A Framework for Robotic Perception

Roger A. Browse and Susan J. Lederman

Abstract

The development of intelligent, flexible robots that can adapt to their environment will require sophisticated perceptual systems. There are drawbacks and limitations to the direct application of computational vision systems as the mainstay of robotic perception. A more promising approach is to consider the possibility of multimodal perception involving an active visual component, along with touch sensors using proprioceptive information. There are several benefits of this broader approach to robotic perception because the perceptual information derived from the different knowledge sources is complementary. There are several interesting and addressable questions about the interaction of the various knowledge sources and about the repercussions of active perception on robotic planning systems. In addition, other areas of psychological research can now be viewed as being directly relevant to robotic issues.