Evolving analytic hierarchy processes for artificial life-based games

Analytic hierarchy process (AHP) is a multiple criteria decision making approach which is widely used in different engineering domains. The hierarchical structure of AHP makes it a very suitable approach for behavior design in artificial life based games. But the details of an AHP are always designed by a series of interviews with experts of the domain, and this is not possible for problems in which an expert does not exist such as computer simulations. In such cases, the behavioral engineering of agents is a very complicated task because the designer may not be able to see all different aspects of agents' life. To overcome this problem, this paper has used an artificial life algorithm to design and optimize the action selection system for a set of animal-like agents for Giti artificial life game, based on evolving AHPs. The approach is compared with evolving case based reasoning and direct human design and it is shown that it has notable advantages versus both other approaches.