This session is aimed at those research groups that develop and use
novel artificial neural network architectures specially adapted to the
necessities of real life interaction with dynamic environments. This includes
neural networks for the implementation of sensing tasks, those dedicated
to sensor fusion and higher level processes and finally those dedicated
to the control and actuation part of a robot. An interesting subarea of
this session would be the use of neural networks for behavior controllers
and how to obtain and/or train networks taking into account the credit
apportitioning problem and the fact that robots operate in very noisy environments
that are very hard to simulate. Finally, a subject to be considered is
the problem of combining networks corresponding to different behaviors
into global controllers that provide advantages over monolithic solutions.