Intergrative Vision and Navigation on Photorealistic Simulated Environments

Abstract

ABSTRACT Robot visual and navigation learning requires extensive hands?on training in physically prepared environments. The presence of highly photorealistic environments in videogames provides the opportunity for a robot’s software to be separately pre?trained in a simulated environment, with the possibility that learned visual primitives will exhibit transfer to real environments with corresponding savings in required training time for the physical robot. We propose to use a set of well studied brain?derived algorithms to demonstrate semi?automated learning in photorealistic videogames, constructing rich representations including structural and temporal relations; the resulting learned representations will be tested for their ability to accelerate robot learning in a real environment, both in a lab?based robot and in robots at the Naval Research Laboratories.

Document Details

Document Type
DoD Grant Award
Publication Date
Aug 12, 2016
Source ID
N000141512132

Entities

People

  • Richard Granger

Organizations

  • Board of Trustees of Dartmouth College
  • Office of Naval Research
  • United States Navy

Tags

Fields of Study

  • Computer science

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Computer Vision.
  • STEM Education

Technology Areas

  • AI & ML
  • AI & ML - Autonomous Systems
  • AI & ML - Neural Networks
  • Autonomy