-
Notifications
You must be signed in to change notification settings - Fork 16
/
Copy pathDC_RIS.py
194 lines (185 loc) · 6.63 KB
/
DC_RIS.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
# -*- coding: utf-8 -*-
#import argparse
#from scipy.optimize import minimize
import copy
import numpy as np
np.set_printoptions(precision=6,threshold=1e3)
import warnings
import cvxpy as cp
import sys
def DC_F(N,K,L,h_d,G,maxiter,theta_DC_RIS,verbose,epislon):
rho=5
h=np.zeros([N,K],dtype=complex)
H=np.zeros([N,N,K],dtype=complex)
# scale=10000
for i in range(K):
h[:,i]=h_d[:,i]+G[:,:,i]@theta_DC_RIS
# h[:,i]=h[:,i]*scale
H[:,:,i]=np.outer(h[:,i],h[:,i].conj())
# print(H)
M = np.random.randn(N,1)+1j*np.random.randn(N,1);
M= copy.deepcopy(np.outer(M,M.conj()))
_,V=np.linalg.eigh(M)
u=V[:,N-1]
# for i in range(K):
# print(np.trace(M@H[:,:,i]))
# define the optimization problem
M_var=cp.Variable((N,N), complex =True)
M_partial=cp.Parameter((N,N), hermitian =True)
M_partial.value = copy.deepcopy(np.outer(u,u.conj()))
# print(M_partial.value)
constraints = [M_var >> 0]
constraints += [cp.real(M_var@H[:,:,k])>=1 for k in range(K)]
cost=cp.real(cp.trace(M_var))+rho*cp.real(cp.trace((np.eye(N)-M_partial)@M_var))
prob = cp.Problem(cp.Minimize(cost),constraints)
obj0=0
#iteritively solve:
for iter in range(maxiter):
if verbose>1:
print('Solving f, Inner iter={}'.format(iter))
with warnings.catch_warnings():
warnings.simplefilter("ignore")
with open('out.log','w+') as f:
sys.stdout.flush()
stream=sys.stdout
sys.stdout=f
prob.solve(solver=cp.SCS,verbose=False,scale=1e-10,max_iters=1000)
sys.stdout.flush()
sys.stdout=stream
if prob.status=='infeasible' or prob.value is None:
break
# print(prob.value)
err=np.abs(prob.value-obj0)
M=copy.deepcopy(M_var.value)
_,V=np.linalg.eigh(M)
u=V[:,N-1]
M_partial.value = copy.deepcopy(np.outer(u,u.conj()))
obj0 = prob.value
if err<epislon:
break
u,_,_=np.linalg.svd(M,compute_uv=True,hermitian=True)
m=u[:,0]
return m/np.linalg.norm(m)
def DC_theta(N,K,L,h_d,G,maxiter,f,verbose,epsilon):
#Compute R,c:
A=np.zeros([L,K],dtype=complex)
c=np.zeros([K,],dtype=complex)
R=np.zeros([L+1,L+1,K],dtype=complex)
for k in range(K):
c[k]=f.conj()@h_d[:,k]
A[:,k]=(f.conj()@G[:,:,k])
R[0:L,0:L,k]=np.outer(A[:,k],A[:,k].conj())
R[0:L,L,k]=A[:,k]*c[k]
R[L,0:L,k]=R[0:L,L,k].conj()
# R[:,:,k]=copy.deepcopy((R[:,:,k].conj().T+R[:,:,k])/2)
#initial V:
V = np.random.randn(L+1,1)+1j*np.random.randn(L+1,1);
V=V/np.abs(V)
V= copy.deepcopy(np.outer(V,V.conj()))
# for k in range(K):
# print(np.trace(R[:,:,k]@V))
# print(np.abs(c[k])**2)
_,v=np.linalg.eigh(V)
# print(v.shape)
u=v[:,L]
u= np.random.randn(L+1,1)+1j*np.random.randn(L+1,1);
#initial other parameters:
infeasible_check=False
#initial the optimization problem:
V_var=cp.Variable((L+1,L+1), hermitian =True)
V_var.value=V
V_partial=cp.Parameter((L+1,L+1), hermitian =True)
V_partial.value = copy.deepcopy(np.outer(u,u.conj()))
# print(M_partial.value)
constraints = [V_var >> 0]
constraints += [V_var[n,n]==1 for n in range(L)]
constraints += [cp.real(V_var@R[:,:,k])+np.abs(c[k])**2>=1 for k in range(K)]
cost=cp.real(cp.trace((np.eye(L+1)-V_partial)@V_var))
prob = cp.Problem(cp.Minimize(cost),constraints)
obj0=0
for iter in range(maxiter):
if verbose>1:
print('Solving theta, Inner iter={}'.format(iter))
with warnings.catch_warnings():
warnings.simplefilter("ignore")
with open('out.log','w+') as f:
sys.stdout.flush()
stream=sys.stdout
sys.stdout=f
prob.solve(solver=cp.SCS,verbose=False,scale=1e-10,max_iters=200,warm_start=True)
sys.stdout.flush()
sys.stdout=stream
print(prob.value)
if verbose:
print('Status={}, Value={}'.format(prob.status,prob.value))
if prob.status=='infeasible' or prob.value is None:
infeasible_check=True
break
err=np.abs(prob.value-obj0)
V=copy.deepcopy(V_var.value)
_,v=np.linalg.eigh(V)
u=v[:,L]
V_partial.value = copy.deepcopy(np.outer(u,u.conj()))
obj0 = prob.value
if err<epsilon:
break
u,_,_=np.linalg.svd(V,compute_uv=True,hermitian=True)
v_tilde=u[:,0]
vv=v_tilde[0:L]/v_tilde[L]
vv=copy.deepcopy(vv/np.abs(vv))
return vv,infeasible_check
def DC_main(N,K,L,h_d,G,maxiter,iter_num,epsilon,verbose,epsilon2):
F_DC_RIS=np.zeros([N,],dtype='complex')
# theta_DC_RIS=np.zeros([L,],dtype='complex')
theta_DC_RIS=np.ones([L],dtype=complex)
h=np.zeros([N,K],dtype=complex)
for i in range(K):
h[:,i]=h_d[:,i]+G[:,:,i]@theta_DC_RIS
obj_pre=min(np.abs(np.conjugate(F_DC_RIS)@h)**2)
infeasible=False
stop=False
for iter in range(maxiter):
if verbose:
print('iter={}'.format(iter))
#Given theta, update F
F_DC_RIS=DC_F(N,K,L,h_d,G,iter_num,theta_DC_RIS,verbose,epsilon2)
# print(F_DC_RIS.shape)
#Given F, update theta
theta_DC_RIS,infeasible=DC_theta(N,K,L,h_d,G,iter_num,F_DC_RIS,verbose,epsilon2)
h=np.zeros([N,K],dtype=complex)
for i in range(K):
h[:,i]=h_d[:,i]+G[:,:,i]@theta_DC_RIS
obj=min(np.abs(np.conjugate(F_DC_RIS)@h)**2)
if verbose:
print('Gain value={}'.format(obj))
if abs(obj-obj_pre)<epsilon or infeasible==True:
stop=True
obj_pre=obj
if stop:
break
return F_DC_RIS,theta_DC_RIS
def DC_RIS(libopt,h_d,G,verbose):
N=libopt.N
M=libopt.M
L=libopt.L
K=libopt.K
K2=K**2
# Ksum2=sum(K)**2
maxiter=50
iter_num=50
# maxiter=1
epsilon=1e-3
epsilon2=1e-8
# obj_DC_RIS =0
F_DC_RIS, theta_DC_RIS = DC_main(N,M,L,h_d,G,maxiter,iter_num,epsilon,verbose,epsilon2)
h=np.zeros([N,M],dtype=complex)
for i in range(M):
h[:,i]=h_d[:,i]+G[:,:,i]@theta_DC_RIS
gain=K2/(np.abs(np.conjugate(F_DC_RIS)@h)**2)*libopt.sigma
obj=np.max(gain)/(sum(K))**2
obj_DC_RIS=copy.deepcopy(obj)
if verbose:
print('obj={:.6f}\n'.format(obj))
return obj_DC_RIS,F_DC_RIS,theta_DC_RIS
if __name__ == '__main__':
pass