In this paper, we propose an experimental validation of a new control strategy based on the use of intuitive control and neural networks CMAC in order to control the under-actuated robot RABBIT. The training of the neural networks is carried out during the simulation. When the training is finished, these neural networks are used to control the real robot. We also present an evaluation of the robustness of this control strategy.

This paper presents the experimental implementation and validation of a framework for the systematic design, analysis, and performance enhancement of controllers that induce stable walking in N-link underactuated planar biped robots. Controllers designed via this framework act by enforcing virtual constraints - holonomic constraints imposed via feedback - on the robot's configuration. The stability properties of the resulting walking motions may be easily analyzed in terms of a two-dimensional sub-dynamic of the full walking model. The experimental validation is performed on RABBIT, a 5-link prototype constructed by the French project Commande de Robots a Pattes of the CNRS-GdR Automatique.