INTERNATIONAL UNDERGRADUATE PROGRAM BINA NUSANTARA UNIVERSITY

Document Sample
INTERNATIONAL UNDERGRADUATE PROGRAM BINA NUSANTARA UNIVERSITY Powered By Docstoc
					              INTERNATIONAL UNDERGRADUATE PROGRAM
                    BINA NUSANTARA UNIVERSITY


                              Major Computer Science
                              Sarjana Komputer Thesis
                            Semester Even year 2004/2005

SOLVING ROBOCUP PROBLEMS USING DIFFERENT METHODOLOGIES:
    Agent Oriented Programming & Design, Expert System, and Genetic
                            Programming


                    David Kartopranoto               0500552985
                    Glenn Cornelius Layaar           0500550632
                    Henry Suryawirawan               0500563521


Abstract

RoboCup is an international competition in which teams of autonomous robots or
software softbots compete in simulated soccer matches to demonstrate cooperative
robotic techniques in a very difficult, dynamic, real-time, and noisy environment. This
thesis will only deal with creating software softbots soccer teams which were developed
using three different methodologies: Agent Oriented Programming and Design, Expert
System, and Genetic Programming. It is hoped that the results of the experiment would
be able to point out some advantages and disadvantages of using each methodology in
solving the RoboCup problem or similar type of problems.
To simplify the problem domain, we created our own simulator including the exclusion
of real time problem restriction that is introduced in the real RoboCup simulator. Each
methodology submitted a five-player soccer team, and along with two other benchmark
teams, would compete in a simulated league in which each team would meet each other.
The results show that it is possible to solve RoboCup problem using those three
methodologies which can perform well. A surprising result came through as the Genetic
Programming team, developed using machine learning technique, finished the league on
top of one methodology which was hand-coded with fully human intervention. This
shows promising outcome of the machine learning techniques to solve similar type of
problems.
Even though the result of the league was also subjective to each strategy used by each
team, we tried to draw a conclusion on the advantages and disadvantages of each
methodology by analyzing the general capability of it in dealing with such problems.
Hopefully, the analysis will be beneficial for further experiments on RoboCup problem
and also on using those three methodologies in solving other problems in different areas
as well.