Walking robots offer a large spectrum of potential advantages over wheeled structures, due to their superior adaptability capabilities on rough terrain and in environments with obstacles, but their control is a challenge because they have a large number of interconnected degrees of freedom. On the other hand, in nature there exist numberless examples of walking creatures which move on unstructured terrain in an efficient and elegant way. Therefore, many robot designers look to biological inspiration to build efficient walking robots. Two key points emerge from the study of the neural control of locomotion in many living creatures:

• the motor system is hierarchically organized;

• a network of neurons, called CPG, is able to generate the rhythm of locomotion independently of sensory feedback and signals from higher centres.

The problem of locomotion control has already been dealt with using bio-inspired approaches in many works. The approaches and the degrees of inspiration in these works vary. In some cases [Arkin (1998); Brooks (1991)], the biological inspiration is not a key point and the animal is often considered as a term of comparison to define what Artificial Intelligence means and to build software architectures able to show complex (or intelligent) behavior.

A very impressive synergy between biology and engineering is, instead, represented by the work of Quinn (for a review see [Webb and Consi (2001)]), where biological data are used for a very accurate design of a bio-mimetic robot with cockroach kinematics. However, the control of the architecture is performed using a traditional approach,...