Evolving non-trivial behavior on autonomous robots:
    Adaptation is more powerful than decomposition and integration

                            Stefano Nolfi
         Institute of Psychology, National Research Council
              15, Viale Marx - 00187 - Rome - Italy
            voice: ++39-6-86090231   fax:++39-6-824737
              e-mail: stefano@kant.irmkant.rm.cnr.it
             http://kant.irmkant.rm.cnr.it/nolfi.html


A new way of building control systems, known as behavior based robotics, has 
recently been proposed  to  overcome  the difficulties of the traditional AI 
approach to robotics. This new approach  is based upon the idea of providing 
the  robot  with  a  range  of  simple behaviors and letting the environment 
determine  which  behavior  should  have control  at any given time. 
In this talk we claim that  the decomposition and integration process should 
be  the  result  of  an  adaptation  process  and  not of the decision of an 
experimenter. To support this  hypothesis  we  show  how  in  the  case of a 
simple task in which a real autonomous robot is supposed to classify objects 
of  different shapes,  by  letting  the  entire  behavior  emerge through an 
evolutionary technique, a  more simple and  robust  solution can be obtained 
than by trying to design a set of modules and to integrate them.
Moreover, we will present a  second  experiment  in  which  neural  networks 
with  different  architectures  have  been trained to control a mobile robot 
designed to keep an  arena  clear  by picking-up trash objects and releasing 
them outside the  arena. We  will compare,  in  simulation  and  on  a  real  
robot, five different network architectures  and  will  show  that a network 
which allows  for  fine-grained  modularity  achieves  significantly  better  
performance.  By comparing the functionality  of each network module and its 
interaction with  a description of  the  simple behavior components, we will 
show that it is not possible to find  simple  correlations;  rather,  module  
switching  and  interaction  is  correlated   with  low-level  sensory-motor  
mappings.