Online Feature Weighting for Human Discrimination in a Person Following Robot - Foundations on Natural and Artificial Computation - page 221

Online Feature Weighting for Human
Discrimination in a Person Following Robot

V. Alvarez-Santos 1 ,X.M.Pardo 1 ,R.Iglesias 1 , A. Canedo-Rodriguez 1 ,
and C.V. Regueiro 2

1 Department of Electronics and Computer Science
Universidade de Santiago de Compostela, Spain
[e-mail]
2 Department of Electronics and Systems
Universidade da Coruna, Spain

Abstract. A robust and adaptive person-following behaviour is an im-
portant ability that most service robots must have to be able to face chal-
lenging illumination conditions, and crowded spaces of non-structured
environments. In this paper, we propose a system which combines a
laser based tracker with the support of a camera, acting as a discrimina-
tor between the target, and the other people present in the scene which
might cause the laser tracker to fail. The discrimination is done using
a online weighting of the feature space, based on the discriminability of
each feature analysed.

Keywords: feature weighting, person follower, guide robot.

1 Introduction

Robots are expected to become part of our everyday's life, either as assistants,
house appliances, collaborating in the care of elderly people, etc. Nevertheless,
service robots will require sophisticated capabilities in order to be able to work
cooperatively, side-by-side with humans in real world scenarios: one of these abil-
ities is person-following. Service robots must be able to detect, track and follow
people in order to interact with them in close proximity. Furthermore, these ser-
vice robots will also have to be able to interpret human commands and react
to them. This is part of the ongoing research under the project intelligent and
distributed control scenario for the fast and easy deployment of robots in diverse
environments , funded by the Spanish Government. Through this project we try
to develop general purpose guide robots able to participate in different events
providing useful information to the visitors. These general purpose guide robots
and their control software must be developed in such a way that their deploy-
ment in different environments should not involve much prior effort or software
development. In particular our general purpose guide robots must be able to
go round providing useful information along routes that have been previously
learnt. In fact to learn these routes our robot will have to follow a person along
them. In this paper we describe a strategy to get a person-following behaviour
that is robust, flexible to the environment and able to operate under dicult