CNN-based control of a robot inspired to snakeboard locomotion

This paper wants to emphasize the role of analog neural processing structures to realize artificial locomotion in mechatronic devices. The approach presented starts by considering locomotion as a complex spatio-temporal phenomena, modelled referring to particular types of reaction-diffusion nonlinear partial differential equations implemented on a Reaction-Diffusion Cellular Neural Network architecture (RD-CNN). Several examples in literature show the usefulness of this methodology applied to generate and control the locomotion in real-time in a number of different robotic structures as multi-legged or worm-like robots. In this paper we apply this technique, using wave-like solutions, obtained by a ring of RD-CNN cells, to generate and control locomotion in a mechanic wheeled structure that exploits the inertia of two masses.