Evolutionary and bio-inspired Robotics

Evolutionary Robotics developed by the intersection among Artificial Life and Artificial Intelligence and by approaches denominated Bottom-Up AI (Brooks, 1986), Animat (Wilson, 1991), Behavior-Based AI (Steels, 1990), Animal Robotics (McFarland, 1992). These words identify a vast group of researchers that share the common objective to understand the intelligent behaviour through the construction of artificial systems. They investigate the behaviour of a single individual or groups of individuals, focusing the attention on what makes the behaviour intelligent and adaptive and how this behaviour emerges. The behaviour is defined as a regularity observed in the dynamics of interaction between the characteristics and the processes both of a system of its environment. Evolutionary Robotics is aimed at building systems using artificial models, putting together a series of physical components (for example a physical body of bird, with all of his parts) being satisfied only if this artefact is actually able to fly. An artificial system cannot elude the comparison with a real complex environment and, for this reasons, it is more difficult to build. 
Starting from the experiments of Synthetic Psychology of Braitenberg and from the studies of  Brooks, in the Ph.D. course the main concepts of Robotics (Arkin, 1998), and those of Evolutionary Robotics will be dealt with. The different methodological approaches in the construction of robotic systems present in the current literature will be privileged.   
Bio-inspired Robotics origins from the realization of analogical structures that using the concepts of the dynamical systems realize emergent behaviours, similar to those that are found in different animal species, in the biological world. The paradigm of the CNN (cellular not linear nets, initially defined Cellular Neural Nets) has been studied both on the theoretical and application level. Their structure is constituted by a set of simple non-linear systems, locally connected, denominated “cells”, whose matrixes of connection, denominated Templates are constituted by programmable parameters. The basic architecture foresees that the non linear dynamics of the single variables of state depends only on input and output of the neighbouring cells and corresponds to the whole of the complex dynamics produced by the well-known Chua’s circuit. The analogic calculus of the CNN can be used with success for the real time control of distributed structures.  A machine built with the paradigm of the CNN is a low cost and high analogic performances computer incorporated in a chip. With these systems, the motor behaviours in robotic agents and in artificial models of the biological brain have been simulated. In addition, some attempts to superior cognitive abilities like perception have been made.