Optimizing "nested" variables with a genetic algorithm

Hi there,

I'm trying to optimize the camera positions for a 3D reconstruction. There are 2 cameras available which can be placed anywhere around the object. The camera position is determined by its 3 spherical coordinates.

I am currently running a ga() optimization where the objective function has an input of a 2x3 matrix, determining camera positions. The output is a scalar characterizing the exactness of the 3D reconstruction.

objfun = @(xarray)reconstruction_fun(xarray);

The problem with this approach is that all entries of the matrix are scored as individuals. I would like the algorithm to take 2 vectors as an entry, each describing one camera position. And then optimize the 2 camera positions. The objective function would have to take 2 inputs as variables.

cam1=(radius1, phi1, theta1)
cam2= (..)
objfun = @(cam1, cam2)reconstruction_fun(cam1, cam2);

Is there a better way to achieve the optimization than with all variables thrown into one array?

Thanks, Simon

Answer 1

[Matt J]

The problem with this approach is that all entries of the matrix are scored as individuals.

No, that shouldn't be. As long as you've specified nvars=6, ga() should know to treat a 1x6 input vector as single population member.

[Simon Dosch]

Thanks for the reply. I got it :)