Brígida Mónica Faria · btf@estsp.ipp.pt1,2
ARTIGO TÉCNICO
Gladys Castillo · gladys@ua.pt3 Nuno Lau · nunolau@ua.pt3 Luís Paulo Reis · lpreis@fe.up.pt4
CLASSIFICATION OF FC PORTUGAL ROBOTIC SOCCER FORMATIONS: A COMPARATIVE STUDY OF MACHINE LEARNING ALGORITHMS ABSTRACT Mobile robots have the potential to become the ideal tool to teach a broad range of engineering disciplines. Indeed, mobile robots are getting increasingly complex and accessible. They embed elements from diverse fields such as mechanics, digital electronics, automatic control, signal processing, embedded programming, and energy management. Moreover, they are attractive for students which increases their motivation to learn. However, the requirements of an effective education tool bring new constraints to robotics. This article presents the e-puck robot design, which specifically targets engineering education at university level. Thanks to its particular design, the e-puck can be used in a large spectrum of teaching activities, not strictly related to robotics. Through a systematic evaluation by the students, we show that the epuck fits this purpose and is appreciated by 90 percent of a large sample of students. KEYWORDS Knowledge Discovery, Data Mining; Support Vector Machines; RoboCup Soccer; Simulation; Formations.
I. INTRODUCTION RoboCup Simulation League has been one of the first competitions integrated on the RoboCup international project. The main goals of this league are concerned with developing the high-level decision and coordination modules of teams of robots [1]. The 2D simulation league has evolved over the years, but the principal architecture of the simulator is the same at it was firstly used in 1997 [1]. The Soccerserver is the simulator that creates a 2D virtual soccer field and the virtual players that are modeled as circles. This simulator implements the movement, stamina, kicking and refereeing models of the virtual world [2]. Another aspect that brings realism is the fact that the models in the simulator are taken both from real robots and from human like characteristics. The team of FC Portugal [2,3] has demonstrated very good results since its creation in 2000 and has won several European and World competitions [1]. The research focused development of the team is one of its main assets and still continues as every year new challenges are introduced. One concept that has been studied it is the usability of formations [4]. One important aspect is to be able to classify and predict the formations that are being used on
1
DETI/UA: Departamento de Electrónica, Telecomunicações e Informática, Universidade de Aveiro, IEETA - Instituto de Engenharia Electrónica e Telemática de Aveiro, Aveiro, Portugal.
2
ESTSP/IPP: Escola Superior de Tecnologia de Saúde do Porto, Instituto Politécnico do Porto, Porto, Portugal.
3
DMAT/UA: Departamento de Matemática, Universidade de Aveiro, Aveiro, Portugal.
4
DEI/FEUP: Departamento de Engenharia Informática, Faculdade de Engenharia da Universidade do Porto, LIACC: Laboratório de Inteligência Artificial e Ciência de Computadores da Universidade do Porto, Porto Portugal.
[4]
robótica
games. Another important aspect is to identify the opponent team that FC Portugal is playing with and its characteristics in the first moments of the game. In this paper a comparative study of three techniques for classification is presented. The following techniques have been used: Support Vector Machines (SVM) [5]; Artificial Neural Network (ANN) and k-Nearest-Neighbor.The environment tool used for machine learning and data mining experiments was RapidMiner [6]. This paper is organized with an initial explanation of the RoboCup Competition with special relevance for the simulation leagues. Next, an explanation and description of the three algorithms is presented. After that the kind of measures used to compare the classifiers and the statistical hypothesis to compare the average performance of the classifiers are presented. Finally the experimental results are presented along with some conclusions and future work.
II. ROBOCUP SOCCER RoboCup is an international cooperative project to promote Artificial Intelligence, Robotics and related fields. It is an attempt to promote artificial intelligence and robotics research by providing a standard problem where a wide range of technologies can be integrated and examined. The known goal of the RoboCup project is to create a soccer team with humanoid robots that can play and win to the world champion soccer team, by the year of 2050 [7]. In this project there are different leagues divided in two main groups: robotics and simulation. The first group involves physical robots with different sizes and different rules based on the competition that they integrate. The second one has the goal of, without the necessity to maintain any robot hardware, being able to research on artificial intelligence, coordination methodologies and team strategy. There is plenty of work performed and