The current default fitness function is based on a bounding box volume. This is not a good option, as the geometry is sized relative to the top-most component in the hierarchy - in any case, volume is not a good fitness function.
An alternative, perhaps based on a similarity or 'tree edit distance' to a particular component hierarchy may be useful here. For example, testing how close the component hierarchy tree is to that of a 787 should hopefully eventually lead to a 787 - a good test to see if the evolutionary algorithm is working. More complex fitness functions may later be added, such as aero-structural and stability analysis.
The current default fitness function is based on a bounding box volume. This is not a good option, as the geometry is sized relative to the top-most component in the hierarchy - in any case, volume is not a good fitness function.
An alternative, perhaps based on a similarity or 'tree edit distance' to a particular component hierarchy may be useful here. For example, testing how close the component hierarchy tree is to that of a 787 should hopefully eventually lead to a 787 - a good test to see if the evolutionary algorithm is working. More complex fitness functions may later be added, such as aero-structural and stability analysis.