Stochastic filtering on mobile devices in complex dynamic environments
Loading...
Date
2014-04-02
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Gathering information, especially about the immediately surrounding world,
is a central aspect of any smart device, whether it is a robot, a partially autonomous
vehicle, or a mobile handheld device. The consequential use of
electrical sensors always implies the need to filter the imperfect sensor data
output in order to gain reliable information. While the challenge of perception
and cognition in machines is not a new one, new technology constantly
opens up new possibilities and challenges. This is stressed further by the
advent of cheap sensor technology and the possibility to use a multitude
of small sensors, with the simultaneous constraint of limited resources on
mobile, battery-powered computing devices.
In this work, stochastic methods are used to filter sensor data, which
is gathered by mobile devices, to model the devices' location and eventually
also relevant parts of their dynamic environment. This is done with a
focus on online algorithms and computation on these mobile devices themselves,
which implies limited available processing power and the necessity for
computational efficiency. This dissertation's purpose is to impart a better
understanding about the conception and design of stochastic filtering solutions,
to propose localization algorithms beyond the current state of the art,
and to show the use of simultaneous localization and mapping algorithms
in the context of cooperatively estimating the surrounding world of a team
of robots in a fast changing, dynamic environment. To achieve these goals,
the concepts are depicted in multiple application scenarios, design choices
and their implications systematically cover all aspects of sensing and estimation,
and the proposed systems are evaluated in real-world experiments on humanoid robots and other mobile devices.
Description
Table of contents
Keywords
Localization, Tracking, Stochastic modeling, Robotics, Kalman, Particle filter