A study of GasNet spatial embedding in a delayed-response task

Abstract

GasNet artificial neural networks can be used as complex neurocontrollers involving virtual chemical neuromodulation as well as synaptic interaction. The aim of this paper is to fur-ther explore the role of space in GasNet models on a delayed-response robot task. Comparative results demonstrate that the use of spatial constraints is not a prerequisite for a good per-formance of the original model in terms of speed of evolution.