Artificial neural network for mobile robot topological localization
Abstract
This paper presents a neural network based approach to a mobile robot localization in front of a certain local object. The robot is equipped with ultrasonic range sensors mounted around the platform. We employ the Fuzzy-ARTMAP network for supervised learning of associations between vectors of sensor readouts and the robot's pose coordinates. In this approach, a world model in the form of a map, as well as its updating routine, become superflous for the considered problem solution. The system, trained on real world data of a door neighborhood region reveals satisfactory performance, sufficient for door-passing task purposes. The proposed method of a mobile robot positioning may be efficiently applied in environments containing natural, geometrical beacons.
A preliminary version of this work has been previously published at the International Workshop on Intelligent Robotic Systems (IRS’94), Grenoble, France, July 1994.
BibTeX
@article{Racz-1995-121667,author = {Janusz Racz and Artur Dubrawski},
title = {Artificial neural network for mobile robot topological localization},
journal = {Robotics and Autonomous Systems},
year = {1995},
month = {November},
volume = {16},
number = {1},
pages = {73 - 80},
}