TARTARUS AND FRACTAL GENE REGULATORY NETWORKS WITH INPUTS

Abstract

Tartarus is a benchmark problem used to evaluate artificial intelligence techniques for solving problems in the field of non-Markovian agent motion planning. In this paper a fractal gene regulatory network with inputs is evolved to act as a virtual robot controller in the Tartarus environment. The proposed technique is compared and contrasted with other previously reported techniques and it is shown that the gene regulatory network that includes input information provides an excellent performance without using any explicit memory or environmental modeling. Detailed experimental studies are presented to illustrate the effectiveness and superiority of the proposed approach.