This monograph is a comprehensive introduction to the field of soccer robotics. Soccer robotics has become an important research area integrating mechatronics, computer science and artificial intelligence techniques to create real-world autonomous systems. It also serves as a popular test arena in which to compare the different approaches, in diverse types of competition and with varying levels of distributed perception and collaboration. The focus of this monograph is the FIRA framework of Soccer Robotics, in particular MiroSot, which uses a central overhead camera to overview the whole soccer field and a central control of the robots. Soccer Robotics completely describes the different requirements to create a soccer team and details the hardware aspects, the computer vision needed, navigation, action selection, basic skills and game strategy. These aspects are described at an advanced undergraduate to beginning graduate level, resulting in a book not only useful as a text for courses but also indispensable for everyone who wants to participate in MiroSot robotics.
The Goals of Soccer Robotics
Fundamental Motion Benchmarks for Robot Soccer
Categories of Robot Soccer
The MiroSoT Robot Soccer System
Classification of MiroSoT Robot Soccer Systems
Purpose of This Book
Robot Soccer System: Hardware and Firmware Components
A Two-Wheel Command-Based Soccer Robot
How to Sense? Use Computer Vision Techniques
Binary Image Processing
Vision System For MiroSoT Robot Soccer
How to Decide and Act? Use Intelligent Systems and Control Techniques
Hybrid Control Architecture
Design of Robot Soccer Actions
Unified Navigation Control
How to Improve Intelligence? Use Soft Computing Techniques
Q-Learning: A Model-Free Reinforcement Learning Method
Fuzzy Logic and Control
Robot Soccer System: Software Components and Programming
MiroSoT Host Software Model
Programming Framework: An Overview
Basic Skill Functions
Applied Skill Functions
Game Strategy Development
Simulated Robot Soccer
How To Run the Simulator
Autonomous robots which are adaptable, communicative and objectiveoriented, and intelligent multi-agent robotic systems in general, are so evidently complex that it has become increasingly necessary to find a domain that can serve as an integrated framework for the complementary purposes of research and education. Robot soccer is one such suitable domain that is representative of intelligent multi-agent robotic systems, in which multiple robotic agents (or simply, multiple robots) need to cooperate in an adversarial environment to achieve specific objectives. It is a game based on the modified rules of human soccer and is played in a scaled down soccer field, in which two soccer robot teams compete by attempting to move a ball into the opponent team’s goal. The team with a higher score at the end of regulation time wins. Technically, robot soccer is a competitive game that makes heavy demands in all the key areas of robot technology, namely, mechanics, control, sensors, communication, and intelligence. On the one hand, it spurs wide-ranging multidisciplinary research work by providing a comprehensive test bed that facilitates the concrete demonstration and performance evaluation of new ideas and concepts. On the other hand, it captivates as an educational tool that helps students better understand and appreciate the scientific knowledge and technological developments in an inherently multidisciplinary setting of intelligent multi-agent robotic systems.
Since its inception in 1995, robot soccer has evolved into a recognized area of its own. This area, called Soccer Robotics, is a subfield of AI Robotics that offers a challenging domain for research and education in a large spectrum of issues integrating the problems of sensing, deciding and acting that are of relevance to the development of complete autonomous agents in general. The hope in Soccer Robotics, of course, is that by discovering how to get a team of robots to sense with acuity, decide collaboratively and act in coordination within the limited context of a soccer game, it will be possible to use the same techniques and technologies to build robots that carry out other more useful tasks. The development of this subfield is actively supported through the Micro-Robot Soccer Tournament (MiroSoT) and Simulated-Robot Soccer Tournament (SimuroSoT) Categories of the FIRA Cup, an international event organized by the Federation of International Robot-soccer Association (FIRA, http://www.fira.net). FIRA Cup, held annually since 1996, has been the ‘examination’ ground for the testing of new techniques and technologies integrated in the game of robot soccer, and has provided much excitement and entertainment for all those who participated. This new book Soccer Robotics is intended to be a comprehensive introduction to the field of soccer robotics, emphasizing breadth of coverage and accessibility of the material to readers with possibly diffe rent backgrounds. Its key feature is the emphasis placed on a robot soccer-programming framework that integrates all the key areas of robot technology. Until now, these areas had been treated mainly in separate books or in research literature only, outside the arena of soccer robotics.
A substantial portion of this book is based on the first author’s lectures EE006 Robot Soccer System at the Korea Advanced Institute of Science and Technology in the period July 13 – August 14, 1998. The material on robot soccer originated with the KAIST postgraduate theses of Dong-Han Kim (2003,1998), Yong-Jae Kim (2003), Hyun-Sik Shim (1998), Mun-Soo Lee (2000), Heung-Soo Kim (1997) and others, together with joint publications with the first author. The experimental robot soccer system program for Small League MiroSoT that supplements this book has been developed with the help of many students in the first author’s Robot Intelligence Technology (RIT) Lab at KAIST, while the simulator package for Large Leagu SimuroSoT has been developed by Bing-Rong Hong’s research team at Harbin Institute of Technology, P.R. China. Both are available for free download from the FIRA website http://www.fira.net. Soccer Robotics is written as a textbook for practical courses at the undergraduate level, and up to the first-year graduate level. It is useful for researchers and practising engineers interested in trying out new techniques in the domain of robot soccer. This book is also suitable for anyone interested in learning and developing robot soccer systems for edutainment purposes. For those interested in participating in either the MiroSoT or SimuroSoT categories of the annual FIRA Cup and other robot-soccer championship events, the material in this book will provide a firm foundation for the development of robot soccer systems to competitive standards. The book will be of interest to scientists, engineers and students in a variety of disciplines besides AI Robotics, where the use of robot soccer as a test bed is relevant: sensors (including computer vision), control, communication, multiagent systems and artificial intelligence.
To review the chapters briefly:
Chapter 1 defines the multi-agent framework of soccer robotics in terms of the three commonly accepted primitives of AI robotics, namely, SENSE, DECIDE and ACT. The goals of soccer robotics in research and education of intelligent multi-agent robotic systems are explained. The various categories of robot soccer created by FIRA, an international regulating body for robot soccer, are described. The classification of robot soccer systems for MiroSoT is also examined.
Chapter 2 presents the basic theoretical background on the mechanical motion of mobile robots, with emphasis on the kinematics of a two-wheel MiroSoT robot. The essentials of hardware and firmware needed to build a two-wheel MiroSoT robot with IR or RF communication are covered in sufficient detail.
Chapter 3 focusses on the (visual) SENSE primitive; in particular, it presents how the postures of target objects in robot soccer can be computed using centralized vision techniques. The basics of computer vision are first introduced. Real examples are then provided to highlight the practical considerations in building a good vision system for a MiroSoT team.
Chapter 4 focusses on the DECIDE and ACT primitives. A hybrid control architecture is introduced that integrates the three primitives of SENSE, DECIDE and ACT in a hierarchy of four interacting levels, namely, role, action, behaviour and execution. To expose the technical challenges involved, example strategies are given at the role level and action level. Action designs for robot soccer, to be implemented at the behavioral level, are classified and explained. An overview of classical PID control, applicable at the behavioral level and execution level, follows. Finally, two different navigation methods, applicable at the behavioral level, are presented.
Chapter 5 motivates the importance of the various aspects of intelligence, namely, search and evolution, knowledge representation and inference and learning and adaptation, as needed by the DECIDE and ACT primitives. Following, it demonstrates, by examples, the applicability of Petri nets, Q-learning, neural networks, evolutionary programming and fuzzy logic to robot soccer under the MiroSoT category. These soft-computing paradigms make concrete (at least one of) the abstract aspects of intelligence. For each paradigm, one or two examples are provided that address some key issues at specific hierarchical levels of the hybrid control architecture introduced in Chapter 4.
Chapter 6 introduces a host software model for MiroSoT robot soccer system. An overview of the programming framework for robot soccer is then presented, in which a number of the robot soccer concepts described in earlier chapters are illustrated through example ‘C’ programs which are the key functions of a robot soccer system for Small League (3-a-side) MiroSoT.
Chapter 7 complements the real-system programming framework presented in the previous chapter with a computer-simulated system programming framework. It presents the core simulator system and programming framework for Large League (11-a-side) SimuroSoT. Example ‘C++’ codes are provided for illustration. As do all authors of technical work, we wish to acknowledge the many contributors on whose work our own presentation is partly based. The list of references gives some indication of those to whom we are in debt. On a more personal level, we would expressly like to thank Hyun-Sik Shim, Myung-Jin Jung, Heung-Soo Kim, Kuk-Hyun Han, Kui-Hong Park, Ming Yu-Chi, Jun- Su Jang, Kang-Hee Lee, Jayyati Ghoshal and many other students in the RIT Lab who have contributed to this book in a wide range of invaluable ways.
This book would never have been possible without the funding that came from a variety of sources to support the research and development work in soccer robotics, and the writing of this book. The first author would like to acknowledge each of these agencies: LG, Samsung, POSCO, KOSEF and MRDEC. The last author would like to acknowledge the award of a ‘Brain Korea 21’ Institute Fellowship in 2002 that supported his joint research and authorship at KAIST.
Finally, the authors are indebted to Dr. Thomas Ditzinger, Engineering Editor at Springer Verlag, for editorial assistance, and quality production of this book.
برای دریافت این کتاب روی دکمهی زیر کلیک نمایین.