Automatic road network generation with L-systems and genetic algorithms

Abstract

This thesis is a research in computer science and artificial intelligence, more precisely a research in biologically inspired methods. Procedural techniques, inspired by biological developmental models, are known to create complex structures from compact descriptions. This motivates exploring procedural techniques on the subject of automatically generating 3D road network models. In this thesis are L-systems applied to procedurally develop vast road networks, and genetic algorithms (GAs) are applied to tune the procedure to target outcomes. The final solution also incorporates a maximum flow algorithm combined with a genetic algorithm to tune intersections in the road network.