Complex robot training tasks through bootstrapping system identification

Authors Organisations
  • Otar Akanyeti(Author)
    University of Essex
  • U. Nehmzow(Author)
    Ulster University
  • S. A. Billings(Author)
    The University of Sheffield
Type Conference Proceeding (Non-Journal item)
Original languageEnglish
Title of host publication2008 IEEE International Conference on Robotics and Biomimetics, ROBIO 2008
PublisherIEEE Press
Pages2168-2173
Number of pages6
ISBN (Print)9781424426799
DOI
Publication statusPublished - 08 May 2009
Externally publishedYes
Event2008 IEEE International Conference on Robotics and Biomimetics, ROBIO 2008 - Bangkok, Thailand
Duration: 21 Feb 200926 Feb 2009

Publication series

Name2008 IEEE International Conference on Robotics and Biomimetics, ROBIO 2008

Conference

Conference2008 IEEE International Conference on Robotics and Biomimetics, ROBIO 2008
Country/TerritoryThailand
CityBangkok
Period21 Feb 200926 Feb 2009
Links
Permanent link
View graph of relations
Citation formats

Abstract

Many sensor-motor competences in mobile robotics applications exhibit complex, non-linear characteristics. Previous research has shown that polynomial NARMAX models can learn such complex tasks. However as the complexity of the task under investigation increases, representing the whole relationship in one single model using only raw sensory inputs would lead to large models. Training such models is extremely difficult, and, furthermore, obtained models often exhibit poor performances. This paper presents a bootsrapping method of generating complex robot training tasks using simple NARMAX models. We model the desired task by combining predefined low level sensor motor controllers. The viability of the proposed method is demonstrated by teaching a Scitos G5 autonomous robot to achieve complex route learning tasks in the real world robotics experiments.