Genetic programming-based geometry reconstruction of conducting cylinders

A general method for optimizing 2D geometries was presented and applied to an inverse scattering problem. This method uses a tree-structured chromosome that represents the geometric combination of several convex polygons. The convex polygons used as inputs are represented by the convex hull of variable-length lists of 2D points. Genetic operators were developed to modify the tree and the point lists. In the two examples given, the method successfully reconstructed the metallic scattering targets resulting in low error in the scattered fields. This method is flexible enough that it could be used in any 2D electromagnetic design problem as presented, and it can also be extended to three dimensions