Robocup3D足球机器人体系结构与基本技能的研究与实现
|
摘要
Robocup组织的目标是到2050年前构建一支仿人形的机器人足球队,使它能够战胜当时的人类世界冠军队。Robocup比赛在推动产、学、研结合方面有着显著作用和极大的意义。厦门大学南强机器人足球队,是本文作者近几年来带领的一支机器人足球队,参加过国内外多次比赛,取得较好比赛成绩,同时在理论研究上也取得不少突破,本文是基于近几年的研究作出的一个总结。
Roboeup 3D是基于人形机器人仿真的一种比赛。Robocup 3D的仿真环境使用SPADES作为仿真基础,SPADES是一种并行智能体离散事件仿真系统(Systemof Parallel Agent Discrete Event Simulation)的简称,它所针对的并不是某一种特殊的仿真,而是抽象意义上的连续空间上的时间序列化仿真。同时Robocup 3D采用较为公认的ODE引擎作为物理仿真基础,使得系统可以更加逼真的模仿真实世界模型。
Robocup 3D客户端是我们研究的重点。厦门大学南强机器人足球队使用自重构机器人系统的特点构建了层结构的模型。机器人智能体在结构上包括底层、技术层和决策层。为了提高机器人的应变能力,在有限计算时间的情况下,采用基于多线程的异步方式来解决机器人世界模型的更新和上层决策之间的时间冲突。
机器人的基本技能包括基于视觉的场上目标定位(包括球的定位和人的定位)和球的跟踪能力。本文给出在采用ODEengine仿真条件下,基于抛体运动的球的轨迹预测算法,并且根据Peter提出的解析算法,实现了截球技能。
机器人的行走是Robocup 3D研究的重点,也是双足机器人研究的重点。机器人的行走问题首先是步态规划问题。步态规划有参考轨迹法、步行数据法、中枢模式发生器等方法。行走稳定问题有动力学方法和ZMP和FZMP理论。本文详述了ZMP、FZMP理论的一般原理,并且使用相关理论,结合Robocup的机器人结构,规划了Robocup机器人的行走动作和轨迹。为了提高在线仿真的计算效率问题,提出了一种基于自动调节的行走算法,并经验证取得较好效果。
在实现机器人行走规划的基础上,为了实现机器人的自适应行走,本文提出使用遗传神经网络来演化一个自适应控制神经网络。为了提高演化效率,提出了使用可缩减结构的神经网络数据结构,通过鼓励神经元数目的减少的进化激励机制,实现了加速演化计算。通过采集机器人行走异常样本,和构建合适的适应值方程,来训练神经网络,最后取得更好的行走效果,实现了机器人的自适应行走。
The ultimate goal of the Robocup Organization is to develop a team of fully autonomous humanoid robots that can defeat the human world champion team in soccer by 2050. Robocup has great significance to promote industry, academia and research. The AmoiensisNQ robot soccer team of Xiamen University is led by the author of this article. In recent years, it has taken part in several competitions at home and abroad, and has achieved favorable results. The team has made some breakthroughs in various areas of theoretical research. This paper summarizes the research of this team over recent years.
Robocup 3D is a competition that is based on humanoid robot simulation. The simulation is based on the software "SPADES", which is an abbreviation for "Simulation of Parallel Agent Discrete Events" It is an abstract time sequenced simulation in continuous space. Robocup 3D uses publicly recognized ODE as the basis of the physical simulation engine, making the system more realistic for imitating the real world.
A Robocup 3D robot is the focus of our study. This robot, according to the characteristics of the publicly supplied "self-reconstruction robot system", is constructed using the "layered structure model". Software agents of the robot include communication, skill, and decision-making layers in this structure. In order to satisfy the constraints imposed on the robot, e.g. limited calculation time, we used a multi-threaded asynchronous approach to resolve time conflicts between updating the robot world model and the time taken with decision-making.
The basic skill of the robot is vision based target tracing of the ball and other robots in the playground. In this paper, we use the ODEengine "simulation conditions and prediction" algorithm to track the ball, based upon its movement. We implemented our "ball chasing" abilities by using the analytic method suggested by Peter.
The "walking" of the robot is the focus of Robocup 3D, and is also important for general bipedal robot study. "Step planning" is the most important component for the "walking" of the robot. Generally speaking, step planning includes such methods as : - "reference path" method, "walk data" method, central pattern generation, and other methods. The stability of the walking involves both "dynamics equations" and the ZMP theory. This paper describes the general principle of the ZMP theory. Applying this theory to the structure of the Robocup robot, we plan the stepping and tracking of the robot. In order to lower the on-line simulation time, we created an "automatic walking adjustment" algorithm and achieved better results.
On the basis of robot walking planning, in order to realize the adaptive walk of robot, the paper proposes that using genetic neural networks to evolve an adaptive control neural network. In order to improve the efficiency of evolution, we use the neural network which data structure can be simplified; By incentive mechanisms that encouraging the reduction of the number of neurons in the evolution, we speed up the evolution of computing. By using the abnormal samples of Robot walk and creating a suitable fitness equation to train the neural network, we obtain a better walk results finally, and realize the adaptive walk of robot.
Roboeup 3D是基于人形机器人仿真的一种比赛。Robocup 3D的仿真环境使用SPADES作为仿真基础,SPADES是一种并行智能体离散事件仿真系统(Systemof Parallel Agent Discrete Event Simulation)的简称,它所针对的并不是某一种特殊的仿真,而是抽象意义上的连续空间上的时间序列化仿真。同时Robocup 3D采用较为公认的ODE引擎作为物理仿真基础,使得系统可以更加逼真的模仿真实世界模型。
Robocup 3D客户端是我们研究的重点。厦门大学南强机器人足球队使用自重构机器人系统的特点构建了层结构的模型。机器人智能体在结构上包括底层、技术层和决策层。为了提高机器人的应变能力,在有限计算时间的情况下,采用基于多线程的异步方式来解决机器人世界模型的更新和上层决策之间的时间冲突。
机器人的基本技能包括基于视觉的场上目标定位(包括球的定位和人的定位)和球的跟踪能力。本文给出在采用ODEengine仿真条件下,基于抛体运动的球的轨迹预测算法,并且根据Peter提出的解析算法,实现了截球技能。
机器人的行走是Robocup 3D研究的重点,也是双足机器人研究的重点。机器人的行走问题首先是步态规划问题。步态规划有参考轨迹法、步行数据法、中枢模式发生器等方法。行走稳定问题有动力学方法和ZMP和FZMP理论。本文详述了ZMP、FZMP理论的一般原理,并且使用相关理论,结合Robocup的机器人结构,规划了Robocup机器人的行走动作和轨迹。为了提高在线仿真的计算效率问题,提出了一种基于自动调节的行走算法,并经验证取得较好效果。
在实现机器人行走规划的基础上,为了实现机器人的自适应行走,本文提出使用遗传神经网络来演化一个自适应控制神经网络。为了提高演化效率,提出了使用可缩减结构的神经网络数据结构,通过鼓励神经元数目的减少的进化激励机制,实现了加速演化计算。通过采集机器人行走异常样本,和构建合适的适应值方程,来训练神经网络,最后取得更好的行走效果,实现了机器人的自适应行走。
The ultimate goal of the Robocup Organization is to develop a team of fully autonomous humanoid robots that can defeat the human world champion team in soccer by 2050. Robocup has great significance to promote industry, academia and research. The AmoiensisNQ robot soccer team of Xiamen University is led by the author of this article. In recent years, it has taken part in several competitions at home and abroad, and has achieved favorable results. The team has made some breakthroughs in various areas of theoretical research. This paper summarizes the research of this team over recent years.
Robocup 3D is a competition that is based on humanoid robot simulation. The simulation is based on the software "SPADES", which is an abbreviation for "Simulation of Parallel Agent Discrete Events" It is an abstract time sequenced simulation in continuous space. Robocup 3D uses publicly recognized ODE as the basis of the physical simulation engine, making the system more realistic for imitating the real world.
A Robocup 3D robot is the focus of our study. This robot, according to the characteristics of the publicly supplied "self-reconstruction robot system", is constructed using the "layered structure model". Software agents of the robot include communication, skill, and decision-making layers in this structure. In order to satisfy the constraints imposed on the robot, e.g. limited calculation time, we used a multi-threaded asynchronous approach to resolve time conflicts between updating the robot world model and the time taken with decision-making.
The basic skill of the robot is vision based target tracing of the ball and other robots in the playground. In this paper, we use the ODEengine "simulation conditions and prediction" algorithm to track the ball, based upon its movement. We implemented our "ball chasing" abilities by using the analytic method suggested by Peter.
The "walking" of the robot is the focus of Robocup 3D, and is also important for general bipedal robot study. "Step planning" is the most important component for the "walking" of the robot. Generally speaking, step planning includes such methods as : - "reference path" method, "walk data" method, central pattern generation, and other methods. The stability of the walking involves both "dynamics equations" and the ZMP theory. This paper describes the general principle of the ZMP theory. Applying this theory to the structure of the Robocup robot, we plan the stepping and tracking of the robot. In order to lower the on-line simulation time, we created an "automatic walking adjustment" algorithm and achieved better results.
On the basis of robot walking planning, in order to realize the adaptive walk of robot, the paper proposes that using genetic neural networks to evolve an adaptive control neural network. In order to improve the efficiency of evolution, we use the neural network which data structure can be simplified; By incentive mechanisms that encouraging the reduction of the number of neurons in the evolution, we speed up the evolution of computing. By using the abnormal samples of Robot walk and creating a suitable fitness equation to train the neural network, we obtain a better walk results finally, and realize the adaptive walk of robot.
引文
[1]A.Mackworth.On Seeing Robots[A].In Computer Vision:Systems,Theory,and Applications[C],1-13.World Scientic Press,Singapore,1992.
[2]The Robocup Federation.What is Robocup[DB/OL].hap://www.Robocup.org,1997-2005.
[3]P.Stone.Layered Learning in Multi-agent System[D].PHD thesis,Computer Science Department,Pittsburgh,Carnegie Mellon University,1998.
[4]P.Carpenter,P.Riley,M.Veloso,and G.Kaminka.AT-Humbolt Team Description[A].In Robocup-2000:Robot Soccer World Cup Ⅳ,Springer Verlag,Berlin,2001.
[5]M.Riedmiller,A.Merke,D.Meier,A.Hoffmann,A.Sinner,O.Thate,and C.Kill.Karlsruhe.Brainstormers2000 Design Principles[A].In Robocup-2000:Robot Soccer World Cup Ⅳ,Springer Verlag,Berlin,2001.
[6]L.P.Reis,J.N.Lau,and L.S.Lopes.FCPortugal Team Description[A].In Robocup-2000:Robot Soccer World Cup Ⅳ,Springer Verlag,Berlin,2001.
[7]Jinyi Yao,Jiang Chen,and Zengqi Sun.An Application in Robocup Combining Q-learning with Adversarial Planning[A].The 4th World Congress on Intelligent Control and Automation[C].WCICA,2002.
[8]Yunpeng Cai,Jiang Chen,Jinyi Yao,and Shi Li.Global Planning from Local Eyeshot:An Implementation of Observation-based Plan Coordination in Robocup Simulation Games[A].In Robocup-2001:Robot Soccer World Cup Ⅴ,Springer Verlag,Berlin,2002.
[9]Jinyi Yao,Jiang Chen,Yunpeng Cai and Shi Li.TsinghuAeolus2002 Basic Source Code[DB/OL].Available in http://166.111.249.36/Robocup/download/tsinghuaeolus/TsinghuAeolus_Source2002.tar.gz..
[10]Mao Chen,Ehsan Foroughi and Fredrik Heintz.Users Manual of Robocup Soccer Server[DB/OL].Available in http://sourceforge.net/projects/sserver.
[11]李实,徐旭明,叶榛,孙增圻.国际机器人足球比赛及其相关技术[J].北京:机器人,2000,22-5:420-426.
[12]石纯一等译.多Agent系统引论[M].北京:电子工业出版社,2003.10.
[13]顾洋.智能体学习与协作及其在Robocup中的应用[D].硕士学位论文,北京理工大学,2003.2.
[14]蔡建怀.Robocup仿真比赛研究[D].厦门大学,硕士学位论文,2005年.
[15]马培荪,曹 曦,赵群飞.两足机器人步态综合研究进展[J],南交通大学学报,2006.41(4).
[16]Sony Corporation.Sony dream robotQR IO[E].http://www.sony.net/Sonylnfo/QR IO/.
[17]SAUNDERS J B,INMAN V T,EBERHART H D.The major determinants in normal and pathological gait[J].Journal of Bone and Joint Surgery,1953,35A:543-558.
[18]Sarcos Inc..Human / Computer Interfac[EB/OL].http://www.sarcos.com/humanintfc.html.
[19]Ascension Technology Corporation.Motionstar reai2time motion cap ture[EB/OL].http://www.ascension2tech.com/products/motionstar_10_04.pdf.
[20]Northern Digital Inc..Optorak centus the competitive advantage[EB/OL].http://www.ndigital,com/certus2benefits,php.
[21]窦瑞军,马培荪,谢玲.两足机器人步态的参数化设计及优化[J].机械工程学报,2002,38(4):36-39.
[22]HUANG Qiang.Planning walking patterns for a biped robot[J].IEEE Transactions on Robotics and Automation,2001,17(3):280-289.
[23]B IZZI E,DpAVELLA A,SALTIEL P,et ai.Modular organization of sp inalmotor systems[J].The Neuroscientist,2002,8;437-442.
[24]NAKAOKA S.Recognition and generation of leg primitive motions for dance imitation by a humanoid robot[EB/OL]Proceedings of the 2nd International Symposium on Adap tiveMotion of Animals andMachines.Kyoto,2003;4-8.http://www.kimura.is.uec.ac.jp/amam2003/ABSTRACTS/E302nakaoka.pdf.
[25]KAZUO H,MASATO H,YUJ I H,et al.The development of honda humanoid robot[C],Proceeding of 1995 IEEE International Conference on Robotics &Automation.Leuven:IEEE Computer Society Press,1998:321-326.
[26]杨晶东,洪炳镕,黄庆成.双足足球机器人行走步态研究[J].哈尔滨工业大学学报,2005,37(7):876-878.
[27]赵晓军,黄强,彭朝琴,等.基于人体运动的仿人型机器人动作的运动学匹配[J].机器人,2005,27(4):358-361,379.
[28]胡洪志,马宏绪.一种双足步行机器人的步态规划方法[J].机器人技术与应用,2002(3):16-18.
[29]窦瑞军,马培荪.基于ZMP点的两足机器人步态优化[J].机械科学与技术,2003,22(1):77-79.
[30]包志军,马培荪,王春雨,等.用ZeroMoment Point描述类人型机器人步行稳定的不完善性探讨[J].上海交通大学学报,2001,35(1):68-71.
[31]Photos from Intel ISEF 2005[EB /OL].http://www.intel.com/pressroom/archive /photos/isef_2005_photos.htm.
[32]MATSUOKA K.Mechanisms of frequency and pattern control in the neural rhythm generators[J].Biol.Cybern.,1987,西南交通大学学报第41卷56:345-353.
[33]TAGA G,YAMAGUCH I Y,SHIMIZU H.Selforganized control of bipedal locomotion by neural oscillators in unpredictable environment[J].Biol.Cybern.,1991,65:147-159.
[34] MASAKIO, YUTAKA K, MASAH IRO A, et al. Reinforcement learning of humanoid rhythmic walking parameters based on visual information[J]. Advanced Robotics, 2004, 18(7) :677-697.
[35] BAY J S, HEMAM IH. Modelling of a neural pattern generatorwith coup led nonlinear oscillators[ J ]. IEEE Transactions on Biomedical Engineering, 1987, 34 (4) :297-306.
[36] M ILLER W T. Control of variable2speed gaits for a biped robot[J]. IEEE Robotics &Automation Magazine, 1999, 6 (3): 19-29.
[37] ZHEN Yuanfang. Reinforcement learning for a biped robot to climb sloping surface[J].Journal of Robotics System, 1997,14 (4) :283-296.
[38] FUMION. A motion learning method using CPG/NP [C]. Proceedings of the 2nd International Symposium on Adaptive Motion ofAnimals and Machines. Kyoto, 2003 : 428,http: //www. kimura. is. uec. ac. jp / amam2003 /PAPERS/E152nagashima. pdf.
[39] J IANG Shan, FUMION. Neural locomotion controller design and implementation for humanoid robot HOAP21[E]. Proceedings of the 20th Annual Conference of the Robotics Society of Japan. Osaka, 2002, http: //www. fujitsu. com /downloads/GLOBAL /Iabs/papers/hoap3. pdf.
[40] N ICHOLLS J G, MARTIN A R, WALLACE B G,et al. 从神经生物学到脑[M ].杨雄里译.北京:科学出版社, 2003 :522.
[41] LOHMEIER S, L; FFLER K, GIENGER M, et al. Computer system and control of biped " Johnnie" [C]. Proceedings ofthe IEEE International Conference on Robotics and Automation. New Orleans: IEEE Computer Society Press, 2004,4 : 222-227.
[42] SHUUJ I K. A realtime pattern generator for biped walking[C]. Proceeding of the 2002 IEEE international Conference on Robotics & Automation. Washington: IEEE Computer Society Press. ,2002:31-37.
[43] McGEER T. Passive dynamic walking[J]. Int. J. Robotics Res., 1990, 9 : 62-82.
[44] PRATT J , CHEW CM, TORRES A, et al. Virtual model control: an intuitive approach for bipedal locomotion [J]. The International Journal of Robotics Research, 2001, 20 (2) :129-143.
[45] STEVE C, MARTIJN W, ANDY R. A three2dimensional passive2dynamic walking robotwith two legs and knees[J]. The International Journal of Robotics Research, 2001, 20 (7) :607-615.
[46] STEVE C, ANDY R, RUSS T, et al. Efficient bipedal robots based on passive2dynamic walkers[ J ]. Science, 2005,307 : 082- 085.
[47] RUSSELL L T. Applied optimal control for dynamically stable legged locomotion [D].Massachusetts Institute of Technology, 2004.
[48] PRATT J . Exploiting inherent robustness and natural dynamics in the control of bipedal walking robots[D]. Massachusetts Institute of Technology, 2000.
[49]HAM ID B.Biped dynamic walking using reinforcement learning[D].University of New Hamp shire,1996.
[50]MOR IMOTO J,DOYA K.Acquisition of standup behavior by a real robot using hierarchical reinforcement learning[J].Robotics and Autonomous Systems,2001,36:37-51.
[51]车玲玲,王志良.ZMP理论在双足步行机器人步态控制中的应用[J],电子器件,2007,(4);.
[52]刘莉,王劲松,陈恳.基于六维力矩传感器的拟人机器人实际ZMP检测[J].机器人,2001,23(5):459.
[53]杨东超,汪劲松,刘莉.基于ZMP的拟人机器人步态规划[J].机器人,2001,23(6):504.
[54]林玎玎,刘莉,赵建东.双足步行机器人的ZMP2COP检测及研究[J].机器人,2004,26(4):368.
[55]殷晨波,周庆敏,徐海涵,杨敏基于虚拟零力矩点FZMP的拟人机器人行走稳定性仿真[J].系统仿真学报,2006,(9).
[56]傅祥志.两足步行机器人动态步行规划及仿真[J].华中理工大学学报,1999,27(3).
[57]SunzHo Lim,JinGeo Kim.Adaptive Gait Algrihm for IWR Biped Robot[E].IEEE Catalogue No.95 TH 8025.
[58]Terence C F,Wong Y S Hung.Stabilization of Biped Dynamic Walking Using Gyroscopic Couple[El.IEEE 1996
[58]张兰霞,解顺强,王立平,汪劲松,龙运佳,递推牛顿-欧拉动力学算法的推广及其应用[J].数学的实践与认识,2005,35(2).
[59]张伟杜,继宏.双足步行机器人的步态规划[J].计算机工程与应用,2002,(13).
[60]彭商贤等.试论国内外机器人机械学的发展趋势.机器人,1991,13(3):48-53.
[61]孙昌国,马香峰,谭吉林.机器人操作器惯性参数的计算.机器人,1990,12(2):19-24.
[62]Tourassis V D,Neuman C P.The inertial characteristics of dynamic robot models.Mechanism and machine theory,1985,20(1):41-52.
[63]曾庆华.结构惯性参数动态测试与识别.航空学报,1994,15(11):1315-1320.
[64]武志云,付利民.机器人操作手惯性参数的有限元算法[J].内蒙古工业大学学报,1995,14(3):7-11.
[65]王树新,张海根,黄铁球等.机器人动力学参数辩识方法的研究.机械工程学报,1999,35(1):23-26.
[66]熊有伦.机器人技术基础.华中理工大学出版社,1996(08).
[69]M Gautier,W Khalil.Direct calculation of minimum set of inertial parameters of serial robots.IEEE transactions on obotics and automation,1990,6(3):368-372.
[70]时小虎.Elman神经网络与进化算法的若干理论研究及应用[D].吉林大学博士论文,2006.
[71]阎平凡,张长水.人工神经网络与模拟进化计算.清华大学出版社,2000.
[72]Hoffmann A.Paradigms of Artificial Intelligence-A Methodolgical a Computational Analysis.Singapore:Springer,1998.
[73]张乃尧,阎平凡.神经网络与模糊控制[M].北京:清华大学出版社,1998.
[74]张立明.人工神经网络的模型及应用[M].上海:复旦大学出版社,1992.
[75]袁曾任.人工神经网络及其应用[M].北京:清华大学出版社,1999.
[76]Hagan,M.T,Demuth H.B.Beale M.Neural Network Design[M].USA:Thomson Learning,1996.
[77]Widrow B,.Neural Networks Application in Industry.Business and Science[J].Communication of the ACM,1994,37:93-105.
[78]Fu L.M.An expert network for DNA sequence analysis[J].IEEE Intelligent Systems&Their Applications,1999,14(1):65-71.
[79]Campos M.and Lopez D.Neural network approach to locate motifs in biosequenes[C].Progress in Pattern Recognition,Image Analysis and Applications,Proceedings 3773,214-221(2005).
[80]Armano G.,Mancosu G.,Milanesi L.,Orro A.,Saba M.and Vargiu E.A hybrid genetic-neural system for predicting protein secondary structure[J].Bmc Bioinformatics 6,(2005).
[81]Kato R.,Nakano H.,Konishi H.,Kato K.,Koga Y.,Yamane T.,Kobayashi T.and Honda H.,Novel strategy for protein exploration:High-throughput screening,assisted with fuzzy neural network.[J].Journal of Molecular Biology 351,683-692(2005).
[82]俞健.基于进化计算的神经网络设计方法[D].浙江大学博士论文,1998.
[83]张晓绩,戴冠中,徐乃平.一种新的优化搜索算法-遗传算法[J].控制理论与应用,1995,3:265-271.
[84]周金荣,黄道,蒋慰孙.遗传算法的改进及其应用研究[J].控制与决策,1995,3:261-264.
[85]孙艳丰,王众托.遗传算法在优化问题中的应用研究进展[J].控制与决策,1996,4:425-431.
[86]挥为民,席裕庚.遗传算法的运行机理分析[J].控制理论与应用,1996,3:297-304.
[87]席裕庚,柴天佑,挥为民.遗传算法综述[J].控制理论与应用,1996,6:697-708.
[88]丁承民,张传生,刘辉.遗传算法纵横谈[J].信息与控制,1997,1:40-47.
[89]张晓绩,方浩戴,冠中.遗传算法的编码机制研究[J].信息与控制,1997:134-139.
[90]孙艳丰,王众托.自然数编码遗传算法的最优群体规模[J].信息与控制,1996,5:317-320.
[91]张博,.小型双足人形机器人机构设计与步态规划[D].哈尔滨工业大学研究生论文2006
[92]P.Stone and D.McAllester.An Architecture for Action Selection in Robotic Soccer[A].In Proceedings of the Fifth International Conference on Autonomous Agents,2001.
[93]Russell Smith.Open Dynamics Engine v0.5 User Guide[E].http://www.ode.org/ode-docs.html.
[94]Vukobratovic.Biped Locomotion:Dynamics,Stability,Control And Application[M].Spring Verlag,Berlin,1990.
[95]M Vukobratovic,B Borovac.Zero-Moment Point-Thirty Five Years of Its Life[J].International Journal of Humanoid Robotics(S0219-8436),2004,1(1):157-173.
[96]K Yoneda,S Hirose.Tumble Stabilit Criterion of Integrated Locomotion and Manipulation[C].Proceedings of IEEE International Conference on Intelligent Robot and Systems.1996:870-876.
[97]A Gowami.Postural Stability of Biped Robots and the Foot-Rotation Indicator(FRI) Point[J].The International Journal of Robotics Research(S0278-3649),1999,18(6):523-533.
[98]K Harada,S Kajita,K Kaneko,H Hirukawa.Pushing Manipulation by Humanoid considering Two-kinds of ZMPs[C].Proceedings of the 2003 IEEE International conference on robotics & Automation.Taipei,Talwan,2003:1627-1632.
[99]A Goswami,V Kallen.Rate of changes of angular momentum and balance maintenance of biped robots[C].Proceedings of the 2004 IEEE international Conference on robotics &Automation.New Orleans,LA,2004:3785-3790.
[100]Honrnik,K.,et al.Mlultilayer Feedforward Networks Are Univesral Approximators[J].Neural Newtorks,1989,2:359-366.
[101]Girosi,F.and Poggio,T.,Newtorks and Best Approximation Proprty.Biological Cybernetics,1990,6:169-176.
[102]Chen T,Chen H,Lui R W.Approximation Capacity in C(R) by Multilayer Feedforward Networks and Related Problems[J].IEEETmas on NeraulNewtorks,1995,6(1).
[103]Chen Tianping et al,Approximation of continuous functions by NN with application to dynamic systems[J].IEEE Trans On Neural Newtokrs,1993,4(6):916-918
[104]Holland J.H.Adaptation in Natural and Artificial System[M].The University of Michigan Press,Ann Arbor,1975.
[105]Goldberg D.E.Genetic Algorithms in Search Optimization and Machine Learning[M].Addison-Wesley,Reading MA,1989.
[106]Narendra,K.S.et al.Identification and Control of dynamical Systems Usmg Neural Networks Networks[J].IEEE Trans.On Neural Networks.1990,1(1):4-27.
[107]韦柳涛,钱治航,夏金安,虞锦江,刘育骥,管家宝.基于遗传交换操作的多层神经网络算法[J].计算机工程,1993,6:22-25
[2]The Robocup Federation.What is Robocup[DB/OL].hap://www.Robocup.org,1997-2005.
[3]P.Stone.Layered Learning in Multi-agent System[D].PHD thesis,Computer Science Department,Pittsburgh,Carnegie Mellon University,1998.
[4]P.Carpenter,P.Riley,M.Veloso,and G.Kaminka.AT-Humbolt Team Description[A].In Robocup-2000:Robot Soccer World Cup Ⅳ,Springer Verlag,Berlin,2001.
[5]M.Riedmiller,A.Merke,D.Meier,A.Hoffmann,A.Sinner,O.Thate,and C.Kill.Karlsruhe.Brainstormers2000 Design Principles[A].In Robocup-2000:Robot Soccer World Cup Ⅳ,Springer Verlag,Berlin,2001.
[6]L.P.Reis,J.N.Lau,and L.S.Lopes.FCPortugal Team Description[A].In Robocup-2000:Robot Soccer World Cup Ⅳ,Springer Verlag,Berlin,2001.
[7]Jinyi Yao,Jiang Chen,and Zengqi Sun.An Application in Robocup Combining Q-learning with Adversarial Planning[A].The 4th World Congress on Intelligent Control and Automation[C].WCICA,2002.
[8]Yunpeng Cai,Jiang Chen,Jinyi Yao,and Shi Li.Global Planning from Local Eyeshot:An Implementation of Observation-based Plan Coordination in Robocup Simulation Games[A].In Robocup-2001:Robot Soccer World Cup Ⅴ,Springer Verlag,Berlin,2002.
[9]Jinyi Yao,Jiang Chen,Yunpeng Cai and Shi Li.TsinghuAeolus2002 Basic Source Code[DB/OL].Available in http://166.111.249.36/Robocup/download/tsinghuaeolus/TsinghuAeolus_Source2002.tar.gz..
[10]Mao Chen,Ehsan Foroughi and Fredrik Heintz.Users Manual of Robocup Soccer Server[DB/OL].Available in http://sourceforge.net/projects/sserver.
[11]李实,徐旭明,叶榛,孙增圻.国际机器人足球比赛及其相关技术[J].北京:机器人,2000,22-5:420-426.
[12]石纯一等译.多Agent系统引论[M].北京:电子工业出版社,2003.10.
[13]顾洋.智能体学习与协作及其在Robocup中的应用[D].硕士学位论文,北京理工大学,2003.2.
[14]蔡建怀.Robocup仿真比赛研究[D].厦门大学,硕士学位论文,2005年.
[15]马培荪,曹 曦,赵群飞.两足机器人步态综合研究进展[J],南交通大学学报,2006.41(4).
[16]Sony Corporation.Sony dream robotQR IO[E].http://www.sony.net/Sonylnfo/QR IO/.
[17]SAUNDERS J B,INMAN V T,EBERHART H D.The major determinants in normal and pathological gait[J].Journal of Bone and Joint Surgery,1953,35A:543-558.
[18]Sarcos Inc..Human / Computer Interfac[EB/OL].http://www.sarcos.com/humanintfc.html.
[19]Ascension Technology Corporation.Motionstar reai2time motion cap ture[EB/OL].http://www.ascension2tech.com/products/motionstar_10_04.pdf.
[20]Northern Digital Inc..Optorak centus the competitive advantage[EB/OL].http://www.ndigital,com/certus2benefits,php.
[21]窦瑞军,马培荪,谢玲.两足机器人步态的参数化设计及优化[J].机械工程学报,2002,38(4):36-39.
[22]HUANG Qiang.Planning walking patterns for a biped robot[J].IEEE Transactions on Robotics and Automation,2001,17(3):280-289.
[23]B IZZI E,DpAVELLA A,SALTIEL P,et ai.Modular organization of sp inalmotor systems[J].The Neuroscientist,2002,8;437-442.
[24]NAKAOKA S.Recognition and generation of leg primitive motions for dance imitation by a humanoid robot[EB/OL]Proceedings of the 2nd International Symposium on Adap tiveMotion of Animals andMachines.Kyoto,2003;4-8.http://www.kimura.is.uec.ac.jp/amam2003/ABSTRACTS/E302nakaoka.pdf.
[25]KAZUO H,MASATO H,YUJ I H,et al.The development of honda humanoid robot[C],Proceeding of 1995 IEEE International Conference on Robotics &Automation.Leuven:IEEE Computer Society Press,1998:321-326.
[26]杨晶东,洪炳镕,黄庆成.双足足球机器人行走步态研究[J].哈尔滨工业大学学报,2005,37(7):876-878.
[27]赵晓军,黄强,彭朝琴,等.基于人体运动的仿人型机器人动作的运动学匹配[J].机器人,2005,27(4):358-361,379.
[28]胡洪志,马宏绪.一种双足步行机器人的步态规划方法[J].机器人技术与应用,2002(3):16-18.
[29]窦瑞军,马培荪.基于ZMP点的两足机器人步态优化[J].机械科学与技术,2003,22(1):77-79.
[30]包志军,马培荪,王春雨,等.用ZeroMoment Point描述类人型机器人步行稳定的不完善性探讨[J].上海交通大学学报,2001,35(1):68-71.
[31]Photos from Intel ISEF 2005[EB /OL].http://www.intel.com/pressroom/archive /photos/isef_2005_photos.htm.
[32]MATSUOKA K.Mechanisms of frequency and pattern control in the neural rhythm generators[J].Biol.Cybern.,1987,西南交通大学学报第41卷56:345-353.
[33]TAGA G,YAMAGUCH I Y,SHIMIZU H.Selforganized control of bipedal locomotion by neural oscillators in unpredictable environment[J].Biol.Cybern.,1991,65:147-159.
[34] MASAKIO, YUTAKA K, MASAH IRO A, et al. Reinforcement learning of humanoid rhythmic walking parameters based on visual information[J]. Advanced Robotics, 2004, 18(7) :677-697.
[35] BAY J S, HEMAM IH. Modelling of a neural pattern generatorwith coup led nonlinear oscillators[ J ]. IEEE Transactions on Biomedical Engineering, 1987, 34 (4) :297-306.
[36] M ILLER W T. Control of variable2speed gaits for a biped robot[J]. IEEE Robotics &Automation Magazine, 1999, 6 (3): 19-29.
[37] ZHEN Yuanfang. Reinforcement learning for a biped robot to climb sloping surface[J].Journal of Robotics System, 1997,14 (4) :283-296.
[38] FUMION. A motion learning method using CPG/NP [C]. Proceedings of the 2nd International Symposium on Adaptive Motion ofAnimals and Machines. Kyoto, 2003 : 428,http: //www. kimura. is. uec. ac. jp / amam2003 /PAPERS/E152nagashima. pdf.
[39] J IANG Shan, FUMION. Neural locomotion controller design and implementation for humanoid robot HOAP21[E]. Proceedings of the 20th Annual Conference of the Robotics Society of Japan. Osaka, 2002, http: //www. fujitsu. com /downloads/GLOBAL /Iabs/papers/hoap3. pdf.
[40] N ICHOLLS J G, MARTIN A R, WALLACE B G,et al. 从神经生物学到脑[M ].杨雄里译.北京:科学出版社, 2003 :522.
[41] LOHMEIER S, L; FFLER K, GIENGER M, et al. Computer system and control of biped " Johnnie" [C]. Proceedings ofthe IEEE International Conference on Robotics and Automation. New Orleans: IEEE Computer Society Press, 2004,4 : 222-227.
[42] SHUUJ I K. A realtime pattern generator for biped walking[C]. Proceeding of the 2002 IEEE international Conference on Robotics & Automation. Washington: IEEE Computer Society Press. ,2002:31-37.
[43] McGEER T. Passive dynamic walking[J]. Int. J. Robotics Res., 1990, 9 : 62-82.
[44] PRATT J , CHEW CM, TORRES A, et al. Virtual model control: an intuitive approach for bipedal locomotion [J]. The International Journal of Robotics Research, 2001, 20 (2) :129-143.
[45] STEVE C, MARTIJN W, ANDY R. A three2dimensional passive2dynamic walking robotwith two legs and knees[J]. The International Journal of Robotics Research, 2001, 20 (7) :607-615.
[46] STEVE C, ANDY R, RUSS T, et al. Efficient bipedal robots based on passive2dynamic walkers[ J ]. Science, 2005,307 : 082- 085.
[47] RUSSELL L T. Applied optimal control for dynamically stable legged locomotion [D].Massachusetts Institute of Technology, 2004.
[48] PRATT J . Exploiting inherent robustness and natural dynamics in the control of bipedal walking robots[D]. Massachusetts Institute of Technology, 2000.
[49]HAM ID B.Biped dynamic walking using reinforcement learning[D].University of New Hamp shire,1996.
[50]MOR IMOTO J,DOYA K.Acquisition of standup behavior by a real robot using hierarchical reinforcement learning[J].Robotics and Autonomous Systems,2001,36:37-51.
[51]车玲玲,王志良.ZMP理论在双足步行机器人步态控制中的应用[J],电子器件,2007,(4);.
[52]刘莉,王劲松,陈恳.基于六维力矩传感器的拟人机器人实际ZMP检测[J].机器人,2001,23(5):459.
[53]杨东超,汪劲松,刘莉.基于ZMP的拟人机器人步态规划[J].机器人,2001,23(6):504.
[54]林玎玎,刘莉,赵建东.双足步行机器人的ZMP2COP检测及研究[J].机器人,2004,26(4):368.
[55]殷晨波,周庆敏,徐海涵,杨敏基于虚拟零力矩点FZMP的拟人机器人行走稳定性仿真[J].系统仿真学报,2006,(9).
[56]傅祥志.两足步行机器人动态步行规划及仿真[J].华中理工大学学报,1999,27(3).
[57]SunzHo Lim,JinGeo Kim.Adaptive Gait Algrihm for IWR Biped Robot[E].IEEE Catalogue No.95 TH 8025.
[58]Terence C F,Wong Y S Hung.Stabilization of Biped Dynamic Walking Using Gyroscopic Couple[El.IEEE 1996
[58]张兰霞,解顺强,王立平,汪劲松,龙运佳,递推牛顿-欧拉动力学算法的推广及其应用[J].数学的实践与认识,2005,35(2).
[59]张伟杜,继宏.双足步行机器人的步态规划[J].计算机工程与应用,2002,(13).
[60]彭商贤等.试论国内外机器人机械学的发展趋势.机器人,1991,13(3):48-53.
[61]孙昌国,马香峰,谭吉林.机器人操作器惯性参数的计算.机器人,1990,12(2):19-24.
[62]Tourassis V D,Neuman C P.The inertial characteristics of dynamic robot models.Mechanism and machine theory,1985,20(1):41-52.
[63]曾庆华.结构惯性参数动态测试与识别.航空学报,1994,15(11):1315-1320.
[64]武志云,付利民.机器人操作手惯性参数的有限元算法[J].内蒙古工业大学学报,1995,14(3):7-11.
[65]王树新,张海根,黄铁球等.机器人动力学参数辩识方法的研究.机械工程学报,1999,35(1):23-26.
[66]熊有伦.机器人技术基础.华中理工大学出版社,1996(08).
[69]M Gautier,W Khalil.Direct calculation of minimum set of inertial parameters of serial robots.IEEE transactions on obotics and automation,1990,6(3):368-372.
[70]时小虎.Elman神经网络与进化算法的若干理论研究及应用[D].吉林大学博士论文,2006.
[71]阎平凡,张长水.人工神经网络与模拟进化计算.清华大学出版社,2000.
[72]Hoffmann A.Paradigms of Artificial Intelligence-A Methodolgical a Computational Analysis.Singapore:Springer,1998.
[73]张乃尧,阎平凡.神经网络与模糊控制[M].北京:清华大学出版社,1998.
[74]张立明.人工神经网络的模型及应用[M].上海:复旦大学出版社,1992.
[75]袁曾任.人工神经网络及其应用[M].北京:清华大学出版社,1999.
[76]Hagan,M.T,Demuth H.B.Beale M.Neural Network Design[M].USA:Thomson Learning,1996.
[77]Widrow B,.Neural Networks Application in Industry.Business and Science[J].Communication of the ACM,1994,37:93-105.
[78]Fu L.M.An expert network for DNA sequence analysis[J].IEEE Intelligent Systems&Their Applications,1999,14(1):65-71.
[79]Campos M.and Lopez D.Neural network approach to locate motifs in biosequenes[C].Progress in Pattern Recognition,Image Analysis and Applications,Proceedings 3773,214-221(2005).
[80]Armano G.,Mancosu G.,Milanesi L.,Orro A.,Saba M.and Vargiu E.A hybrid genetic-neural system for predicting protein secondary structure[J].Bmc Bioinformatics 6,(2005).
[81]Kato R.,Nakano H.,Konishi H.,Kato K.,Koga Y.,Yamane T.,Kobayashi T.and Honda H.,Novel strategy for protein exploration:High-throughput screening,assisted with fuzzy neural network.[J].Journal of Molecular Biology 351,683-692(2005).
[82]俞健.基于进化计算的神经网络设计方法[D].浙江大学博士论文,1998.
[83]张晓绩,戴冠中,徐乃平.一种新的优化搜索算法-遗传算法[J].控制理论与应用,1995,3:265-271.
[84]周金荣,黄道,蒋慰孙.遗传算法的改进及其应用研究[J].控制与决策,1995,3:261-264.
[85]孙艳丰,王众托.遗传算法在优化问题中的应用研究进展[J].控制与决策,1996,4:425-431.
[86]挥为民,席裕庚.遗传算法的运行机理分析[J].控制理论与应用,1996,3:297-304.
[87]席裕庚,柴天佑,挥为民.遗传算法综述[J].控制理论与应用,1996,6:697-708.
[88]丁承民,张传生,刘辉.遗传算法纵横谈[J].信息与控制,1997,1:40-47.
[89]张晓绩,方浩戴,冠中.遗传算法的编码机制研究[J].信息与控制,1997:134-139.
[90]孙艳丰,王众托.自然数编码遗传算法的最优群体规模[J].信息与控制,1996,5:317-320.
[91]张博,.小型双足人形机器人机构设计与步态规划[D].哈尔滨工业大学研究生论文2006
[92]P.Stone and D.McAllester.An Architecture for Action Selection in Robotic Soccer[A].In Proceedings of the Fifth International Conference on Autonomous Agents,2001.
[93]Russell Smith.Open Dynamics Engine v0.5 User Guide[E].http://www.ode.org/ode-docs.html.
[94]Vukobratovic.Biped Locomotion:Dynamics,Stability,Control And Application[M].Spring Verlag,Berlin,1990.
[95]M Vukobratovic,B Borovac.Zero-Moment Point-Thirty Five Years of Its Life[J].International Journal of Humanoid Robotics(S0219-8436),2004,1(1):157-173.
[96]K Yoneda,S Hirose.Tumble Stabilit Criterion of Integrated Locomotion and Manipulation[C].Proceedings of IEEE International Conference on Intelligent Robot and Systems.1996:870-876.
[97]A Gowami.Postural Stability of Biped Robots and the Foot-Rotation Indicator(FRI) Point[J].The International Journal of Robotics Research(S0278-3649),1999,18(6):523-533.
[98]K Harada,S Kajita,K Kaneko,H Hirukawa.Pushing Manipulation by Humanoid considering Two-kinds of ZMPs[C].Proceedings of the 2003 IEEE International conference on robotics & Automation.Taipei,Talwan,2003:1627-1632.
[99]A Goswami,V Kallen.Rate of changes of angular momentum and balance maintenance of biped robots[C].Proceedings of the 2004 IEEE international Conference on robotics &Automation.New Orleans,LA,2004:3785-3790.
[100]Honrnik,K.,et al.Mlultilayer Feedforward Networks Are Univesral Approximators[J].Neural Newtorks,1989,2:359-366.
[101]Girosi,F.and Poggio,T.,Newtorks and Best Approximation Proprty.Biological Cybernetics,1990,6:169-176.
[102]Chen T,Chen H,Lui R W.Approximation Capacity in C(R) by Multilayer Feedforward Networks and Related Problems[J].IEEETmas on NeraulNewtorks,1995,6(1).
[103]Chen Tianping et al,Approximation of continuous functions by NN with application to dynamic systems[J].IEEE Trans On Neural Newtokrs,1993,4(6):916-918
[104]Holland J.H.Adaptation in Natural and Artificial System[M].The University of Michigan Press,Ann Arbor,1975.
[105]Goldberg D.E.Genetic Algorithms in Search Optimization and Machine Learning[M].Addison-Wesley,Reading MA,1989.
[106]Narendra,K.S.et al.Identification and Control of dynamical Systems Usmg Neural Networks Networks[J].IEEE Trans.On Neural Networks.1990,1(1):4-27.
[107]韦柳涛,钱治航,夏金安,虞锦江,刘育骥,管家宝.基于遗传交换操作的多层神经网络算法[J].计算机工程,1993,6:22-25