Evolutionary perspective for Goal-Babbling Exploration

Goal-Babbling Exploration consists in the generic task of sensory-motor coordination learning through goal space exploration. A typical drawback in Goal-Babbling Exploration algorithms remains the human involvement during the adjustment phase. In order to tackle this issue, we propose in this article to study an evolutionary perspective. It takes the form of applying an optimization approach to enhance the learning process. Our goal is to disengage the modeler without efficiency loss. We use a simple chromosome describing every parameters on which the model is built. The experimental results obtained demonstrate its feasibility: a self-adjustment of these parameters bringing up good enough developmental trajectories.