A neural network based testbed for modelling sensorimotor integration in robotic applications

Preliminary work toward the integration of sensorimotor interactions using the Neural Simulation Language (NSL) and the Rapid Robotics Application Development (R²AD) environment for computing sensory information and creating motor behavior, respectively, for the purpose of visuomotor coordination in the real world is described. It is noted that, by combining these environments, one gains greater flexibility in designing neural networks to model various perceptions and behaviors in biological systems. This system also gives the computational neurobiologist a new avenue of investigation: the ability to see one´s algorithms, based upon neurophysiological and behavioral data, come to life within an artificial creature. Such a capability can allow a better behavioural comparison between the real system and the artificial algorithm. As an application example, the modeling of the visual sensory modality in predator avoidance behavior found in frogs and toads is considered