A High Level Language Based on Rules for the RoboCup
Contents
1. we have eight available files each one with the name of a different type of player according to those found on the UvA Trilearn team We could expect for example that rules specifying the behavior for the goalie should be written into the goalie rules file and so on However it is important to point out that these files are only intended to provide an organized manner for describing the players It does not mean that the goalie rules file must have the rules for the goalie After all rules define the behavior of the player not the name of the file Furthermore it is not necessary for the user to use all the available files The user could choose to write rules for all the players in just one file This is possible if we use the changing of modes explained earlier The compiler used to parse the rules was programmed in C for Unix Fig 1 illustrates a schematic view of the compilation process The compiler analyzes the rules specified in the files for each player if everything was correct it will generate a file named PlayerTeams cpp which will contain the C code which will be linked along with the UvA Trilearn RULES sweeper win defender wing defender attacker attacker wing midfielder center midfielder centra defender Playerfeams cpp Fig 1 Schematic view of the compilation process system B Examples It is important to know that the order of the2. For example getProbToLoseBall returns the numeric probability of being tackled by some opponent 10 Server In order to make our language more robust it allows the user to have access to the values the soccer server handles directly Although all the high level conditions provide enough information to make decisions in certain situations it might be desirable to get detailed information in order to make better decisions For example serverSpeed returns the current speed of the player value between 0 and 1 2 Of course the use of this set of conditions requires knowledge about the soccer server C Commands Commands are the actions we want the player to perform when some conditions are satisfied As with the conditions we have defined what we think is a large number of commands A player might perform almost sixty commands during the game We provide the following classification 1 Formations Areas These commands are designed to work with the player s formation and its position For example moveToArea makes the player to move to a certain area and or zone Once the player reaches the area it will stop moving 2 Ball These commands are related with the use of the ball during the game For example kickToGoal makes the player to kick the ball to some desired position and speed to the opponent s goal interceptBall makes the player to move toward the ball as fast as possible in order to intercept the ball passToClosest all
3. Thus a lot of effort was put into solving certain low level aspects of the server such as the synchronization problem and the implementation to represent the player s world model It follows an object oriented approach so it is easy to modify and update with functionalities that were not planned originally Their code is also very well structured and documented And it was done considering the most recent version of the soccer server version 10 It can be argued that students could just work on some free source code such as the UvA Trilearn system However it still requires time to understand and get familiar with the server and their code Also it requires good programming knowledge in order to make changes and create their own players B Rules Rules are a natural way to describe situations where conditions must be satisfied in order to perform some action A rule is of the form If conditions then perform some action Rules can easily be used to code the behavior of a soccer player For example we could define the rule If I can score a goal then kick the ball to the goal Rules allow to create reactive players in the sense that given the current circumstances the player will decide which action to perform Rules are easy to read and understand especially if they can be written in natural language such as English And thus rules are also easy to create and modify For these reasons we decided to develop a language based
4. parameters that should be used within our language These will let us specify situations objects positions directions values etc We have classified these constants and parameters as follows 1 Play modes These constants are used to refer to some specific mode of play during the game The referee specifies these modes For example BEFORE_KICK_OFF tells us that the game has not started yet Other examples include FREE_KICK_US OFFSIDE_THEM PENALTY_US CORNER_KICK_THEM etc 2 Objects These constants refer to players and the ball For example TEAMMATE_1 specifies the reference to our first player Other examples include BALL OPPONENTS OPPONENT_6 ATTACKER GOALKEPPER etc 3 Directions For example EAST specifies the east direction Directions are relative to coordinates of the field from the player s perspective This means that they do not change even if teams swap sides in the halftime of the game Constants for distance are also specified such as FAR and NEAR 4 Positions These constants refer to specific areas and zones inside the field For example OPP_PENALTY determines the opponent s penalty area Other examples include OWN_CENTER ANY_ZONE BOTTOM_ZONE OPP_TOP etc 5 Actions These constants are used for actions such as Passing kicking dribbling tackling and marking For example A player may try to kick the ball into the goal at the TOPCORNER CENTER BOTTOMCORNER or RANDOM 6 Stamina We use the
5. A High Level Language Based on Rules for the RoboCup Soccer Simulator Jose I Nunez Varela Abstract The RoboCup Soccer Server is an excellent framework where students can learn and understand more about artificial intelligence A I application problems However getting to work with the server is not a straightforward task they are required to know specific details about the server which is time consuming and complex This paper describes the development of a high level language based on rules for the Soccer Server This language provides a full set of conditions and commands to create rules that will determine the behavior of a player in an easy and quick manner Thus students can focus on problems such as player s behaviors strategies coordination etc Index Terms A I Education Multi Agent Systems Rule Based Systems Soccer Simulation I INTRODUCTION OBOCUP is a research and educational initiative Its objective is to foster the A I and intelligent robotics research by providing a standard problem 4 There has been a continuous effort to increase the motivation participation and interest of students in science and technology The RoboCup soccer server provides an excellent platform for students to learn and apply A I techniques It does not require the use of real robots so students or universities who do not have the resources to participate in other leagues are not exempt for participating actively in the R
6. IHI IMPLEMENTATION Each condition and command that we defined for our language has to be translated into C code and be added into the UvA Trilearn system to create the executable file Therefore modifications and additions to the UvA Trilearn code were needed in order to allow a clean and organized translation Furthermore modifications were also needed to increase the functionalities of their code to allow the definition of more conditions and commands for our rules As mentioned before the UvA Trilearn code follows an object oriented approach One of our main additions was a class containing all the methods needed in order to perform every condition and command We consider this class to be a good addition to their code Anyone familiar with their code could create players easily using those methods Other classes were modified to add more functionality For instance the class defining the formations of the players was modified in order to have better formation stances among players A class defining areas and zones inside the soccer field was added This class is used to tell the players where to go or to specify their position We defined ten areas inside the field and each area is divided into nine zones Thus a player can easily be positioned in any part of the field A Constants and Parameters Before we start describing the conditions and commands available in our language we will briefly describe some of the constants and
7. implementing a high level language for the 3D simulation server And also implementing a high level language for coaches As we mentioned before we are planning to use the soccer server as part of an undergraduate A I course We hope that the students will take advantage of our implementation And also we expect that the interest in A I and robotics research can grow with this kind of projects Since it is also our goal to form a group of students with the desire to participate in some RoboCup competition in the future ACKNOWLEDGMENT The author wish to thank Dr Milos Hauskrecht for his helpful advice and assistance throughout this project REFERENCES 1 R de Boer J R Kok The incremental development of a synthetic multi agent system the UvA Trilearn 2001 robotic soccer simulation team M S thesis University of Amsterdam The Netherlands 2002 2 R de Boer J R Kok F Groen UvA Trilearn 2001 team description RoboCup 2001 Robot Soccer World Cup V Lecture Notes in Computer Science vol 2377 Springer Verlag pp 551 554 2002 3 E Foroughi F Heintz S Kapetanakis K Kostiadis J Kummeneje I Noda O Obst P Riley T Steffens 2001 RoboCup soccer server user manual for soccer server version 7 06 and later Available http sourceforge net projects sserver 4 H Kitano M Asada Y Kuniyoshi I Noda E Osawa RoboCup the robot world cup initiative Proc of The First Internationa
8. l Conference on Autonomous Agent Marina del Rey California United States 1997 pp 340 347 5 J Kok R de Boer 2003 The Official UvA Trilearn Source Code University of Amsterdam Available http www science uva nl jellekok robocup 6 I Noda H Matsubara K Hiraiki I Frank Soccer server a tool for research on multiagent systems Applied Artificial Intelligence vol 12 2 3 pp 233 250 1998 7 The RoboCup Federation the official RoboCup website Available http www robocup org
9. ndle this synchronization problem The sensory information contains only local information for every player So in order to make good decisions a player must have a global perspective of what is going on inside the soccer field For instance it is desirable to know our position and the positions of our teammates opponents the ball etc Therefore users should build a software system capable of creating a global representation of the game Players must send commands to the server in order to do something These commands are primitive and must contain local information also Good plays are coded with the combination of these commands However users must calculate the parameters needed to perform each command For example the kick command uses two parameters power and local angle then we must first calculate these two parameters in order to perform that kick We can see that the soccer server is a robust and excellent simulator system 3 Unfortunately it might be complex and difficult to understand and utilize for some people Especially if the simulator is intended to be used as an educational project For example during an A I course where students do not have enough time to familiarize with its functionality Also students might not be familiar with sockets programming and real time systems at all Therefore the development and implementation of a full soccer team requires a lot of effort and time that could sometimes be impossible
10. oboCup domain A RoboCup Soccer Server The soccer server is a powerful tool that simulates games of soccer in the best possible way The simulation is executed in a client server style via UDP IP sockets A soccer team will normally consist on 11 players 10 players and 1 goalie and a coach Each player and the coach are clients that would connect to the server Also a monitor is needed to visualize the simulation which is another client too Once all players are connected to the server the server will send sensory information to every player This information will help the player to make a decision on what action to perform next This action command is then communicated to the server which will simulate it on behalf of the player Manuscript received August 28 2007 This work was conducted at the Computer Science Department University of Pittsburgh J Nunez Varela was a M S student at the Computer Science Department University of Pittsburgh Pittsburgh PA 15260 USA He is now with the School of Engineering Universidad Autonoma de San Luis Potosi San Luis Potosi 62490 Mexico e mail jnunez uaslp mx The soccer server sends three types of sensory information at different intervals in time Furthermore the players should send their commands at certain intervals also so the server could simulate them correctly The consequence is that users should be familiar with real time systems and also they should code routines to ha
11. on rules The specific syntax for our language is the following if condition s then command s end We can combine conditions using logical operators amp amp for AND operation for OR operation and for negation or relational operators lt is greater than gt is less than lt is greater equal to gt is less equal to for is equal and for is different C Player Modes We have mentioned that the use of rules produces reactive behaviors But in certain circumstances that kind of behavior might not be useful Thus a planned strategy could work better Therefore we have incorporated the use of player modes Modes allow the user to group rules and to activate them later in situations they may find appropriate For example we could plan a certain strategy in the case of a corner kick or free kick We could also change the player s behavior if certain condition is satisfied For example if there are no opponents a defense could change its mode to attacker to try to score a goal even if it is not its primarily function Once it notices that it has been blocked by some opponent it will return to its normal behavior A mode is specified as follows mode name_of_mode rules endmode When we change to some mode we could also specify the frequency to which we would like to use that mode This means that a player may or may not change its mode adding a surprise factor to its behavior
12. ows the player to pass the ball to the closest teammate 3 Player This set of commands control the player s body For example turnBodyTo makes the player to turn its body toward some object 4 Goalie This set of commands are used exclusively by the goalie For example catchBall allows the goalie to catch the ball if it is within the catchable area 5 Communication These commands are used to communicate with other teammates for coordination and planning For example sayMessage makes the player to say some message 6 Game This set of commands are used for defending searching and moving For example searchFor makes the player to search for some object in the field Another example markTo allows the player to mark some specific opponent 7 Server As with conditions we decided to provide commands that the server understands directly We may want in some situations to have direct control of what the player should do For example serverDash makes the player to dash in the direction of its body with the supplied power 0 100 Users should know what kind of parameters and ranges the server accepts in order to use these commands IV FUNCTIONALITY Basically the user should write the rules for the players into a text file Then that file is compiled and a C file is generated which will be linked together with the UvA Trilearn code A Compiling the Rules Rules should be written into a predefined text file In fact
13. r 4 Communication This set of conditions are used for coordination and planning For example isHeardMessage determines whether the player heard a particular message 5 Player We use this set of conditions to know certain information about the player s perspective For example amICloseTo determines whether our player is close to some object Another example isHeadingToMe determines whether the ball a teammate or an opponent is moving towards the player s position 6 Game This set of conditions have to do with the status and some situations in the game and about the status of the ball For example isBallKickable determines whether the player is capable of kicking the ball Another example isInOffside determines whether some teammate is currently in an offside position 7 Direction These conditions require the specification of the direction in which they are to be verified For example numPlayersAroundMe returns the number of teammates or opponents inside a circle with center at the current player s position and with some radius 8 Pass These conditions help the player to decide whether or not is good to pass the ball For example canIPassToClosest determines whether or not it is possible to pass the ball to the closest teammate since it is probable we do not know its exact position 9 Tackle These conditions allow the player to decide whether is good to perform a tackle or to verify if somebody is going to tackle us
14. reate soccer players in a straightforward manner with little knowledge of the soccer server This is desirable for certain situations Here we talked about the possibility of using the soccer server as an educational project Students could work on the server to learn about multi agent systems coordination planning decision making among other things We believe that RoboCup competitions are attractive for a large number of students However if they find that using the soccer server requires too much effort they could lost interest very quickly Hence by using a language that allows them to create players easy and fast it will help them to maintain their interest Furthermore involving the students faster into the real problem might motivate them to want to know more about the server learn how it really works and how a player can be enhanced There are some upgrades we are planning to implement in our language First of all increase the set of conditions and commands to comprise more situations and to make the rules more flexible Also the whole compilation process could be done in more user friendly manner And the most important thing is to release our code so that it can be downloaded and used by anyone who might be interested We have not done it because there is no user manual that can be attached to the code yet So we have planned to complete the manual and set up a website to upload the project Further work might also include
15. rules determines their priority We can think of rules as being inside an infinite loop The program will check from the beginning of the loop for the satisfaction of some rule Once a rule is satisfied the command s will be executed and the loop will start all over again from the beginning This kind of logic is similar to the form humans play If we are running to some position but suddenly we can be able to kick the ball we are expected to do so and not to keep running Thus we set priorities in our actions To demonstrate how to create rules and how easy is to define players let s review some examples 1 Kiddie Soccer This is a common term used to refer to the way small kids play soccer At that age they do not have a good sense of coordination and teamwork and thus everybody try to kick the ball and to follow the ball through all the field We can build such kind of player with the following rules if isBallKickable then kickToGoal POS_RANDOM MAX_SPEED end if isKnownPosition BALL then searchFor BALL end if isKnownPosition OBJECT_BALL amp amp isMyCondition GOOD isMyCondition AVERAGE then moveToObject BALL end This example is also good to show that the ordering of the rules is important If we write the first rule at the end the player will never kick the ball it will only chase the ball Unless it gets tired and the ball is in a kickable distance 2 Attacker This example shows a ba
16. se The intention is to use the soccer simulator so that students can create their own soccer team and then organize a tournament at the end of the term II A HIGH LEVEL LANGUAGE While the functionality of the soccer server is somehow easy to comprehend the construction of just one player is complex and time consuming There is a large number of considerations that must be taken into account in order to build a software system that could work with the soccer server Hence the availability of a high level language easy to understand and use would result in an excellent way of allowing a large number of users the possibility of creating their own teams in a straightforward manner Requiring a small amount of knowledge about how the soccer server works A UVA Trilearn Team Since the establishment of the RoboCup competition several universities and research teams have spent a lot of time and effort in developing good software players And in an effort to expand the interest for this problem they have been releasing their source code so that other people could use it and work on top of it Avoiding the overhead of developing a new software system Since our interest was on the high level language we decided to work on top of an existing system We found that the UvA Trilearn simulation code 5 was our best choice One of their main objectives was to have a good base code that could be used in the following years after its creation in 2001
17. se constants to check the player s tiredness during the game The stamina level may be GOOD AVERAGE BAD VERY_BAD TERRIBLE 7 Parameters We can specify Boolean values and numeric values directly to evaluate some condition Boolean values are TRUE and FALSE And numeric values can be natural or real numbers B Conditions Conditions allow the user to verify situations or events during the game The number of rules that can be created is proportional to the number of conditions available Therefore our intention was to create a full set of conditions Trying to cover every possible situation in a soccer game We have defined around one hundred conditions Obviously they can be combined with relational or logical connectors to form more conditions We have classified the conditions as follows 1 Formations This set of conditions are used to determine the information about the team formation For example isInFormation determines whether a certain player is on its defined position as the current formation specifies 2 Areas Zones These conditions are designed to work with the areas and zones inside the field For example isClearAtArea determines whether or not there are teammates or opponents inside a certain area and or zone 3 Stamina These conditions are defined to let the player know its current and future condition For example isMyCondition determines whether the condition of the player is the same as the given paramete
18. sic behavior expected for an attacker The first rule makes the player to kick the ball whenever it is possible inside the opponent penalty area It will kick the ball to the goal randomly The second rule applies when there are no opponents in the direction of the penalty area The third rule will try to approach to some area where the ball can be kicked to the goal The fourth rule makes the player to try to get control of the ball if the player is closest to the ball with respect to its teammates The fifth rule makes the player to go to its position if it is idle And finally the sixth tule allows the player to know the position of the ball if isBallKickable amp amp amlInArea OPP_PENALTY ANY_ZONE then kickToGoal RANDOM MAX end if isBallKickable amp amp isRectClearAtDir OPPONENTS EAST FAR then kickToGoal RANDOM MAX end if isBallKickable then dribbleToDir EAST SLOW end if isKnownPosition BALL amp amp amIClosestTo BALL TEAMMATES then interceptBall end if amIInArea OPP_CENTERTOP CENTER_ZONE then moveToArea OPP_CENTERTOP CENTER_ZONE end if isKnownPosition BALL then searchFor BALL end V CONCLUSIONS AND FUTURE WORK This paper described the development and implementation of a high level language based on rules for the RoboCup soccer server Its main objective is to eliminate the inherent complexities found on the soccer server Thus allowing the users to c
19. to spend This might lead to the total loss of interest not only from the students but also from the professor B Objectives of Our Project Based on the idea of using the soccer server as an educational project we have developed and implemented a high level language based on rules that could be easy to use and understand Trying to eliminate the aforementioned problems and allowing the students to create a full team of players in a straightforward manner with little knowledge about the server and its complexities thus saving time and effort The language provides a complete set of directives in the form of conditions and commands These are used to create rules to specify the behavior of a player and ultimately the behavior of a full team This allows the students to quickly explore the creation of strategies team coordination and planning not worrying about such things as how to send a message to the server Since the main objective was to create a language with a large and functional set of conditions and commands there was no point in creating from scratch a working soccer system Therefore we chose the UvA Trilearn robotic soccer simulation team 1 as our platform to build on top of their system our language To this end we modified certain parts of their code and added some other functionalities that were not available in the original code It has already been planned to use the language as part of an undergraduate A I cour
Download Pdf Manuals
Related Search