fmincon after solve error

I'm having some problem using fmincon. I have three variables:

 var(1) = sym('pilotPower');
 var(2) = sym('dataPower');
 gamma = sym('gamma');

With the vpasolve I solve the equation in function of gamma so fun has two variable var(1) and var(2). Then I take one of the solution and I put it in fmincon after the conversion in a Matlab function (if I don't do that I receive a different error).

These are the constraints:

% Starting evaluation point for fmin con
x0 = [1,249];
% pilotPower + dataPower <= powerBudgetLin
A = [1,1];
b = powerBudgetLin;

And this is the remaining part of the code

eqn = usersNum*alpha^2*var(2)*q+dataSigma2 ==...
receiversNum*alpha^2*var(2)*d/gamma-(usersNum 1)*alpha^2*var(2)*d/(1+gamma);
fun = -vpasolve(eqn,gamma);
myMatlabFunction = matlabFunction(fun(1));
[pilotPowerOpt,fval] = fmincon(myMatlabFunction,x0,A,b);

If I inspect myMatlabFunction I see that it has two variable "pilotPower" and "dataPower"

At the moment I have as a error "Not enough input arguments. Error in optimizationtest (line 106) [pilotPowerOpt,fval] = fmincon(myMatlabFunction,x0,A,b); Caused by: Failure in initial objective function evaluation. FMINCON cannot continue. "

My aim is to maximize the positive result of "fun".

Answer 1

[Torsten]

fun = -solve(eqn,gamma);
[pilotPowerOpt,fval] = fmincon(@(x)myMatlabFunction(x(1),x(2)),x0,A,b);

[Daniele Davoli]

Hi, thank you very much, so, now it works but the results do not respect my constraints -1.77634611474494e+20 6.55702664354688e+19 these are my two results, but their sum should be 250

[Walter Roberson]

Their sum is 250 to within the accuracy of floating point numbers.

>> eps(-1.77634611474494e+20)
ans =
       32768

We will need your full code (all variables initialized) to test further.

[Daniele Davoli]

This is my code, and my request for the fmincon is that pilot power + data power <= 250 with pilot power >= 0 and data power >=0.

%%Parameter
% Power Budget (pilot + data) for each users [mW]
powerBudgetLin = 250;
% Number of Users
usersNum = 3;
% Number of receivers
receiversNum = 10;
% Number of TX pilot symbols
pilotSymbolsNum = usersNum;
% Number of Subcarriers
subcarriersNum = 24;
% Channel distribution power
rho = 1;
% Channel Covariance Matrix
C = rho*eye(receiversNum);
% Channel distrubution mean
muC = 0;
% Noise Variance [mW] value found on Ming thesis
sigma2 = 1.99e-10;
% Noise Variance (pilot)
pilotSigma2 = sigma2;
% Noise Variance (data)
dataSigma2 = sigma2;

%%Simulation variables
x(1) = sym('pilotPower');
x(2) = sym('dataPower');
gamma = sym('gamma');

% Starting evaluation point for fmin con
x0 = [1,249];
% pilotPower + dataPower <= powerBudgetLin
A = [1,1];
b = powerBudgetLin;

%%Tx Power
% Pilot TX Power
pilotPowerSym = x(1)/pilotSymbolsNum;
% Data TX Power
dataPowerSym = x(2)/(subcarriersNum-pilotSymbolsNum);

alpha = sqrt(db2pow(-100));

R = C + pilotSigma2./(alpha^2*x(1))*eye(receiversNum);
r = norm(R);
D = C/R;
d = norm(D);
Q = C-C/R*C;
q = norm(Q);

eqn = usersNum*alpha^2*x(2)*q+dataSigma2 ==...
    receiversNum*alpha^2*x(2)*d/gamma-(usersNum-1)*alpha^2*x(2)*d/(1+gamma);
fun = - solve(eqn,gamma);
myMatlabFunction = matlabFunction(fun(1));
[pilotPowerOpt,fval] = fmincon(@(x)myMatlabFunction(x(1),x(2)),x0,A,b);

I think that I made a mistake because I haven't defined the second and the third inequality. So I tried to put

   lb = [0,0]
   ub = [powerBudgetLin, powerBudgetLin]
   [pilotPowerOpt,fval] = fmincon(@(x)myMatlabFunction(x(1),x(2)),x0,A,b,[],[],lb,ub);

And now seems to work. Thank you!

[Walter Roberson]

So problem solved?

[Daniele Davoli]

I think so, thank you again.