International Journal of Scientific & Engineering Research, Volume 5, Issue 12, December-2014

ISSN 2229-5518

247

Robot Soccer: A Review

Miss Tanvi Mahajan, Prof. A. S. Bhide, Mr. Ambarish Pundlik
Abstract : The robot soccer competition is a popular game that offers a set of challenges for intelligent agent researchers using a friendly competition in
a dynamic, real-time, multi-agent domain. Soccer is a dynamic game where things change fast and randomly so it cannot be predicted i.e. the robots
cannot be pre-programmed. A lot of work has been done in the field of robot soccer to keep the game as close as possible to th e real game of soccer.
This paper is concentrated on various strategies used in robot soccer.
Key Words : Robotic Soccer, Fira, Kalman Filter, Q-learning, Co-ordinate theory, Android

1. INTRODUCTION
The development of robots not only provide a great deal
of business opportunities of technology industry, but also
have great impact on areas such as national defense,
security, domestic life, medical treatment, rescue works,
exploitation in deep ocean, etc. According to different
applications, robots can have different contribution in
different areas. One of the challenging issues in robotic
research is the cooperation, coordination, and negotiation
among distributed agents in a multi-agent system.
To speed up the development on this challenging
issue, a group of Korean researchers initiate a robotics
soccer game called Fira in 1995. Later the league becomes a
well-known world competition, called Federation of
International Robot Association (Fira). A soccer robot is a
specialized autonomous robot and mobile robot that is used
to play variants of soccer.
A robot soccer system is multi-agent intelligent control
system composed of two or more robots, vision system,
communication equipment and a computer. Each robot in
this system has its own movement mechanism and
therefore it can move independently as well as cooperate
with other robots. In order to allow researchers to consider
many different possibilities, several various leagues are
formed by FIRA Soccer Simulation League
Small Size League
Standard Platform League
Middle Size League
Humanoid League.
Besides the above mentioned leagues some more leagues
have been introduced during recent years which focus on
different applications than soccer playing, i.e. rescue
scenarios, educational robots, and human-robot interaction.

Fig. General System Configuration

The vision system consists of a camera which is

used to capture the real time images from the field &
transmits the information to computer. The strategy system
allows the robot to determine the next action to be taken
depending on the information obtained from vision system.
The on-board control system receives data from
communication system and accordingly action is taken. The
robot system actually controls mechanical elements like
actuators, sensors, etc.
Up till now, various platforms & techniques are
used in order to make the robot soccer as same as possible
to the real soccer game. This paper is focused on the
following different techniques used in robot soccer 3D coordinates theory & measurement of angle
between robot & object or goal.
Adaptive Q-learning.
Color tag identification.
Artificial immune network based action selection
method & reinforcement learning mechanism.
Path planning algorithm based on Bezier curves.
Rete algorithm.
Kalman filter algorithm.
Fuzzy logic & Genetic algorithm.
2.1 3D Coordinates Theory & Measurement of Angle
between Robot & Object or Goal: In this method, the
obstacle can be avoided by developing an algorithm using
3-dimensional coordinate system. Here few calculations are
made by using coordinate theory inclusive of
i) Distance of goal from current position of robot.
ii) Distance between the robot & nearest obstacle. e
iii) Angle between the current position of robot & goal.
iv) Angle between the current position & nearest obstacle.
The above mentioned calculations are done on the basis of
the real time data received from vision system. Depending
on this data, the robot generates linear as well as angular
velocity.
The object is detected by the real time image data
& then reconstructed to interpret the original information.
For this purpose, three dimensional coordinate system is

IJSER

2. OVERVIEW
In general, the architecture of robot soccer has
following different sections Vision System
Strategy System
Communication System
On board Control System
Robot System
Following figure shows the general configuration as
mentioned above-

IJSER 2014
http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 5, Issue 12, December-2014
ISSN 2229-5518

used as it does not depend on current position of robot. The

origin of the 3-D camera coordinate system can be selected
randomly by the focal point of the camera. The coordinates
can be determined by using stereo technique which
requires two images of the same object from different
orientation. Both the images should be such that
corresponding pixels from the image region must detect the
desired object. Stereo triangulation uses these geometrical
conditions & in this way, the distance of the robot from the
object is determined. When the coordinates has been
determined, the robot will follow the algorithm & execute
the best action possible according to the algorithm.
2.2 Adaptive Q-Learning: Using adaptive Q-learning
method, robots are able to learn from past experiences.
These experiences are then used to enhance their
performance. This method is comprised of hierarchical
structure. In the first layer of hierarchy, role (attacker or
defender) is assigned to each robot as per their position. In
second layer, two algorithms are developed to assign the
role depending on the decision from first layer. In third
layer, each robot takes an action & completes the task as per
their role. This last layer is known as behavior layer.
As the game is dynamic & random in nature, no
robot is having a fixed role. Instead their roles are
exchanged randomly which is based on the real time data
input from vision system. Such a random exchange of roles
can be accomplished by using Adaptive Q-learning
strategy.
2.3 Color Tag Identification: This method uses robots
wearing a jacket of specific color. Due to specific color
jacket, vision system can identify the robots of home team
effectively & efficiently. It also reduces the complexity of
algorithms. This makes the game to be played in a real time
environment.
The jacket of a robot consists of two colors - one
represents team & another color represents an individual
player. The vision system scans the image for these two
colors & calculates the centroid of each color. The centre of
gravity is then calculated by calculating the mean of both
the centroids. As the central portion of the robot is painted
with a light as well as dark color, by calculating the
grayscale values of pixels, a value & thereby the role is
assigned to each robot. This not only improves the accuracy
but also reduces the complexity.
2.4 Artificial Immune Network Based Action Selection
Method & Reinforcement Learning Mechanism: In robot
soccer, assignment of role & action to be executed play an
important role. The role assignment & action execution can
be done using geometrical approach i. e. coordinate theory.
But it is quite complicated to design an algorithm for action
selection using coordinate theory. This can be overcome by
using artificial immune network.
In this technique, the perception of the biological
immune system is used to represent the perception of the
robot soccer team while reaction of biological immune
system is used to represent the response of robot soccer
team. The field is divided into four quadrants. Like
perception of biological immune system, the robot detects
the closest obstacle as well as ball in the quadrants using

248

coordinate system. Depending on this information, robots

determine the strongest & nearest defender or attacker in
the field.
Reinforcement learning mechanism helps the
robots to decide the priority order. This method decides
whether the response of the robot matches to the priority
order or not. If it matches reward is given and if it does not
penalty is given.
The use of artificial intelligence as well as
reinforcement learning mechanism reduces the overall
complexity.
2.5 Path Planning Algorithm Based on Bezier Curves:
Bezier curve is generally used for animation as well as
computer graphics. Bezier curves have an important
property for trajectory generation. The algorithm
developed using Bezier curve makes the trajectory of the
robot to follow the desired path.
2.6 Rete Algorithm for Robot Soccer: The Rete algorithm is
mainly used for designing expert systems those based on
particular rules. Expert systems are the part of artificial
intelligence capable of simulating human decision skills.
The Rete algorithm is used to define the architecture of the
system. It treats the low level information received from the
field as high level soccer concept. As the expert system
operates in an infinite loop, the Rete algorithm does not
check all the rules in each cycle, rather it takes reference
from previous results. Also, the rete algorithm reduces the
number of tests to be carried out in each cycle as few of the
rules can be shared in network.
2.7 Kalman Filter Algorithm: In soccer, the coordinates as
well as the direction of the robot must be known in order to
determine the position of the robot in real time. This can be
done by using trigonometric functions. But such functions
do not consider the noise in real time which results in an
error in the position of the robot in the field.
The Kalman filter integrates the measured data
into an estimate by considering that measurements are
noisy. The Kalman filter smoothens the effect of noise by
employing large information from the reliable source. The
Kalman filter algorithm can easily combine the information
from vision system & locomotive system in real time.
2.8 Fuzzy Logic & Genetic Algorithm: Fuzzy logic is
generally used when the information is uncertain or vague.
Fuzzy system makes decision by applying approximate
reasoning depending on the vague & dynamic data. But the
fuzzy system is not able to modify the rules on its own for
the changing environment. Genetic algorithms can adopt
based on natural selection.
Thus by combining the property of fuzzy system of
approximate reasoning & the property of genetic algorithm
of self adaptive learning, a hybrid system is designed that
meets the expectations of robot soccer.
2.9
Android Based Robotic Soccer: Till now, the
communication, object detection was based on computer
technology. The same thing can be implemented on
android platform. Various algorithms which are stated
above can be employed on android base.
In this methodology, tagging scheme is used. The
object to be followed is identified from the image by

IJSER

IJSER 2014
http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 5, Issue 12, December-2014
ISSN 2229-5518

tagging it. & then depending upon the role of each robot
the task is accomplished.
3. Conclusion & Future Work
In this paper, different techniques as well as
various algorithms are described which are successfully
used in robotic soccer. Todays era is Smartphones era. Use
of android platform makes communication faster &
efficient. The use of android can be further extended for
robotic soccer. Android based object detection algorithm
could be used as one of the strategy in real robot soccer
competition.

REFERENCES
[1] Awang Hendrianto Pratomo, Anton Satria Prabuwono, Mohd.
Shanudin Zakaria, Khairuddin Omar, Md. Jan Nordin,
Shahnorbanun Sahran, Siti Norul Huda Sheikh Abdullah and
Anton Heryanto, Position and Obstacle Avoidance Algorithm in
Robot Soccer, Journal of Computer Science 6 (2): 173-179, 2010
ISSN 1549-3636
[2] Kao-Shing Hwang, Shun-Wen Tan, and Chien-Cheng Chen,
Cooperative Strategy Based on Adaptive -Learning for Robot Soccer
Systems, IEEE Transactions On Fuzzy Systems, Vol. 12, No. 4
[3] Ayan Banerjee, Ramanpreet Singh Arora, Aruna Chakrabarty, A
Dynamic Identification Method for Robotic Soccer Game,
International Journal of Wisdom Based Computing, Vol. 1 (3).
[4] Yin-TienWang, Zhi-Jun You, and Chia-Hsing Chen, AIN-Based
Action Selection Mechanism for Soccer Robot Systems, Hindawi
Publishing Corporation Journal of Control Science and
Engineering Volume 2009, Article ID 896310
[5] Michael Angelo A. Pedrasa, A Path Planning Algorithm for Soccer
Playing Robots Based on Repeated Modification of Bezier Polynomials,
Philippine Engineering Journal Pej 2005; Vol. 26 No. 1:120
[6] Savitha G Venugopal P S Dr. Sarojadevi Dr. Niranjan Chiplunkar,
An Approach for Object Detection in Android Device, 2014 Fifth
International Conference on Signals and Image Processing.
[7] K. Matusiak1, P.Skulimowski1 And P. Strumillo1, Object
Recognition In A Mobile Phone Application for Visually Impaired
Users, Sopot, Poland, June 06-08, 2013

IJSER

IJSER 2014
http://www.ijser.org

249