Visual learning, adaptive expectations, and behavioral conditioning of the mobile robot MAVIN

Abstract

We have developed a control system for mobile robots which utilizes dynamical neural networks for learning and performance at all stages, from visual object learning and recognition to behavioral conditioning, and have implemented it on a mobile robot, the Mobile Adaptive VIsual Navigator (MAVIN). The system includes networks for early visual perception, eye motion control, pattern learning and object recognition, object associations and delayed expectation learning, emotional states, behavioral actions, and associative switching. The system takes as visual input various patterns of light ( i.e., 3D objects), and learns these objects invariant to location, size, orientation, angle of gaze (foreshortening effects), and aspect on the visual field. The system associates reflex motor behaviors with certain learned visual objects. A robot eye motion system that can detect and adaptively track moving objects is also incorporated in MAVIN. These sensory-motor reflex associations and learned expectations are finally used to demonstrate various classical conditioning paradigms on MAVIN, in which a new visual stimulus is associated with the behavior triggered by either the on-set (conditioned excitor) or off-set (conditioned inhibitor) of some other stimulus. Extinction of a conditioned excitor, and the nonextinction of a conditioned inhibitor are also demonstrated. Behavioral conditioning phenomena are emergent consequences of the overall system neurodynamics.