FMINCON requires the following inputs to be of data type double

I have the following code and I can't understand how to fix the error.

Answer 1

[Torsten]

Although I don't understand your problem setup because the variables in the objective function are not influenced by your constraint, here is a running code:

%Parameters of the system%
Fk = 10^10;
Bk = 20*10^6;
p = 0.1; %Watts
g0 = 10^(-4);
w0 = -100;
v = 0.2*10^(-11); %J/cycle
C = 8*500; %cycles/bit
fe = 100*10^9; %Hz
fc = 500*10^9; %Hz
s_enc = 240*10^6; %cycles, random number
s_dec = 240*10^6; %cycles, random number
w = 0.5; %decision weight

%Parameters of the system that change%
fl = (0.4:0.1:1)*10^9; %Hz
D = linspace(0,10,100)*10^6; %MB
theta = randi([0,1]);

r = randi([1 length(fl)],1,1); %generate random nuber
fl_temp = fl(r);

q = randi([1 length(D)],1,1); %generate random nuber
D_temp = D(r);

%variables of the system%
w0_watt = 10^(w0/10)/1000;
R = Bk*log2(1+(p*g0^2)/(w0_watt*Bk));
% E_enc = v*s_enc*D_temp*a3;
% t_end_dec = (v*s_enc*D_temp/fc)*a3 + (v*s_dec*D_temp/fe)*a3;
% E_l = v*C*D_temp*a1;
% T_l = (C*D_temp/fl_temp)*a1;
% T_s = (C*D_temp/fe)*a2;
% T_c = (C*D_temp/fc)*a3;
% T_trans = D_temp/R*(a2 + a3);
% E_r = p*T_trans;
% E_sec = theta*(E_enc + E_r) + (1-theta)*E_r;
% T_sec = theta*(t_end_dec + T_trans) + (1-theta)*T_trans;
%
% E = E_l + E_sec;
% T = max([T_l, T_sec + T_s, T_sec + T_c]);

%Soving the problem%
%set initial guess values for the problem variables%
a1 = 1/3;
a2 = 1/3;
a3 = 1/3;
Emax = 1.0;
Tmax = 1.0;

%Load guess values into an array%
x0 = [a1 a2 a3 Emax Tmax];

E_enc = v*s_enc*D_temp*a3;
t_end_dec = (v*s_enc*D_temp/fc)*a3 + (v*s_dec*D_temp/fe)*a3;
E_l = v*C*D_temp*a1;
T_l = (C*D_temp/fl_temp)*a1;
T_s = (C*D_temp/fe)*a2;
T_c = (C*D_temp/fc)*a3;
T_trans = D_temp/R*(a2 + a3);
E_r = p*T_trans;
E_sec = theta*(E_enc + E_r) + (1-theta)*E_r;
T_sec = theta*(t_end_dec + T_trans) + (1-theta)*T_trans;

E = E_l + E_sec;
T = max([T_l, T_sec + T_s, T_sec + T_c]);

%set lower and upper bounds%
lb = [0 0 0 E T];
ub = [1.0 1.0 1.0 1000.0 1000.0];

%call solver to minimize the objective function given the constraints%
% fmincon(fun,x0,A,b,Aeq,beq,lb,ub,nonlcon)%
A = [1 1 1 0 0];
b = 1.0;
Aeq = [];
beq = [];
[x,fval] = fmincon(@(x)objective(x,w),x0,A,b,Aeq,beq,lb,ub)
end
%define objective function for optimization%
function obj = objective(x,w)
Emax = x(4);
Tmax = x(5);
obj = w*Emax + (1-w)*Tmax;
end

[Konstantina Symeonidou]

This is what I am trying to solve

[Torsten]

Your problem statement can't be correct since T_max and E_max are given - thus the solution to minimize the objective is obvious without calculation.

Maybe you mean

min: w*E + (1-w)*T

instead of

min: w*E_max + (1-w)*T_max

?

[Konstantina Symeonidou]

I am gonna discuss it with my supervisor. I see your point! Thank you very much for your time! I appreciate it!!

[Torsten]

@Konstantina Symeonidou comment moved here:

if I want to implement this method for 2 users, how can i do it? do I use a for loop?

I mean that I have e.g. a1 = [a1(1 user) a1(2 user)] etc, so every variable is an

array

[Torsten]

I don't understand.

Did you clarify the question about the objective ?

[Konstantina Symeonidou]

yes, he told me to keep it as it was

[Torsten]

Then the solution (at least following your setup in the code) is

x(4) = E, x(5) = T, x(1) = x(2) = x(3) = 1/3

[Konstantina Symeonidou]

and what happens if R takes multiple values and I want to implement fmincon for all of them. Do I apply a for loop?

[Torsten]

a1, a2 and a3 are used to calculate quantities before you call fmincon.

This is incorrect. All these calculations have to be done within the objective function.

So you will first have to change this basic problem before looping or further changes to your code.