by Thorsten Belker, Michael Beetz and Armin Cremers
Abstract:
Most state-of-the-art navigation systems for autonomous service robots decompose navigation into global navigation planning and local reactive navigation. While the methods for navigation planning and local navigation themselves are well understood, the plan execution problem, the problem of how to generate and parameterize local navigation tasks from a given navigation plan, is largely unsolved. This article describes how a robot can autonomously learn to execute navigation plans. We formalize the problem as a Markov Decision Process (MDP) and derive a decision theoretic action selection function from it. The action selection function employs models of the robot's navigation actions, which are autonomously acquired from experience using neural network or regression tree learning algorithms. We show, both in simulation and on a RWI B21 mobile robot, that the learned models together with the derived action selection function achieve competent navigation behavior.
Reference:
Thorsten Belker, Michael Beetz and Armin Cremers, "Learning Action Models for the Improved Execution of Navigation Plans", In Robotics and Autonomous Systems, vol. 38, no. 3–4, pp. 137–148, 2002.
Bibtex Entry:
@article{Bel02Lea,
author = {Thorsten Belker and Michael Beetz and Armin Cremers},
title = {Learning Action Models for the Improved Execution of Navigation Plans},
journal = {Robotics and Autonomous Systems},
volume = {38},
number = {3--4},
pages = {137--148},
month = {March},
year = {2002},
bib2html_pubtype = {Journal},
bib2html_rescat = {Robot Learning, Plan-based Robot Control},
bib2html_groups = {IAS},
bib2html_funding = {ignore},
bib2html_keywords = {Learning, Robot, Planning},
abstract = {Most state-of-the-art navigation systems for autonomous service robots decompose navigation into
global navigation planning and local reactive navigation. While the methods for navigation planning
and local navigation themselves are well understood, the plan execution problem, the problem of how
to generate and parameterize local navigation tasks from a given navigation plan, is largely
unsolved.
This article describes how a robot can autonomously learn to execute navigation plans. We formalize
the problem as a Markov Decision Process (MDP) and derive a decision theoretic action selection
function from it. The action selection function employs models of the robot's navigation actions,
which are autonomously acquired from experience using neural network or regression tree learning
algorithms. We show, both in simulation and on a RWI B21 mobile robot, that the learned models
together with the derived action selection function achieve competent navigation behavior.}
}