TABLE OF CONTENTS In this Section it is shown how connections among CNN neurons lead to complex systems able to constitute the CPG for artificial locomotion control Firstly we discuss the approach based on RD-CNN that is the starting point for the more general approach based on the chemical synapse introduced above, then we show some guidelines for the CNN-based CPG design.
Reaction-Diffusion (RD) phenomena occur in many systems (biological, chemical, physical systems) [Murray (1993)], but they obey the same laws that can be represented under the following general RD equations:
| (2.15) |  |
where u is a vector of at least two elements. For instance u can represent the dynamics of an activator/inhibitor system diffusing in a nonlinear medium and showing pattern formation. Another example is the propagation of autowaves in a nonlinear medium. This term was first coined by R.V.Khorhlov to indicate autonomous waves [Krinsky (1984)]. They represent a particular class of nonlinear waves which propagate without forcing in active nonlinear mediums. Autowaves have some typical characteristics, basically different from those of classical waves in conservative systems: their shape remains constant during propagation and reflection and interference do not take place, while diffraction is a common property between classical and autowaves. These autowaves are the basis of the RD-CNN approach.
CNNs are able to show complex phenomena arising in space-distributed fields such as Reaction-Diffusion phenomena in 2D arrays [Arena et. al. (1997)]. Examples are autowaves travelling in nonlinear media or...
