Z = mtimes( x, y, 'ldivide' )

I have only recently come into contact with cvx and am eager to use it to complete the trajectory optimization of aircraft. Therefore, my convex optimization theory is relatively weak and it is difficult to find problems in the program. I have the following error in use:

Error using * (line 306)
Disciplined convex programming error:
This expression produced one or more invalid numeric values (NaNs).

Error in \ (line 14)
z = mtimes( x, y, ‘ldivide’ );

Error in slack_inturn (line 133)
X(:,i) = Ak-(Bk*(t0-tf)U1(:,1) + (Tk-BkU1k(:,i))U2 + Wk + 2DX(:,i));

My code is here.

cvx_begin
cvx_solver mosek
    variable U1(3,n);        %u1 u2 u3
    variable U2(1,1);        %tf
    expression X(6,n+1);
    expression AX(6,1);
    expression DX(n,6);
    obj = U2;
    minimize(obj);
    
    subject to
    X(:,1)== X0;

X = X';

for num_x = 1:NX
DX(:,num_x) = D*X(:,num_x);
end
DX = DX’;
X = X’;

 for i=1:n
     
   Ak = [ - (R0*S*V(i)*cdjian*rou0*exp(-R0*beta*y(i)))/(2*m) - (R0*S*V(i)*cdjian*rou0*u3(i)*exp(-R0*beta*y(i)))/(2*m),                                                                 -cos(theta(i))/r(i)^2,                                0, 0, (R0^2*S*V(i)^2*beta*cdjian*rou0*exp(-R0*beta*y(i)))/(4*m) + (R0^2*S*V(i)^2*beta*cdjian*rou0*u3(i)*exp(-R0*beta*y(i)))/(4*m), 0;
                                    cos(theta(i))/(V(i)^2*r(i)^2) + (R0*S*cljian*rou0*u1(i)*exp(-R0*beta*y(i)))/(2*m),                                                           sin(theta(i))/(V(i)*r(i)^2),                                0, 0,                                                              -(R0^2*S*V(i)*beta*cljian*rou0*u1(i)*exp(-R0*beta*y(i)))/(2*m), 0;
                                                     -(R0*S*cljian*rou0*u2(i)*exp(-R0*beta*y(i)))/(2*m*cos(theta(i))), -(R0*S*V(i)*cljian*rou0*u2(i)*exp(-R0*beta*y(i))*sin(theta(i)))/(2*m*cos(theta(i))^2),                                0, 0,                                                 (R0^2*S*V(i)*beta*cljian*rou0*u2(i)*exp(-R0*beta*y(i)))/(2*m*cos(theta(i))), 0;
                                                                                            cos(psi(i))*cos(theta(i)),                                                       -V(i)*cos(psi(i))*sin(theta(i)),  -V(i)*cos(theta(i))*sin(psi(i)), 0,                                                                                                                           0, 0;
                                                                                                        sin(theta(i)),                                                                    V(i)*cos(theta(i)),                                0, 0,                                                                                                                           0, 0;
                                                                                           -cos(theta(i))*sin(psi(i)),                                                        V(i)*sin(psi(i))*sin(theta(i)),  -V(i)*cos(psi(i))*cos(theta(i)), 0,                                                                                                                           0, 0];
   Bk = [                                                0,                                                               0, -(R0*S*V(i)^2*cdjian*rou0*exp(-R0*beta*y(i)))/(4*m);
          (R0*S*V(i)*cljian*rou0*exp(-R0*beta*y(i)))/(2*m),                                                               0,                                                   0;
                                                         0, -(R0*S*V(i)*cljian*rou0*exp(-R0*beta*y(i)))/(2*m*cos(theta(i))),                                                   0;
                                                         0,                                                               0,                                                   0;
                                                         0,                                                               0,                                                   0;
                                                         0,                                                               0,                                                   0];
                                                                           
   Tk = -1*[-0.5*0.5*R0*rou0*exp(-beta*R0*y(i))*V(i)*V(i)*S*cdjian/m-sin(theta(i))/(r(i)*r(i))-0.5*0.5*R0*rou0*exp(-beta*R0*y(i))*V(i)*V(i)*S*cdjian*u3(i)/m;
           -cos(theta(i))/(V(i)*r(i)*r(i))+u1(i)*0.5*R0*rou0*exp(-beta*R0*y(i))*V(i)*V(i)*S*cljian/(m*V(i));
           -u2(i)*0.5*R0*rou0*exp(-beta*R0*y(i))*V(i)*V(i)*S*cljian/(m*V(i)*cos(theta(i)));
           V(i)*cos(theta(i))*cos(psi(i));
           V(i)*sin(theta(i));
           -V(i)*cos(theta(i))*sin(psi(i))];
     
   fk = [-0.5*0.5*R0*rou0*exp(-beta*R0*y(i))*V(i)*V(i)*S*cdjian/m-sin(theta(i))/(r(i)*r(i))-0.5*0.5*R0*rou0*exp(-beta*R0*y(i))*V(i)*V(i)*S*cdjian*u3(i)/m;
           -cos(theta(i))/(V(i)*r(i)*r(i))+u1(i)*0.5*R0*rou0*exp(-beta*R0*y(i))*V(i)*V(i)*S*cljian/(m*V(i));
           -u2(i)*0.5*R0*rou0*exp(-beta*R0*y(i))*V(i)*V(i)*S*cljian/(m*V(i)*cos(theta(i)));
           V(i)*cos(theta(i))*cos(psi(i));
           V(i)*sin(theta(i));
           -V(i)*cos(theta(i))*sin(psi(i))];
 
Xk = [V;theta;psi;x;y;z];
U1k = [u1(1:n);u2(1:n);u3(1:n)];

   Wk = (t0-tf)*fk + (t0-tf)*Bk*U1k(:,i) - Ak*Xk(:,i) - Tk*tf - (t0-tf)*Bk*U1k(:,i) + Bk*U1k(:,i)*tf;
   X(:,i) = Ak\-(Bk*(t0-tf)*U1(:,1) + (Tk-Bk*U1k(:,i))*U2 + Wk + 2*DX(:,i));
 end
cvx_end

Please tell me where the error is and how I should modify the program. I will be very grateful for your guidance.

Look at (apply isnan to) all the input data on the RHS of X(:,i) = Ak-(Bk*(t0-tf)U1(:,1) + (Tk-BkU1k(:,i))U2 + Wk + 2DX(:,i)); The error is probably due to at least one element of input data being `NaN``.

If your CVX program is occurring within a Successive Convex Approximation (SCA) loop, one way NaN in input data can occur is if the argmin from a previous SCA iteration is used as input to the current CVX problem. If the previous iteration did not solve to optimality (infeasible, unbounded, or numerical difficulty), the argmin will have value NaN, which then becomes NaN input data on the next iteration.