gerald.foliot@univ-lyon2.fr
L.E.A.C.M. -- Universite Lumiere Lyon 2
5, avenue Pierre Mendes France, 69500 Bron, France
Olivier.Michel@di.epfl.ch
LAMI-EPFL, Swiss Federal Institute of Technology
CH-1015 Lausanne, Switzerland
This paper presents an emotion based process generating cognition through learning. Two models are described which investigate the possibilities of designing cogniti ve systems based on emotion. In order to test these models through experiments, we used a mobile robot simulator, Webots, modelling the real Khepera robot. The first proposed model is the sim pliest since it uses only positive and negative reinforcement signals as well as proximity sensors. The robot learns to avoid obstacles based on reflex reaction and on reinforcement signals. The second model relying on appraisal theory is more complex. It involves the use of reinforcement signals, proximity sensors and a simulated color camera to learn, in particular, to associate visual stimulus with the appropriate behavior. Interesting simulation results demonstrate the significance of this promising approach.