How to add constraints to simulannealbnd minimization

How to add constraint to minimization problem?

Simply x(3) >10*x(4)

ConstraintFunction = @simple_constraint;  %constraint here does not affect minimization
hybridopts = optimoptions('fmincon','OptimalityTolerance',1e-14);
options = optimoptions(@simulannealbnd,'InitialTemperature',[1e4 1e4 1e4 1e4     ],...
    'PlotFcn',{@saplotbestf,@saplottemperature,@saplotf,@saplotstopping},'HybridFcn',{'fmincon',hybridopts},...
    'AnnealingFcn','annealingboltz','TemperatureFcn','temperatureboltz');
options.ReannealInterval = 400;
[x,fval] = simulannealbnd(ObjectiveFunction,X0,LB,UB,options);  

function [c, ceq] = simple_constraint(x)    
c = [-x(3) + 10*x(4)];
% c = [];    
   ceq = [];  

Answer 1

[Walter Roberson]

You cannot add constraints directly. However you can set AnnealingFcn to the handle of a function that only generates points that meet the constraints.

Alternately you could permit those points to be generated, but set the AcceptanceFcn to reject them.

[Ole]

Thank you. How is this implemented ?

AnnealingFcn  = @simple_constraint
function [c, ceq] = simple_constraint(x)    
c = [-x(3) + 10*x(4)];
% c = [];    
   ceq = [];  

[Walter Roberson]

No, nothing like that. AnnealFcn needs to return data points; it is not designed as a constraint function.

Give the command

type annealingboltz

to see the code for the default function.

You would need to add a loop that generated new y and newx and keep going generating new ones if the constraint was not met.

[Ole]

thank you