Evolvability and analysis of robot control networks*

Identification of control system classes capable of generating adaptive behaviour over time is a black art. Many practitioners rely on systems that have “always worked in the past”, others may use trial-and-error until success, but carry out no subsequent analysis of why the particular system actually worked. Here we apply the techniques of dynamical systems theory to the analysis of successfully evolved robot control systems, allowing us to identify potentially useful properties of network classes