Two Methods for Autonomous Robot Obstacle
Sensing and Application Programming Interface for
Fuzzy Rule Interpolation
Roland Bartók
Department of Automation and Infocommunication
University of Miskolc
Miskolc, Hungary
qgeroli5@uni-miskolc.hu
József Vásárhelyi
Department of Automation and Infocommunication
University of Miskolc
Miskolc, Hungary
vajo@mazsola.iit.uni-miskolc.hu
Abstract— Environment detection is important task for an
autonomous robot. Obstacle avoidance is a must when the robot
do indoor activity. I this case robot movement is done in
corridors and rooms, which is similar to a maze. Distance from
the walls can be sensed with different sensors, in the experiment
infra-red sensors were used. There are several methods for
obstacle detection and calculating the distance from it. In the
experiment Fuzzy Rule Interpolation (FRI) and Bayes classifier
were used. On the other side a programmer friendly API was
created for relive the Declarative Description Language used for
different type of hardware. Using the FRI method to obstacle
detection a rulebase was defined for each wall position (front,
left, right). The Bayes classifier makes use of a big amount of
data, which are collected from the sensors. The collected data are
clustered for noise reduction and the amount of data is reduced.
The method gave 7 classes according to possible wall positions,
which appear in the maze around the robot. Both methods were
tested on a robot. The results were compared and described in
this paper.
Keywords—fuzzylogic; fuzzy interpolation; embedded system;
mobile robot; behavior based detection;
I.
I
NTRODUCTION
Today there are more and more mobile robots developed in
the world for different applications. Mobile robots can
transport the litter or patrol inside a building. There are robots
what transport the medical resources and hazardous waste in
the hospital [7]. Other robots are social robots, they are
interacting with humans and they are “living” together with
them [8] or working in offices. Some robots are not only social
robots, but they can reproduce a behavior of an animal and are
used ethological and behavior control researches [5].
An important task for mobile robots is environment
sensing. The main obstacles – except the moving people or
objects – are the walls around the robot. Position sensing of
obstacles helps to avoid a collision and the shape of walls can
be a reference point for the robot in indoor navigation.
This paper presents a robot movement in an example
environment, which is a maze, similar to interior movement in
a building. There will be presented two methods for the wall
sensing task. One of them is an application example using the
developed Fuzzy Rule Interpolation API. The second method
uses the Bayes theorem.
II. F
UZZY
R
ULE INTERPOLATION
AND BEHAVIOR DESCRIPTION
L
ANGUAGE
The easier way to describe a mobile robot behavior is using
a Behavior Description Language based on fuzzy automaton.
The fuzzy automaton uses fuzzy rule interpolation, which is
why enough to define the significant rules of behavior. In this
case the interpolation method is Fuzzy Interpolation Based on
Vague Environment (FIVE). [1]
A. The FIVE method
Fuzzy logic is a way of description for physical quantities
in vague expressions. For example the speed can be estimated
such as “The speed is bit slowly” or “The speed is too fast”.
The processes which are too complicated to be described by
mathematical equations, they can be described by fuzzy rules.
The fuzzy rules are like “IF … THEN …” sentences. The set of
these rules is the knowledge base. The complicated functions
are approximated by many of rules.
The rule input data is called antecedent and the output is
called consequent. In the classical fuzzy reasoning method all
the possible combinations of input variables must be defined
by a rule.
The complete rulebase generates a huge number of rules; it
needs large storage and large computation power. The number
of rules depends on task complexity. In that case when not all
possible rules are filled, the rulebase is called sparse or
incomplete. In this case there are observations, which not hit
the rule and no conclusion for observation. In this case
unpredictable side effect is generated.
Fuzzy Rule Interpolation (FRI) methods help to reduce the
size of the rulebase. Because of interpolation the significant
rules defined only. The FIVE method is an application oriented
simple method uses Shepard interpolation between similar
rules. The similarity is based on the distance between the rule
points. The FIVE method eliminates the fuzzyfication and
defuzzyfication steps in the controller design [1].
The described study was carried out as part of the EFOP-3.6.1-16-00011
“Younger and Renewing University – Innovative Knowledge City –
institutional development of the University of Miskolc aiming at intelligent
specialization” project implemented in the framework of the Szechenyi 2020
program. The realization of this project is supported by the European Union,
co-financed by the European Social Fund.
978-1-5090-4862-5/17/$31.00 ©2017 IEEE
87