Tile-world — A case study of Genetic Network Programming with automatic program generation

Abstract

Genetic Network Programming (GNP) is a novel evolutionary algorithm. It has graph-based structures which is extended from Genetic Algorithm (GA) and Genetic Programming (GP). Up to now, GNP has been applied to many research fields such as data mining and elevator control systems. On the other hand, automatic program generation is a way to obtain a program without explicitly programming it, and Genetic Programming is the traditional paradigm in this field. Drawn from the inspiration of GP, GNP for Automatic Program Generation (GNP-APG) has been proposed. In this paper, GNP-APG is applied to the Tile-world, which is a famous test bed with dynamic and uncertain characteristics. GNP-APG uses a kind of genotype-phenotype mapping process to create program. The procedure of the program generation based on evolution is demonstrated in this paper. In simulations, different tile-worlds between the training phase and the testing phase are used for performance evaluations and the results shows that GNP-APG could have better performances than the conventional GNP methods.