A Generic Stochastic Method to Arrange Virtual Creatures in a 3D-Scenario

We propose a generic stochastic method to handle the space planning problem [Eastman73] inherent in declarative modeling [Donikian93] of virtual scenes. Genetic algorithms [Holland75] is an adaptive and heuristic search algorithm based on the evolutionary ideas of natural selection; it is capable to find good solutions within a reasonable time. In this work, solution representation and fitness calculation are chief issues. The proposed method has been implemented as part of a framework useful to design virtual scenes [Piza04J. The aim of this work is to let users arrange virtual creatures in a 3D-world by using geometric constraints [Leroux00] (behind, between, close to, ...) instead of geometric data. Although the platform offers a fixed set of constraints currently, it provides means to let users add more, by sub-classing class constraint and overriding its method sigma().