Galapogos: suboptimal solution & best practice


There is an example of truss optimisation using Galapogos on the karamba3d site. The truss has a uniform load applied to the entire length of the top beam; i.e. this is a uniform and symetrical problem. However the solution generated by Galapogos is not symetric and not very simple. Galapogos is optimising 16 variables so I imagine there is a large fitness landscape and potentially the higher peaks are being missed.

My questions are:

  1. What are good ways to define/ constrain problems before they are fed into Galapogos i.e. ways to reduce the size/ dimensions of the fitness landscape?
  2. Do people have other related thoughts on how best to define Galapogos problems?