Why Status: Unbounded Optimal value (cvx_optval): +Inf

Hello, everyone. When I use cvx, I encounter some problems, as follows

Calling Mosek 9.1.9: 6250 variables, 884 equality constraints
   For improved efficiency, Mosek is solving the dual problem.
------------------------------------------------------------

MOSEK Version 9.1.9 (Build date: 2019-11-21 11:34:40)
Copyright (c) MOSEK ApS, Denmark. WWW: mosek.com
Platform: Windows/64-X86

MOSEK warning 710: #12 (nearly) zero elements are specified in sparse col '' (32) of matrix 'A'.
MOSEK warning 710: #3 (nearly) zero elements are specified in sparse col '' (46) of matrix 'A'.
MOSEK warning 710: #3 (nearly) zero elements are specified in sparse col '' (100) of matrix 'A'.
MOSEK warning 710: #4 (nearly) zero elements are specified in sparse col '' (133) of matrix 'A'.
MOSEK warning 710: #1 (nearly) zero elements are specified in sparse col '' (147) of matrix 'A'.
MOSEK warning 710: #1 (nearly) zero elements are specified in sparse col '' (201) of matrix 'A'.
MOSEK warning 710: #8 (nearly) zero elements are specified in sparse col '' (430) of matrix 'A'.
MOSEK warning 710: #2 (nearly) zero elements are specified in sparse col '' (444) of matrix 'A'.
MOSEK warning 710: #2 (nearly) zero elements are specified in sparse col '' (498) of matrix 'A'.
MOSEK warning 710: #4 (nearly) zero elements are specified in sparse col '' (531) of matrix 'A'.
Warning number 710 is disabled.
Problem
  Name                   :                 
  Objective sense        : min             
  Type                   : CONIC (conic optimization problem)
  Constraints            : 884             
  Cones                  : 16              
  Scalar variables       : 1482            
  Matrix variables       : 20              
  Integer variables      : 0               

Optimizer started.
Presolve started.
Linear dependency checker started.
Linear dependency checker terminated.
Eliminator started.
Freed constraints in eliminator : 24
Eliminator terminated.
Eliminator started.
Freed constraints in eliminator : 0
Eliminator terminated.
Eliminator - tries                  : 2                 time                   : 0.00            
Lin. dep.  - tries                  : 1                 time                   : 0.00            
Lin. dep.  - number                 : 0               
Presolve terminated. Time: 0.02    
Problem
  Name                   :                 
  Objective sense        : min             
  Type                   : CONIC (conic optimization problem)
  Constraints            : 884             
  Cones                  : 16              
  Scalar variables       : 1482            
  Matrix variables       : 20              
  Integer variables      : 0               

Optimizer  - threads                : 16              
Optimizer  - solved problem         : the primal      
Optimizer  - Constraints            : 855
Optimizer  - Cones                  : 17
Optimizer  - Scalar variables       : 743               conic                  : 735             
Optimizer  - Semi-definite variables: 20                scalarized             : 9832            
Factor     - setup time             : 0.00              dense det. time        : 0.00            
Factor     - ML order time          : 0.00              GP order time          : 0.00            
Factor     - nonzeros before factor : 6.38e+04          after factor           : 6.46e+04        
Factor     - dense dim.             : 2                 flops                  : 6.95e+07        
ITE PFEAS    DFEAS    GFEAS    PRSTATUS   POBJ              DOBJ              MU       TIME  
0   1.3e+02  1.3e+00  1.3e+01  0.00e+00   1.531135369e+01   3.509105336e+00   1.0e+00  0.05  
1   3.5e+01  3.5e-01  5.6e+00  -6.88e-01  4.678028163e+01   3.862443819e+01   2.7e-01  0.13  
2   5.1e+00  5.1e-02  2.1e+00  -8.83e-01  3.321001950e+02   3.385156753e+02   4.0e-02  0.14  
3   6.6e-01  6.6e-03  7.1e-01  -9.45e-01  2.187548819e+03   2.293596718e+03   5.1e-03  0.16  
4   1.1e-01  1.1e-03  2.8e-01  -8.77e-01  1.198185847e+04   1.261379843e+04   8.6e-04  0.19  
5   2.3e-02  2.3e-04  1.2e-01  -9.35e-01  5.195964428e+04   5.470518257e+04   1.8e-04  0.20  
6   2.3e-03  2.3e-05  3.8e-02  -9.87e-01  5.213111896e+05   5.486210202e+05   1.8e-05  0.23  
7   2.8e-04  2.8e-06  1.3e-02  -9.98e-01  4.418116376e+06   4.645422545e+06   2.2e-06  0.25  
8   9.6e-05  9.7e-07  7.7e-03  -9.93e-01  1.444187640e+07   1.509070424e+07   7.5e-07  0.27  
9   2.3e-05  2.3e-07  3.8e-03  -9.91e-01  6.798052006e+07   7.070005298e+07   1.8e-07  0.30  
Optimizer terminated. Time: 0.31    


Interior-point solution summary
  Problem status  : PRIMAL_INFEASIBLE
  Solution status : PRIMAL_INFEASIBLE_CER
  Dual.    obj: 1.8023539098e+01    nrm: 4e+04    Viol.  con: 0e+00    var: 4e+04    barvar: 2e-07    cones: 0e+00  
Optimizer summary
  Optimizer                 -                        time: 0.31    
    Interior-point          - iterations : 9         time: 0.30    
      Basis identification  -                        time: 0.00    
        Primal              - iterations : 0         time: 0.00    
        Dual                - iterations : 0         time: 0.00    
        Clean primal        - iterations : 0         time: 0.00    
        Clean dual          - iterations : 0         time: 0.00    
    Simplex                 -                        time: 0.00    
      Primal simplex        - iterations : 0         time: 0.00    
      Dual simplex          - iterations : 0         time: 0.00    
    Mixed integer           - relaxations: 0         time: 0.00    

------------------------------------------------------------
Status: Unbounded
Optimal value (cvx_optval): +Inf

My code
cvx_clear
cvx_begin
% cvx_solver mosek
cvx_save_prefs
variable W(M,M,K) hermitian semidefinite
variables relax_S1_lamda(L) relax_S2_lamda(L) tau_max(K) gammaa(K,L) R(K,K)
expressions LMI_S1(M+N_all+1,M+N_all+1,K) LMI_S2(M+N_all+1,M+N_all+1,K)
P=0;

    S=0;
    for k=1:K
        for l=1:L
            E(:,l)=[h_rE(:,l)',h_BE(:,l)'];
           A(:,:,k)=[Theta*G*sum(W(:,:,k+1:K),3)*G'*Theta',Theta*G*sum(W(:,:,k+1:K),3);...
                sum(W(:,:,k+1:K),3)*G'*Theta',sum(W(:,:,k+1:K),3)];
            a(:,:,k,l)=A(:,:,k)+relax_S1_lamda(l)*eye(M+N_all);
            b(:,k,l)=E(:,l)'*A(:,:,k);
            c(k,l)=E(:,l)'*A(:,:,k)*E(:,l)+noise-gammaa(k,l)-relax_S1_lamda(l)*epsilong(l);
            LMI_S1(:,:,k,l)=[a(:,:,k,l)   b(:,k,l);...
                b(:,k,l)'    c(k,l)];
            right_up(k,l)=(exp(tau_max_t(1,k))-1)*gammaa(k,l)+gamma_t(k,l)*exp(tau_max_t(1,k))*...
                (tau_max(k)-real(tau_max_t(1,k)));
            B(:,:,k)=[Theta*G*W(:,:,k)*G'*Theta',Theta*G*W(:,:,k);...
                W(:,:,k)*G'*Theta',W(:,:,k)];
            a0(:,:,k,l)=relax_S2_lamda(l)*eye(M+N_all)-B(:,:,k);
            b0(:,k,l)=-E(:,l)'*B(:,:,k);
            c0(k,l)=-E(:,l)'*B(:,:,k)*E(:,l)+right_up(k,l)-relax_S2_lamda(l)*epsilong(l);
            LMI_S2(:,:,k,l)=[a0(:,:,k,l)   b0(:,k,l);...
                b0(:,k,l)'    c0(k,l)];
        end
          S=S-rel_entr(1,R(k,k))/log(2)-tau_max(k)*log2(exp(1));
end
    maximize real(trace(R))
    subject to
          for k=1:K
              for l=1:L
              LMI_S1(:,:,k,l) ==hermitian_semidefinite(M+N_all+1);
              LMI_S2(:,:,k,l) ==hermitian_semidefinite(M+N_all+1);
              relax_S1_lamda(l)>0; 
              relax_S2_lamda(l)>0;
              end

              P=P+trace(W(:,:,k)); 
             tau_max(k)*log2(exp(1))<=M_E;
          end

        for j=2:K
             for k=1:j-1
                  W_temp3=0;
               for i=k+1:K
                    W_temp3=W_temp3+h_k(:,j)'*W(:,:,i)*h_k(:,j);
               end
                pow_pos((R(j,k)-1)*real(A_t(j,k)),2)+pow_pos((real(W_temp3)+noise)/real(A_t(j,k)),2)<=...
                    2*real(h_k(:,j)'*W(:,:,k)*h_k(:,j));%Convert R (j, k) to convex constraint by AMG
               R(j,k)>=R(k,k);
            end
        end

cvx_end

Some initial values



What is the reason for the unbounded?

The infeasibility certificate is of pretty bad quality. I would suggest you first try with the latest Mosek 10.

And perhaps improve the input data scaling to not have near zero elements, which Mosek warned about.