Hi,
I'm having a problem using Fmincon and I cannot find a solution online. I have to minimize a function ('ikSolver' in the code below) whose input is a [10x1] vector of doubles:
[lb,ub] = getLegJointLimits('A')
rng('shuffle');
qlegs = rand(10,1);
qlegs = lb + (ub-lb).*qlegs;
options = optimset('Display','iter','MaxFunEvals',50000,...
'TolFun',1e-06,...
'MaxIter',50000,'LargeScale','on');
[fqs,Fval,EXITFLAG] = fmincon('ikSolver',qlegs,[],[],[],[],lb,ub,'constraint',options);
The values of the boundary vectors are:
lb =
-0.6685
-0.9745
0.1193
-1.5464
-0.6229
-0.2624
-1.5480
0.1282
-0.9782
-0.2811
ub =
0.2712
0.7202
1.8977
0.2598
0.3003
0.6735
0.2582
1.8921
0.7114
0.6523
and the functions appearing (simplified for simplicity):
function F = ikSolver(theta)
global Rmatr
global gt gteq
global H
Qplus = Rmatr*theta;
F = sum(Qplus)^2;
gt = [ ];
gteq = [ ];
end
with:
Rmatr =
0 0 0 0 0 0 0 0 0 1
0 1 1 1 0 0 -1 -1 0 0
0 0 0 0 0 0 0 1 0 0
0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 1 0 0 0 0
0 0 0 0 1 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0 0 0
And 'constraint':
function [gtr,gteqr] = constraint(Jsolcons)
global gt gteq
gtr = gt;
gteqr = gteq;
return
The warning is the following:
Warning: To use the default trust-region-reflective algorithm you must supply the gradient in the objective function and set the GradObj option to 'on'. FMINCON will use the active-set algorithm instead. For information on applicable algorithms, see Choosing the Algorithm in the documentation. > In fmincon at 492
The same warning (and consequent wrong result) appears if I change the arguments of the function fmincon as for example:
[fqs,Fval,EXITFLAG] = fmincon('ikSolver',qlegs,[],[],[],[],lb,ub);
[fqs,Fval,EXITFLAG] = fmincon('ikSolver',qlegs,eye(10),ub);
options = optimset('Display','off','MaxFunEvals',50000,...
'TolFun',1e-06,...
'MaxIter',50000);
I don't understand where am I asking for the trust-region-reflective algorithm which requires the gradient...
Thank you very much!!
