CAREER: Neurally Inspired Active Vision: Theory, Models, and Applications in Mobile Robotics

University Of Washington, Seattle WA

Investigators

Abstract

The goal of the project is to bridge the gap between human and machine vision by designing active vision systems modeled after neurobiology. Such systems can benefit from the ability to make eye, head, and body movements for active sensing, and can learn robust models of the visual world directly from interaction with the environment. Computational models of active vision will be formulated within the context of neurobiological data, addressing the problems of object identification, detection, attention, tracking, modeling self-motion, and oculomotor learning. An anthropomorphic real-time binocular active vision system will be designed based on commercially available camera platforms and real-time image processing hardware. The performance of the approach will be evaluated on both wheeled and legged robots in tasks involving visual navigation and multi-robot collaboration. The education part of the project includes development of an interdisciplinary curriculum in computational neuroscience.