The followings are my codes
cvx_begin
variables B_l(N_user,ch) P_l(N_user,ch) B_u(N_user,WAP,ch_u) P_u(N_user,WAP,ch_u);
variable v;
expressions R_l(N_user,ch) R_u(N_user,WAP,ch_u) inter(ch);
for i=1:N_user
for j=1:ch
R_l(i,j)=-rel_entr(B_l(i,j),B_l(i,j)+P_l(i,j)h_l(i,j)/((I_mu(i,j)+N0)10^6));
end
for k=1:WAP
for n=1:ch_u
R_u(i,k,n)=-rel_entr(betaB_u(i,k,n),betaB_u(i,k,n)+P_u(i,k,n)h_u(i,k,n)/(N010^6));
end
end
end
for j=1:ch
for i=1:N_user
inter(j)=inter(j)+h_MS(j)P_l(i,j);
end
end
minimize v;
subject to
for j=1:ch
inter(j)<=v;
sum(B_l(:,j))<=1;
end
for i=1:N_user
sum(R_l(i,:))+sum(R_u(i,:,:))>=R_th;
end
sum(sum(P_l))+sum(sum(sum(P_u)))<=Pt;
sum(sum(sum(P_u)))<=Pt_u;
for k=1:WAP
for n=1:ch_u
sum(B_u(:,k,n))<=1;
end
end
P_l>=0;
P_u>=0;
B_l>=0;
B_u>=0;
R_l>=0;
R_u>=0;
cvx_end
As you can see all the variables are positive, bu the R_L and R_U maybe negative. It is quite strange. since the
R_L=B_Llog(1+(P_Lh_L)/(B_LNo)), R_U=betaB_Ulog(1+(P_Uh_U)/(betaB_U*N0))
N0 is the noise.
anyone can help. Thanks