On the design of genetic algorithms for geographical applications

In many geographical optimization problems, the linkage (which determines the structure of building blocks) is determined by the spatial relationships between the components of a solution. Therefore the linkage can be identified easily, unlike in most other problems. Based ox this observation, we develop a hybrid GA that uses a geometrically local optimiser-one that computes good solutions to subproblems that are local in the geometric sense. One of the main advantages of our method is that it leads to GA's that are easily adapted to slight changes in the problem definition, without the need to tune many parameters in the fitness function. We apply our method to the map labeling problem, (placing as many names as possible on a map, without overlap), where it leads to good results.