[:, :user_num - 1]
return cap_val_D2D, tf.transpose(tf.convert_to_tensor(CUE_cap))
def cal_inter_tf(channel, tx_power, tx_max, noise, num_samples, log_data_mean, log_data_std):
chan_num = channel.shape[1]
user_num = channel.shape[2]
cap_val = tf.constant(0.0)
Inter_cap = []
channel_rev = tf.exp(channel * log_data_std + log_data_mean)
for i in range(chan_num):
tx_power_w_CUE = tf.concat([tx_power[:, :, i], tx_max * tf.ones((num_samples, 1))], axis=1)
tot_ch = tf.multiply(channel_rev[:, i], tf.expand_dims(tx_power_w_CUE, -1))
sig_ch = tf.linalg.diag_part(tot_ch)
inter_ch = tot_ch - tf.linalg.diag(sig_ch)
inter_ch = tf.reduce_sum(inter_ch, axis=1)
Inter_cap.append(inter_ch[:, -1])
return tf.transpose(tf.convert_to_tensor(Inter_cap))
def cal_EE_tf(cap_val_D2D, tx_power, p_c):
tx_power_d2d = tf.reduce_sum(tx_power, axis=2)
EE = tf.div(cap_val_D2D, p_c+tx_power_d2d)
return EE
def cal_EE_tf_temp(cap_val_D2D, tx_power, p_c):
tx_power_d2d = tf.reduce_sum(tx_power, axis=2)
EE = tf.div(1.0, p_c+tx_power_d2d)
return EE
def cal_LOSS_Total_SE_tf(channel, tf_output, noise, DUE_thr, I_thr, tx_max, num_samples, log_data_mean, log_data_std,
lambda_mat):
## Output of DNN should be divided proerly.
## [power alloc, resource alloc on channel]
chan_num = int(channel.shape[1])
user_num = int(channel.shape[2] - 1)
power_granu = int(Num_power_level)
tx_pow_chan = tf.minimum(tf_output, 1.0)*tx_max
D2D_rate, CUE_rate = cal_RATE_tf(channel, tx_pow_chan, tx_max, noise, num_samples, log_data_mean, log_data_std)
I_val = cal_inter_tf(channel, tx_pow_chan, tx_max, noise, num_samples, log_data_mean, log_data_std)
D2D_vio = tf.nn.tanh(tf.nn.relu(DUE_thr-D2D_rate) / (DUE_thr + 1e-10))
CUE_vio = tf.nn.tanh(tf.nn.relu(I_val - I_thr) / (I_thr + 1e-10))
D2D_vio_sum = tf.reduce_mean(D2D_vio)
CUE_vio_sum = tf.reduce_mean(CUE_vio)
Loss = -lambda_mat[0] * tf.reduce_mean(D2D_rate, axis=1) + lambda_mat[1] * D2D_vio_sum + lambda_mat[2] * CUE_vio_sum
return Loss
def cal_LOSS_Total_EE_tf(channel, tf_output, noise, DUE_thr, I_thr, tx_max, num_samples, log_data_mean, log_data_std,
lambda_mat, P_c):
## Output of DNN should be divided proerly.
## [power alloc, resource alloc on channel]
chan_num = int(channel.shape[1])
user_num = int(channel.shape[2] - 1)
power_granu = int(Num_power_level)
tx_pow_chan = tf.minimum(tf_output, 1.0) * tx_max
D2D_rate, CUE_rate = cal_RATE_tf(channel, tx_pow_chan, tx_max, noise, num_samples, log_data_mean, log_data_std)
EE_rate = cal_EE_tf(D2D_rate, tx_pow_chan, P_c)
I_val = cal_inter_tf(channel, tx_pow_chan, tx_max, noise, num_samples, log_data_mean, log_data_std)
D2D_vio = tf.nn.tanh(tf.nn.relu(DUE_thr-D2D_rate) / (DUE_thr + 1e-10))
CUE_vio = tf.nn.tanh(tf.nn.relu(I_val - I_thr) / (I_thr + 1e-10))
D2D_vio_sum = tf.reduce_mean(D2D_vio)
CUE_vio_sum = tf.reduce_mean(CUE_vio)
Loss = -1000.0*lambda_mat[0] * tf.reduce_mean(EE_rate, axis=1) + lambda_mat[1] * D2D_vio_sum + lambda_mat[2] * CUE_vio_sum
#Loss = -100.0 * lambda_mat[0] * tf.reduce_mean(EE_rate, axis=1)
return Loss
def cal_LOSS_Total_PW_tf(channel, tf_output, noise, DUE_thr, I_thr, tx_max, num_samples, log_data_mean, log_data_std,
lambda_mat):
chan_num = int(channel.shape[1])
user_num = int(channel.shape[2] - 1)
power_granu = int(Num_power_level)
tx_pow_chan = tf.minimum(tf_output, 1.0) * tx_max
D2D_rate, CUE_rate = cal_RATE_tf(channel, tx_pow_chan, tx_max, noise, num_samples, log_data_mean, log_data_std)
I_val = cal_inter_tf(channel, tx_pow_chan, tx_max, noise, num_samples, log_data_mean, log_data_std)
D2D_vio = tf.nn.tanh(tf.nn.relu(DUE_thr-D2D_rate) / (DUE_thr + 1e-10))
CUE_vio = tf.nn.tanh(tf.nn.relu(I_val - I_thr) / (I_thr + 1e-10))
D2D_vio_sum = tf.reduce_mean(D2D_vio)
CUE_vio_sum = tf.reduce_mean(CUE_vio)
Loss = lambda_mat[0] * tf.reduce_mean(tx_pow_chan, axis=1) + lambda_mat[1] * D2D_vio_sum + lambda_mat[2] * CUE_vio_sum
#Loss = -100.0 * lambda_mat[0] * tf.reduce_mean(EE_rate, axis=1)
return Loss
def cal_LOSS_rate_tf(channel, tf_output, noise, tx_max, num_samples, log_data_mean, log_data_std):
tx_pow_chan = tf_output*tx_max
D2D_rate, CUE_rate = cal_RATE_tf(channel, tx_pow_chan, tx_max, noise, num_samples, log_data_mean, log_data_std)
Loss = -tf.reduce_mean(D2D_rate, axis=1)
return Loss
def cal_LOSS_ee_tf(channel, tf_output, noise, tx_max, num_samples, log_data_mean, log_data_std, P_c):
tx_pow_chan = tf_output * tx_max
D2D_rate, CUE_rate = cal_RATE_tf(channel, tx_pow_chan, tx_max, noise, num_samples, log_data_mean, log_data_std)
EE_rate = cal_EE_tf(D2D_rate, tx_pow_chan, P_c)
Loss = -tf.reduce_mean(EE_rate, axis=1)
return Loss
def cal_LOSS_ee_tf_temp(channel, tf_output, noise, tx_max, num_samples, log_data_mean, log_data_std, P_c):
tx_pow_chan = tf_output * tx_max
D2D_rate, CUE_rate = cal_RATE_tf(channel, tx_pow_chan, tx_max, noise, num_samples, log_data_mean, log_data_std)
EE_rate = cal_EE_tf_temp(D2D_rate, tx_pow_chan, P_c)
Loss = -tf.reduce_mean(EE_rate, axis=1)
return Loss
def cal_LOSS_pw_tf_temp(channel, tf_output, noise, tx_max, num_samples, log_data_mean, log_data_std):
tx_pow_chan = tf_output * tx_max
Loss = tf.reduce_mean(tx_pow_chan, axis=1)
return Loss
def cal_LOSS_DUE_CONST_tf(channel, tf_output, noise, DUE_thr, CUE_thr, tx_max, num_samples, log_data_mean, log_data_std,
lambda_mat):
## Output of DNN should be divided proerly.
## [power alloc, resource alloc on channel]
chan_num = int(channel.shape[1])
user_num = int(channel.shape[2] - 1)
tx_pow_chan = tf_output*tx_max
D2D_rate, CUE_rate = cal_RATE_tf(channel, tx_pow_chan, tx_max, noise, num_samples, log_data_mean, log_data_std)
D2D_vio = tf.cast(DUE_thr>D2D_rate, tf.float32)
D2D_vio_sum = tf.reduce_mean(D2D_vio, axis=1)
Loss = D2D_vio_sum
return Loss
def cal_LOSS_CUE_CONST_tf(channel, tf_output, noise, DUE_thr, I_thr, tx_max, num_samples, log_data_mean, log_data_std,
lambda_mat):
## Output of DNN should be divided proerly.
## [power alloc, resource alloc on channel]
chan_num = int(channel.shape[1])
user_num = int(channel.shape[2] - 1)
power_granu = int(Num_power_level)
tx_pow_chan = tf_output*tx_max
I_val = cal_inter_tf(channel, tx_pow_chan, tx_max, noise, num_samples, log_data_mean, log_data_std)
CUE_vio = tf.cast(I_val>I_thr, tf.float32)
CUE_vio_sum = tf.reduce_mean(CUE_vio, axis=1)
Loss = CUE_vio_sum
return Loss
'''
This function calculates the loss for RATE
'''
def cal_LOSS_init_tf(channel, tf_output, y_true):
Loss = tf.reduce_mean(tf.reduce_mean(tf.pow(tf_output-y_true, 2), axis=2))
return Loss
def Total_SE_loss_wrapper(input_tensor, noise, DUE_thr, I_thr, tx_max, num_sample, log_data_mean, log_data_std, lambda_mat):
def TOTAL_SE_loss(y_true, y_pred):
Loss = cal_LOSS_Total_SE_tf(input_tensor, y_pred, noise, DUE_thr, I_thr, tx_max, num_sample, log_data_mean, log_data_std, lambda_mat)
return Loss
return TOTAL_SE_loss
def Total_EE_loss_wrapper(input_tensor, noise, DUE_thr, I_thr, tx_max, num_sample, log_data_mean, log_data_std, lambda_mat, P_c):
def TOTAL_EE_loss(y_true, y_pred):
Loss = cal_LOSS_Total_EE_tf(input_tensor, y_pred, noise, DUE_thr, I_thr, tx_max, num_sample, log_data_mean, log_data_std, lambda_mat, P_c)
return Loss
return TOTAL_EE_loss
def Total_PW_loss_wrapper(input_tensor, noise, DUE_thr, I_thr, tx_max, num_sample, log_data_mean, log_data_std, lambda_mat):
def TOTAL_PW_loss(y_true, y_pred):
Loss = cal_LOSS_Total_PW_tf(input_tensor, y_pred, noise, DUE_thr, I_thr, tx_max, num_sample, log_data_mean, log_data_std, lambda_mat)
return Loss
return TOTAL_PW_loss
def Rate_loss_wrapper(input_tensor, noise, tx_max, num_sample, log_data_mean, log_data_std):
def RATE_loss(y_true, y_pred):
Loss = cal_LOSS_rate_tf(input_tensor, y_pred, noise, tx_max, num_sample, log_data_mean, log_data_std)
return Loss
return RATE_loss
def EE_loss_wrapper(input_tensor, noise, tx_max, num_sample, log_data_mean, log_data_std, P_c):
def EE_loss(y_true, y_pred):
Loss = cal_LOSS_ee_tf(input_tensor, y_pred, noise, tx_max, num_sample, log_data_mean, log_data_std, P_c)
return Loss
return EE_loss
def PW_loss_wrapper(input_tensor, noise, tx_max, num_sample, log_data_mean, log_data_std):
def PW_loss(y_true, y_pred):
Loss = cal_LOSS_pw_tf_temp(input_tensor, y_pred, noise, tx_max, num_sample, log_data_mean, log_data_std)
return Loss
return PW_loss
def INT_CONST_loss_wrapper(input_tensor):
def INT_loss(y_true, y_pred):
chan_num = int(input_tensor.shape[1])
rounded_value = tf.round(y_pred)
Loss = tf.reduce_mean(tf.reduce_mean(tf.square(y_pred-rounded_value), axis=2), axis=1)
return Loss
return INT_loss
def RA_CONST_loss_wrapper(input_tensor):
def RA_loss(y_true, y_pred):
chan_num = int(input_tensor.shape[1])
rounded_value = tf.round(y_pred)
Loss = tf.reduce_mean(tf.reduce_mean(tf.square(y_pred[:,:,-chan_num:]-rounded_value[:,:,-chan_num:]), axis=2), axis=1)
return Loss
return RA_loss
def DUE_CONST_loss_wrapper(input_tensor, noise, DUE_thr, I_thr, tx_max, num_sample, log_data_mean, log_data_std, lambda_mat):
def DUE_loss(y_true, y_pred):
Loss = cal_LOSS_DUE_CONST_tf(input_tensor, y_pred, noise, DUE_thr, I_thr, tx_max, num_sample, log_data_mean,
log_data_std, lambda_mat)
return Loss
return DUE_loss
def CUE_CONST_loss_wrapper(input_tensor, noise, DUE_thr, I_thr, tx_max, num_sample, log_data_mean, log_data_std, lambda_mat):
def CUE_loss(y_true, y_pred):
Loss = cal_LOSS_CUE_CONST_tf(input_tensor, y_pred, noise, DUE_thr, I_thr, tx_max, num_sample, log_data_mean,
log_data_std, lambda_mat)
return Loss
return CUE_loss
def Init_loss_wrapper(input_tensor):
def Init_loss(y_true, y_pred):
Loss = cal_LOSS_init_tf(input_tensor, y_pred, y_true)
return Loss
return Init_loss
'''
DNN MODELS
'''
def DNN_basic_module(Input_layer, Num_weights_inner, Num_outputs, Num_layers=3, activation='relu'):
Inner_layer = layers.Dense(Num_weights_inner)(Input_layer)
Inner_layer_in = layers.Activation('relu')(Inner_layer)
## Number of layers should be at least 2
assert Num_layers > 1