I-Corps: Accurate GPS-free Navigation and Localization Grant uri icon

abstract

  • The broader impact/commercial potential of this I-Corps project is to develop autonomous navigation technology that will enable systems to robustly operate in uncertain environments without a Global Positioning System (GPS). The project is a result of a confluence of astronomy, aerospace, computational science and artificial intelligence. Commercialization of this technology has the potential to revolutionize space exploration, self-driving cars, Unmanned Aerial Vehicles (UAVs) and other such systems which need accurate position estimation. A key advantage of this project''s technology is enhanced cybersecurity as it does not rely on external signals for navigation. Further, this project will contribute open-source software to the scientific community. It is envisioned that development of a software toolbox that integrates with the popular ROS (Robot Operating System) library will allow researchers to simulate autonomous navigation without GPS.This I-Corps project is a result of research into the problem of Simultaneous Localization and Mapping (SLAM). In SLAM, a robot is not given prior knowledge of its environment, it must use its sensory data and actions to simultaneously build a map of its environment and position itself within its uncertain map. Competing methods in this area exhibit positioning errors which may be unsuitable for long-term navigation. The work developed here shows that by fusing orientation sensing with short-range sensing, the system attain a simplification of the underlying optimization problem. This allows fast and globally optimal solutions. In this approach, a vehicle uses a camera to track celestial bodies in the sky which allows the vehicle to estimate its orientation in space, this information is fused with short-range sensors such as lasers and cameras which track features in vicinity of the vehicle. Using the proposed approach, a system can achieve 100x improvement in position error over existing methods.

date/time interval

  • 2017 - 2018