bhuman10说明书

【实例简介】
这是世界冠军机器人足球队Bhuman发布的说明文档，对于理解代码和阅读代码很有帮助
CONTEN①S B-Human 2010 3.3.1 Blackboard 3.3.2 Module definition 24 3.3.3 Configuring providers 3.3.4 Pseudo-Module default 3.4 Streams 26 3.1.1 Streams Available 3.4.2 Streaming Data 3.4.3 Making Classes Streamable 3.5 Communication 3.5.1 Message Queues 3.5.2 Inter-process Communication 31 3.5.3 Debug Communication 32 3.5.4 Tcam communication 32 6 Debugging Support 3.6. 1 Debug requests 3.6.2 Debug Imag 33 3.6.3 Debug Drawings 35 3.6.4 Plot 3.6.5 Modify 3.6.6 Stopwa 4 Cognition 38 1.1 Perception 4.1.11efi nition of coordinate syste 4.1.1.1 Camera Matrix 39 4.1.1.2 Image Coordinate Systelll 41 4.1.2 Body Contour Provider 4.1.3 Image Processing 42 1.1.3.1 Segmentation and Region-Building 13 4.1.3.2 Region Classification 4.1.3.3 Detecting line 47 4.1.3.4 Detecting the Goal 4.1.3.5 Dctccting the Ball 50 4.1.3.6 Detecting Other Robots 51 4.2 Modeling 52 4.2.1 Self localization 52 4.2.2 Robot Pose validation 53 4.2.3 Ball Tracking 4 B-Human 2010 CONTENTS 42.4 Ground truth 4.2.5 Obstacle Mode 56 4.2.6 Robot Tracking 4.2.7 Largest Free Part of the Opponent goal 59 otion 60 5. 1 Sensing 61 5.1.1 Joint Data filtering 61 5.1.2 Ground Contact Recognition 61 5.1.3 Robot Model generation 62 5.1.4 Inertia Sensor Dala calibration 5.1.5 Inertia Sensor Data Filtering 5.1.6 Torso matrix 5.1.7 Detecting a Fall 64 5.2 Motion Cont rol 65 5.2.1 MoLion seleclion 5.2.2 Head Motions 66 5.2.3 Walking 5.2.3. 1 Inverse Kinematic 5.2.4 Special Actions 5.2.5 Motion Combination 72 6 Behavior Control 74 6.1 XABSL 6.2 Set ling Up a New behavior 6.3 Bchavior Uscd at RoboCup 2010 80 6.3.1 Button interface 6.3.2 Body Control 81 6.3.3 Head Control 6.3. 4 Kick Pose provider 83 6.3.5 Tactics 6.3.6 Role selector 6.3.7 Different roles 6.3. 7.1 Striker 6.3.7.2 Supporter 90 6.3.7. 3 Defender 6.3. 7.4 Keeper 93 6.3.8 Penalty Control 94 4 CONTEN①S B-Human 2010 6.3.9 Display cont 95 6.3.9.1 Right Eye 6. 3.9.2 Left eye 95 6. 3.9.3 Torso(Chest Button 6.3.9.4Feet 6.3.9.5E 7 Challe 97 7. 1 Dribble Challenge 7.2 Passing Challenge 7.3 Open Challenge 8 Irobot 101 8.1 Introduction 101 8.2 SceneⅤie 101 8.3 Information Views 101 8.3.1 Iimage Views 102 8.3.2 Color Space views 103 8.3.3 FieldⅤiews,.,,,,,,,,,,,.,,,,,,,,,,,,,,,,,,,,,104 8.3.4 Xabsl vicw 105 8.3.5 Sensor Data View 106 8.3.6 Joint Data view ..,106 8.3.7PlotⅤiews. 106 3.8 Timing View 107 8.3.9 Module views 8.3.10KickⅤiew, 107 8.4 Sccnc Dcscription Files 110 Console commands 111 8.5.1 Initialization Commands 111 8.5.2 Global Commands .112 8.5.3 Robot commands 112 8.5.4 Input Selection Dialog 119 8.6 Exarnples 119 8.6.1 Rccording a log fi 119 8.6.2 Replaying a log file 120 8.6.3 Remote Control 121 9 Acknowledgements 123 B-Human 2010 CONTENTS Bibliography 125 6 Chapter 1 Introduction 1.1 About us B-Human is a joint RoboCup team of the Universitat Bremen and the german Research Center for Artilicial Intelligence (DFKI. The teall was founded in 2006 and it consists ol numerous undergraduate students as well as of researchers of these two institutions. The latter have already bccn activc in a numbcr of roboCup tcams, such as the Gcrman Tcam and the Bremen Bytcrs (both Four-Legged League), B-Human in the Humanoid Kid-Size League, the BreDoBrothers (both in the Humanoid League and the Standard Platform League), and B-Smart( Small-Size League) The senior team members have also been part of a number of other successes, such as winning the RoboCup World Championship thrcc timos with the Gcrman Tcam(2004, 2005, and 2008) winning the roboCup german Open also three Uiles(2007 and 2008 wilh the GerimanTealll 2008 with B-Smart), and winning the Four-Legged League Technical Challenge twice(2003 and 2007 with thc gcrman tcam In parallel to these activities, B-Iluman started as a part of the joint team BreDoBrothers that has bccn a coopcration of the Technische Universitat Dortmund and the Universitat Bremen The team participated in the Humanoid League in RoboCup 2006. The software was based on previous works of the German Team [12 This team was split into two single Ilumanoid teams bccausc of difficultics in devcloping and maintaining a robust robot platform across two locations The dol! Bots from Dortmund as well as B-Human from Bremen participated in roboCup 2007 B-Iluman reached the quarter finals and was undefeated during round robin. In addition to the participation in the Humanoid League at the roboCup 2008, B-Human also attended a new cooperation with the Technische Universitat Dortmund. Hence, B-Human took part in the Two- Legged Competition of the Standard Platform League as part of the team BreDoBrothers, who reached the quarter finals. After the robo Cup 2008, we concentrated our work exclusively on the Two-Legged sPL. By integrating all the students of the Humanoid League team B-Human, the BreDobrothers would have had more than thirty members. Therefore we decided to end the cooperation by mutual agreement to facilitate a better workflow and work-sharing In 2009, we participated in the RoboCup German Open Standard Platform League and won the competition. We scored 27 goals and received none in five matches against different teams Furthermore, B-Human took part in the roboCup World Championship and won the compe lilion, achieving a goal ralio of 64: 1. In addilion, we could also win lirst place in the technical challenge, shared with Nao Team HTWK from Leipzig We repeated our successes in 2010 and won the German Open with a goal ratio of 54: 2 as well B-Human 2010 1.2. ABOUT THE DOCUMENT GMN00:800h7 P岛 詹 GERMAN OPEN 2010 Magdeburg 15.-18.Apri2010 MnI magd Figure 1.1: The majority of the team members at the german Open 2010 awards ceremony. as the RoboCup with an overall goal ratio of 65: 3 The current team consists of the following persons Students. Alexander Fabisch, Arne Humann, Benjamin Markowsky, Carsten Konemann Danicl Honscl, Emil Huscynli, Fclix Wenk, Fynn Fcldpausch, Jonas Peter, Martin Ring, Max Trocha, Michael Mester, Ole Jan Lars Riemann, Philipp Kastner, Bastian Reich, homas liebschwager, Tobias Kastner, Wiebke Sauerland Senior Students. Alexander Hartl. Armin Burchardt, Colin graf, Ingo Sieverdingbeck. Katha rina gillman, Thijs Jeffry de haas Researchers. Tim Laue. Judith Mille Senior Researcher. Thomas Rofer(team leader) 1.2 About the document As we wanted to revive the tradition of an annual code release two vears ago, it is obligatory for us to continue wit h it this year. This document, which is partially based on last years code release 23, gives a survey about the evolved system we used at RoboCup 2010. The changes made to code which was used at RoboCup 2009 are shortly enumerated in 1.3 Chapter 2 starts with a short introduction to the software required, as well as an explanation of how to run the Nao with our software. Chapter 3 gives an introduction to the software framework. Chapter 4 deals with the cognition system and will give an overview of our perception and modeling components. In Chapter 5, we describe our walking approach and how to create special motion patterns. Chapter 6 gives an overview about the behavior which was used at the 1.3. CHANGES SINCE 2009 B-Human 2010 RoboCup 2010 and how to create a, new behavior. Chapter 7 explains our approaches that we used to compete for the Challenges. Finally, Chapter 8 describes the usage of SimRobot, the prograln that is both used as simulator and as debugging FrOntend when controlling real robots 1. 3 Changes since 2009 The changcs madc sincc RoboCup 2009 arc dcscribcd in thc following scctions 4.1.3. 4 Detecting the Goal We developed a new GoalPerceptor that works on the image instead of the Region Percept and thus is morc efficient 4.1.3.5 Detecting the Ball In comparison to the ballPerceptor of the last years code release, the new BallPerceptor is morc indcpcndent of color tables, although not complctcly. In addition, it is morc cfficicnt and easier to maintain 4.1. 3.6 Detecting Other Robots We are developing a robot detection via vision which we did not have last year 4.2.2 Robot pose validation We introduced a new module to refine the loca lization provided by the selflocator 4.2.3 Ball Tracking The new BallLocator uses multiple Kalman filters instead of a Particle filter. 4.2.5 Obstacle Model The Obstacle Model incorporates vision informalion in addition to ultrasonic measurements 4.2.6 Robot Tracking To use the robots detected in our behavior, we implemented a Kalman filter that tracks the robots that are recognized through vision 4.2.7 Largest Free Part of the Opponent Goal The nodule is based on the visual robot detection and provides the part of the opponent goal to shoot at 5.1.7 Detecting a Fall To reduce the number of broken joints, we are now preparing a robot for the impact On the ground, i. e. we lower the power of the joints and turn the head 5.2. 3 Walking This years Walking Engine is a further development of last years approach with an im- proved method for controlling the center of mass motion and altered usage of sensor feed- back, which almost doubled the walking speed 6.3 Behavior Used at RoboCup 2010 The changes in this year's behavior are composed of new features(mainly the introduction of tactics, hand-to-hand handling, a new role assignment, a completely revised supporter, and a new way of approaching the ball) as well as using new modules replace parts of the behavior, namely the Kick Pose Provider and the kick engine biKe ng existing IIn the CodcRclcasc, therc is only a striker with a simplcgo-to-ball-and-kick"behavior implcmentcd which is not described in this document B-Human 2010 1. 3. CHANGES SINCE 2009 6.3.4 Kick pose provider The best robot pose to kick the ball will be calculated by the Kick Pose Provider 7.1 Dribble Challenge Our approach to solve the dribble challenge is mainly based on the obstacle model 7.3 Open Challenge We presented an approach to throw the ball back in that is fully integrated in our behavior 8.3.10 Kick view We developed a dynamic kick engine, the user interface of which is described in this report The general principles of the engine are published in 20 10 【实例截图】
【核心代码】