四足机器人步态规划与平衡控制研究
|
摘要
四足步行机器人是机器人家族的一个重要分支,其不仅承载能力强,而且容易适应不平的地形。它既能使用静态稳定的步态缓慢平滑地行走,又能以动态稳定的步态跑动;如果安装了操作手,在站立时还可以成为一个稳定和灵活的工作平台。因此,四足步行机器人是步行机器人中最有可能首先实用化的机型之一。在对四足步行机器人的研究中,步态规划及运动平衡控制是具有重要意义的课题。为了有效地提高机器人的行走速度和能量效率,增强机器人在行走过程中对不平地形的适应能力,本文以AIBO四足机器人为研究平台,从以下三个方面对四足步行机器人的运动控制进行了研究:
1.四足机器人的步态规划及优化
四足机器人最常用的步态规划方法是基于运动学模型的步态控制方法。实验表明,使用该方法规划的轨迹与实际运动轨迹不相符,这一问题造成机器人无法准确按照预先设计的方式运动。因此,本文通过对机器人行走时身体摇摆现象的研究,在运动学模型控制法的基础上引入动力学规划,提出一种使用零力矩点轨迹规划的步态控制方法:控制机器人运动时的身体姿态来改变机器人足部与地面形成的支撑多边形,同时使用零力矩点轨迹规划方法生成机器人在支撑相时足部的轨迹。该方法有效地使用了基于模型的步态控制方法,降低了采用完全动力学规划方法的计算复杂度,满足了算法实时性的要求。设计的控制器具有大量的参数,因而本文使用基于遗传算法的进化学习方法对步态控制参数进行优化。由于有效地使用了基于运动学规划的步态参数进行种群初始化,算法收敛较快。此外,在适应度函数的选取上充分考虑了稳定性这一因素,因而大大提高了最优步态的稳定性。实验使用AIBO机器人进行测试,机器人使用该进化学习方法可自主地得到最优步态,其最优步态在保证稳定性的基础上速度达到了455 mm/s。实验结果表明,应用该方法进行步态控制,机器人获取的最优步态不仅满足稳定性要求,而且对不平地形也具有较好的适应能力。
2.四足机器人的平衡控制及优化
针对机器人在正常运动过程中突然遭遇外力作用引发翻倒危险的问题,各国科学家们致力于提出各种“参考点”作为步行机器人的稳定性判定标准。诸多成果都是使用仿真机器人进行模拟实验,证明了相关理论的有效性,而在实体机器人上,机器人自身传感器不可避免地存在各种噪声和误差,导致机器人无法确切地感知外部世界并得到自身状态的准确信息,因此本文针对实体机器人引入回归映射模型有效地预测翻倒状态,提高了机器人状态检测的精度与准确度。在此基础上,为了迅速恢复机器人的站立状态,本文提出一种基于轨道能量模型的步行机器人平衡控制方法:采用姿态控制和“迈单步策略”两种控制策略实现机器人的平衡控制。同时,采用离线监督学习方法对理论落脚点模型进行优化。实验使用AIBO机器人进行测试,机器人实际的翻倒次数减少了一半以上,证明了该方法的有效性和可行性。
3.四足机器人运动控制系统
四足机器人需要能够实现快速的直线行走、灵活的转弯和紧急制动;当机器人遇到突发状况身体失去平衡时,需要判断出身体是否处于将要翻倒的危险状态,并主动恢复身体平衡;当机器人不幸翻倒后,需要实现翻倒后的快速站立恢复;当机器人成为一个操作平台时,需要通过控制其全身的运动,完成操作物体等各项任务。因此,本文综合考虑四足机器人实际的应用需求,设计了一整套运动控制系统,各层之间任务明确,要求具体,相互关联,相互制约,能够有效地协调四足机器人的各种运动行为,满足多变的应用需求。
第六章给出结论和一些值得进一步研究的问题。
Quadruped walking robot is an important branch of robot. Quadruped walkingrobot not only has heave loading capacity, but also adapts to the all-terrain variety moreeasily. It could not only move slowly and smoothly with statically stable gait, but alsorun with dynamically stable gait. If installed with a manipulator, it will become a sta-ble and flexible working platform while standing. Therefore, quadruped walking robotcould be one of the first practical physical types. On the study of quadruped walk-ing robot, gait planning and balance control of quadruped robot is the subject withgreat significance. In order to improve speed of travel and energy efficiency effec-tively, and enhance the ability of adapting the rough terrain during walking, we takethe quadrupedal walking robot AIBO as research platform and study motion control ofquadrupedal legged robot in the following three aspects:
1. The gait planning and optimization of quadruped robotThe most common method of gait planning of quadruped robots is the controlmethod based on the kinematic model. Experimental results show that the planning tra-jectory using this method is not identical with the actual trajectory of locomotion. Thisproblem causes that the robot can not move following the preliminary mode designedexactly. Therefore, this paper presents a gait control method using ZMP (zero-momentpoint) trajectory planning by studying the body sway phenomenon of robots duringwalking, which introduces the dynamic planning based on the the control method ofthe kinematic model: control the body posture during walking to change the supportpolygon formed by the robot foot and the ground, and in the mean time use ZMPtrajectory planning to generate the foot trajectory during stance phase. This methoduses the model-based gait control method effectively, thus reduces the computationalcomplexity of totally using dynamic planning method, and satisfies the need of real-time algorithm. Since the designed controller contains a large number of parameters,we also use an evolutionary learning method based on genetic algorithm to optimizethose gait control parameters. Since we effectively use the gait parameters based on thekinematic planning to initialize generation, the algorithm constringency is much better.Moreover, we take full account of the stability factor when choosing the fitness func-tion, the stability of the optimal gait is improved. In the experiments, the AIBO robotcan obtain the optimal gait autonomously, and its maximum speed achieves 455 mm/s with stability. The experimental results show that the robots can obtain the optimal gaitwith better stability and stronger adaptability to uneven terrain by using the proposedgait control method.
2. The balance control and optimization of quadruped robot
Concentrated on the danger of turnover imposed by sudden external shocks ofrobots during normal walking, scientists in various countries devote their lives to propos-ing different”reference point”which could be used as the stability criteria of walkingrobot. Most results are carried out with simulation experiments using simulated robots,which can prove the validity of related theory. However, for the entity robot, due to allkinds of errors and noises caused by sensors themselves, the robots can’t tell exactlythe outer world and finally obtain the accurate information of their status. Therefore,for physical robot, this paper introduces a linear regress model to predict the conditionof being turnover effectively, which improves the precision and accuracy of robot statedetection. On this basis, we present a balance control method based on the orbital en-ergy model to achieve stand-restoration after turnover, using both attitude control and“taking a step”strategy to achieve balance control. Here we use an of?ine supervisedlearning method to optimize the theoretic foothold model. In the experiment of AIBOrobots, this method reduces the frequency of turnover by more than half, accordinglyproving its validity and feasibility.
3. The motion control system of quadruped walking robot
Quadruped walking robot needs to achieve fast lineal locomotion, well turning andemergency braking; when the robot loses its balance in emergency, it needs to determinewhether the body will be in the danger of turnover, then could recover itself automati-cally; when the robot turns over unfortunately, it needs to achieve fast stand-restorationafter turnover; when it becomes a working platform, it needs to carry out various taskssuch as griping objects through controlling whole-body movement. Therefore, a com-plete motion control system is designed based on the actual application requirementsof quadrupedal walking robot. Each layer has clearly defined duties and specific re-quirements, and also mutually related and restraint, which will effectively coordinateall kinds of motion behaviors, and satisfy various application requirements.Summary and some open problems that require further investigation are given inChapter 6.
1.四足机器人的步态规划及优化
四足机器人最常用的步态规划方法是基于运动学模型的步态控制方法。实验表明,使用该方法规划的轨迹与实际运动轨迹不相符,这一问题造成机器人无法准确按照预先设计的方式运动。因此,本文通过对机器人行走时身体摇摆现象的研究,在运动学模型控制法的基础上引入动力学规划,提出一种使用零力矩点轨迹规划的步态控制方法:控制机器人运动时的身体姿态来改变机器人足部与地面形成的支撑多边形,同时使用零力矩点轨迹规划方法生成机器人在支撑相时足部的轨迹。该方法有效地使用了基于模型的步态控制方法,降低了采用完全动力学规划方法的计算复杂度,满足了算法实时性的要求。设计的控制器具有大量的参数,因而本文使用基于遗传算法的进化学习方法对步态控制参数进行优化。由于有效地使用了基于运动学规划的步态参数进行种群初始化,算法收敛较快。此外,在适应度函数的选取上充分考虑了稳定性这一因素,因而大大提高了最优步态的稳定性。实验使用AIBO机器人进行测试,机器人使用该进化学习方法可自主地得到最优步态,其最优步态在保证稳定性的基础上速度达到了455 mm/s。实验结果表明,应用该方法进行步态控制,机器人获取的最优步态不仅满足稳定性要求,而且对不平地形也具有较好的适应能力。
2.四足机器人的平衡控制及优化
针对机器人在正常运动过程中突然遭遇外力作用引发翻倒危险的问题,各国科学家们致力于提出各种“参考点”作为步行机器人的稳定性判定标准。诸多成果都是使用仿真机器人进行模拟实验,证明了相关理论的有效性,而在实体机器人上,机器人自身传感器不可避免地存在各种噪声和误差,导致机器人无法确切地感知外部世界并得到自身状态的准确信息,因此本文针对实体机器人引入回归映射模型有效地预测翻倒状态,提高了机器人状态检测的精度与准确度。在此基础上,为了迅速恢复机器人的站立状态,本文提出一种基于轨道能量模型的步行机器人平衡控制方法:采用姿态控制和“迈单步策略”两种控制策略实现机器人的平衡控制。同时,采用离线监督学习方法对理论落脚点模型进行优化。实验使用AIBO机器人进行测试,机器人实际的翻倒次数减少了一半以上,证明了该方法的有效性和可行性。
3.四足机器人运动控制系统
四足机器人需要能够实现快速的直线行走、灵活的转弯和紧急制动;当机器人遇到突发状况身体失去平衡时,需要判断出身体是否处于将要翻倒的危险状态,并主动恢复身体平衡;当机器人不幸翻倒后,需要实现翻倒后的快速站立恢复;当机器人成为一个操作平台时,需要通过控制其全身的运动,完成操作物体等各项任务。因此,本文综合考虑四足机器人实际的应用需求,设计了一整套运动控制系统,各层之间任务明确,要求具体,相互关联,相互制约,能够有效地协调四足机器人的各种运动行为,满足多变的应用需求。
第六章给出结论和一些值得进一步研究的问题。
Quadruped walking robot is an important branch of robot. Quadruped walkingrobot not only has heave loading capacity, but also adapts to the all-terrain variety moreeasily. It could not only move slowly and smoothly with statically stable gait, but alsorun with dynamically stable gait. If installed with a manipulator, it will become a sta-ble and flexible working platform while standing. Therefore, quadruped walking robotcould be one of the first practical physical types. On the study of quadruped walk-ing robot, gait planning and balance control of quadruped robot is the subject withgreat significance. In order to improve speed of travel and energy efficiency effec-tively, and enhance the ability of adapting the rough terrain during walking, we takethe quadrupedal walking robot AIBO as research platform and study motion control ofquadrupedal legged robot in the following three aspects:
1. The gait planning and optimization of quadruped robotThe most common method of gait planning of quadruped robots is the controlmethod based on the kinematic model. Experimental results show that the planning tra-jectory using this method is not identical with the actual trajectory of locomotion. Thisproblem causes that the robot can not move following the preliminary mode designedexactly. Therefore, this paper presents a gait control method using ZMP (zero-momentpoint) trajectory planning by studying the body sway phenomenon of robots duringwalking, which introduces the dynamic planning based on the the control method ofthe kinematic model: control the body posture during walking to change the supportpolygon formed by the robot foot and the ground, and in the mean time use ZMPtrajectory planning to generate the foot trajectory during stance phase. This methoduses the model-based gait control method effectively, thus reduces the computationalcomplexity of totally using dynamic planning method, and satisfies the need of real-time algorithm. Since the designed controller contains a large number of parameters,we also use an evolutionary learning method based on genetic algorithm to optimizethose gait control parameters. Since we effectively use the gait parameters based on thekinematic planning to initialize generation, the algorithm constringency is much better.Moreover, we take full account of the stability factor when choosing the fitness func-tion, the stability of the optimal gait is improved. In the experiments, the AIBO robotcan obtain the optimal gait autonomously, and its maximum speed achieves 455 mm/s with stability. The experimental results show that the robots can obtain the optimal gaitwith better stability and stronger adaptability to uneven terrain by using the proposedgait control method.
2. The balance control and optimization of quadruped robot
Concentrated on the danger of turnover imposed by sudden external shocks ofrobots during normal walking, scientists in various countries devote their lives to propos-ing different”reference point”which could be used as the stability criteria of walkingrobot. Most results are carried out with simulation experiments using simulated robots,which can prove the validity of related theory. However, for the entity robot, due to allkinds of errors and noises caused by sensors themselves, the robots can’t tell exactlythe outer world and finally obtain the accurate information of their status. Therefore,for physical robot, this paper introduces a linear regress model to predict the conditionof being turnover effectively, which improves the precision and accuracy of robot statedetection. On this basis, we present a balance control method based on the orbital en-ergy model to achieve stand-restoration after turnover, using both attitude control and“taking a step”strategy to achieve balance control. Here we use an of?ine supervisedlearning method to optimize the theoretic foothold model. In the experiment of AIBOrobots, this method reduces the frequency of turnover by more than half, accordinglyproving its validity and feasibility.
3. The motion control system of quadruped walking robot
Quadruped walking robot needs to achieve fast lineal locomotion, well turning andemergency braking; when the robot loses its balance in emergency, it needs to determinewhether the body will be in the danger of turnover, then could recover itself automati-cally; when the robot turns over unfortunately, it needs to achieve fast stand-restorationafter turnover; when it becomes a working platform, it needs to carry out various taskssuch as griping objects through controlling whole-body movement. Therefore, a com-plete motion control system is designed based on the actual application requirementsof quadrupedal walking robot. Each layer has clearly defined duties and specific re-quirements, and also mutually related and restraint, which will effectively coordinateall kinds of motion behaviors, and satisfy various application requirements.Summary and some open problems that require further investigation are given inChapter 6.
引文
[1]日本机器人学会.机器人技术手册[M].北京:科学出版社, 1996.
[2]蒋新松.未来机器人技术发展方向的探讨[J].机器人, 1996, 18(5): 285-291.
[3]陈学东,孙翊,贾文川.多足步行机器人运动规划与控制[M].武汉:华中科技大学出版社,2006.
[4] Raibert Marc H. Legged robots that balance[J]. Communications of the ACM, 1986, 29(6):499-514.
[5] Raibert Mh, Brown Hb, Murthy Ss. 3D balance using 2D algorithms[M]. International Sym-posiun of Robotics Research. MIT Press. 1984: 279-301.
[6] Raibert M. H., Brown H. B., Chepponis M., et al. Dynamically stable legged locomotion[EB/OL]. 1989. http://hdl.handle.net/1721.1/6820.
[7] Raibert Mh, Brown Jr Hb, Chepponis M. Experiments in balance with a 3D one-legged hop-ping machine[J]. The International Journal of Robotics Research, 1984, 3(2): 75.
[8] Pratt Ga, Williamson Mm. Series elastic actuators[M]. IEEE/RSJ International Conference onIntelligent Robots and Systems. Pittsburgh, PA: IEEE. 1995.
[9] Pratt G, Williamson M, Dillworth P, et al. Stiffness isn’t everything[M]. Proceedings ofISER’95. Stanford CA: Springer-Verlag. 1995: 253-262.
[10] Pratt Jerry, Dilworth Peter, Pratt Gill. Virtual model control of a biped walking robot[M].International Conference on Robotics and Automation. Albuquerque, New Mexico: IEEE.1997.
[11] Pratt J, Torres A, Dilworth P, et al. Virtual actuator control[M]. Intelligent Robots and Systems.IEEE. 1996.
[12] Pratt Jerry, Chew Chee-Meng, Torres Ann, et al. Virtual model control: An intuitive approachfor bipedal locomotion[J]. The International Journal of Robotics Research, 2001, 20(2): 129-143.
[13] Chestnutt J., Lau M., Cheung G., et al. Footstep planning for the Honda ASIMO hu-manoid[M]. Proceedings of the IEEE International Conference on Robotics and Automation.IEEE. 2005: 629-634.
[14] Sakagami Y., Watanabe R., Aoyama C., et al. The intelligent ASIMO: System overview andintegration[M]. Intelligent Robots and Systems. IEEE. 2002.
[15] Hirose R., Takenaka T. Development of the humanoid robot ASIMO[R], 2001.
[16] Fujita M. AIBO: Toward the era of digital creatures[J]. International Journal of Robotics Re-search, 2001, 20(10): 781-791.
[17] Pransky J. AIBO-the No. 1 selling service robot[J]. Industrial Robot, 2001, 28(1): 24-26.
[18] Hirose S, Arikawa K. Development of quadruped walking robot TITAN-VIII for commerciallyavailable research platform[J]. Journal of the Robotics Society of Japan, 1999, 17(8): 1191-1197.
[19] Kato K, Hirose S. Development of the quadruped walking robot, TITAN-IX mechanical designconcept and application for the humanitarian de-mining robot[J]. Advanced Robotics, 2001,15(2): 191-204.
[20] Fukuoka Y, Kimura H, Cohen Ah. Adaptive dynamic walking of a quadruped robot on irreg-ular terrain based on biological concepts[J]. The International Journal of Robotics Research,2003, 22(3-4): 187-202.
[21] Buehler M., Playter R., Raibert M. Robots step outside[A]. International symposium on Adap-tive Motion of Animals and Machines[C] IImenau, Germany: EPFL, 2005.
[22] Raibert M., Blankespoor K., Nelson G., et al. Bigdog, the rough-terrain quaduped robot[J].International Federation of Automation Control, 2008.
[23] Altendorfer R., Moore N., Komsuoglu H., et al. Rhex: A biologically inspired hexapod run-ner[J]. Autonomous Robots, 2001, 11(3): 207-213.
[24] Saranli U, Buehler M, Koditschek De. Rhex: A simple and highly mobile hexapod robot[J].The International Journal of Robotics Research, 2001, 20(7): 616.
[25]梶田秀司.仿人机器人[M].北京:清华大学出版社, 2007.
[26] Kuffner James, Nishiwaki Koichi, Kagami Satoshi, et al. Motion planning for humanoidrobots[J]. Transactions in Advanced Robotics, 2005, 15: 365-374.
[27] Latombe Jean-Claude. Robot motion planning[M]. Boston: Kluwer Academic Publishers,1991.
[28] Chevallereau C., Aoustin Y. Optimal reference trajectories for walking and running of a bipedrobot[J]. Robotica, 2001, 19: 557-569.
[29] Huang Qiang, Yokoi Kazuhito, Kajita Shuuji, et al. Planning walking patterns for a bipedrobot[J]. IEEE Transactions on Robotics and Automation 2001, 17: 280-289.
[30] Kajita Shuuji, Yamaura Tomio, Kobayashi Akira. Dynamic walking control of a biped robotalong a potential energy conserving orbit[J]. IEEE Transactions on Automation, 1992, 8(4):84-101.
[31] Park Jong H., Rhee Yong K. ZMP trajectory generation for reduced trunk motions of bipedrobots[M]. IEEE/ RSJ International Conference on Intelligent Robots and Systems. IEEE.1998: 90-95.
[32] Bay John S., Hemami Hooshang. Modeling of a neural pattern generator with coupled nonlin-ear oscillators[J]. IEEE Transactions on Biomedical Engineering, 1987, 34(4): 297-306.
[33] Kun Andrew L., W. Thomas Miller Ill. Control of variable-speed gaits for a biped robot[J].IEEE Robotics & Automation Magazine, 1999, 6(3): 19-29.
[34] Matsuoka Kiyotoshi. Mechanisms of frequency and pattern control in the neural rhythm gen-erators[J]. Biological Cybernetics, 1987, 413(56): 345-353.
[35] Ogino M, Katoh Y, Aono M, et al. Reinforcement learning of humanoid rhythmic walkingparameters based on visual information[J]. Advanced Robotics, 2004, 18(7): 677-697.
[36] Taga G., Yamaguchi Y., Shimizu H. Self-organized control of bipedal locomotion by neuraloscillators in unpredictable environment[J]. Biological Cybernetics, 1991, 65: 147-159.
[37] Zhen Yuanfang. Reinforcement learning for a biped robot to climb sloping surface[J]. Journalof Robotics System, 1997, 14(4): 283-296.
[38] Kajita S. , Kanehiro F. , Kaneko K. , et al. A realtime pattern generator for biped walking[M].IEEE international Conference on Robotics and Automation. Washington, D.C.: IEEE Com-puter Society Press. 2002: 31-37.
[39] Lohmeier S. , Lof?er K. , Gienger M. , et al. Computer system and control of biped“John-nie”[M]. IEEE International Conference on Robotics and Automation. New Orleans, LA,USA: IEEE. 2004: 4222-4227.
[40] Mcgeer Tad. Passive dynamic walking[J]. The International Journal of Robotics Research,1990, 9(2): 62-82.
[41] Park Jong Hyeon. Fuzzy-logic zero-moment-point trajectory generation for reduced trunk mo-tions of biped robots[J]. Journal of Fuzzy Sets and Systems 2003, 134(1): 189-203.
[42] Salatian A.W. , Zheng Y.F. . Gait synthesis for a biped robot climbing sloping surfaces usingneural networks part II: Dynamic learning[M]. IEEE International Conference on Roboticsand Automation. IEEE. 1992: 2601-2606.
[43] Capi Genci, Nasu Yasuo, Barolli Leonard, et al. Application of genetic algorithms for bipedrobot gait synthesis optimization during walking and going up-stairs[J]. Advanced Robotics,2001, 15(6): 675-694.
[44] Cheng M.-Y., Lin C.-S. Genetic algorithm for control design of biped locomotion[J]. Journalof Robotic System, 1997, 14(5): 365-373.
[45] Russell S., Norvig P.人工智能——一种现代方法[M].北京:人民邮电出版社, 2004.
[46] Siegwart R., Nourbakhsh I. R.自主移动机器人导论[M].西安:西安交通大学出版社,2006.
[47] Mcghee Rb, Frank Aa. On the stability properties of quadruped creeping gaits[J]. Mathemati-cal Biosciences, 1968, 3(3-4): 331-351.
[48] Yoneda K, Hirose S. Tumble stability criterion of integrated locomotion andmanipulation[M].Intelligent Robots and Systems. IEEE. 1996.
[49] Hirose S, Tsukagoshi H, Yoneda K. Static stability criterion for walking robots on irregularterrains[J]. Journal of the Robotics Society of Japan, 1998, 16: 48-54.
[50] Papadopoulos Eg, Rey Da. A new measure of tipover stability margin for mobile manipula-tors[M]. IEEE International Conference on Robotics and Automation. IEEE. 1996.
[51] Hirose S, Tsukagoshi H, Yoneda K. Normalized energy stability margin and its contour ofwalking vehicles on rough terrain[M]. IEEE International Conference on Robotics and Au-tomation. Seoul, Korea: IEEE. 2001: 181-186.
[52] Ghasempoor A., Sepehri N. A measure of machine stability for moving base manipulators[M].IEEE International Conference on Robotics and Automation. Nagoya-shi, Japan: IEEE. 1995.
[53] Kajita S, Espiau B. Springer handbook of robotics[M]. Springer, 2008.
[54] Vukobratovic M, Borovac B. Zero-moment point: thirty five years of its life[J]. InternationalJournal of Humanoid Robotics, 2004, 1(1): 157-173.
[55] Vukobratovic M, Borovac B, Surdilovic D. Zero moment point proper interpretation and newapplication[M]. IEEE/ RAS International Conference on Humanoid Robots. IEEE. 2001: 237-244.
[56] Maki Be, Mcilroy We. The role of limb movements in maintaining upright stance: the“change-in-support”strategy[J]. Physical Therapy, 1997, 77(5): 488.
[57] Atkeson Christopher G., Stephens B. Multiple balance strategies from one optimization crite-rion[A]. IEEE-RAS International Conference on Humanoid Robots[C] Daejeon, Korea: IEEE,2007.
[58] Pratt J., Carff J., Drakunov S., et al. Capture point: A step toward humanoid push recovery[M].IEEE-RAS International Conference on Humanoid Robots. IEEE. 2006: 200–207.
[59] Kudoh S, Komura T, Ikeuchi K. Stepping motion for a humanlike character to maintain bal-ance against large perturbations[M]. IEEE International Conference Robotics and Automation.Orlando, Florida, USA: IEEE. 2006.
[60] Yin K., Pai D. K., Van De Panne M. Data-driven interactive balancing behaviors[M]. PacificConference on Computer Graphics and Applications. Macao, China: Springer. 2005: 118-121.
[61] Chestnutt Joel, Kuffner James, Nishiwaki Koichi, et al. Planning biped navigation strategies incomplex environments[M]. IEEE International Conference on Humanoid Robotics. Munich,Germany: IEEE. 2003: 322-329.
[62] Jean Fre′de′ric. Complexity of nonholonomic motion planning[J]. International Journal of Con-trol, 2001, 74(8): 776-782.
[63] Lavalle Steven M. Planning algorithms[M]. Chicago: Cambridge University Press, 2004.
[64] Mackworth Ak, Vancouver Bc. On seeing robots[J]. Mind Readings: Introductory Selectionson Cognitive Science, 1998: 271.
[65] Burkhard Hd, Duhaut D, Fujita M, et al. The road to RoboCup 2050[J]. IEEE Robotics &Automation Magazine, 2002, 9(2): 31-38.
[66] Kitano H, Asada M, Kuniyoshi Y, et al. Robocup: A challenge problem for ai and robotics[M].RoboCup-97: Robot Soccer World Cup I. Springer. 1998: 1-19.
[67] Kitano H, Asada M, Noda I, et al. RoboCup: Robot world cup[J]. IEEE Robotics & Automa-tion Magazine, 1998, 5(3): 30-36.
[68]多智能体系统实验室. http://ai.ustc.edu.cn/[EB/OL].
[69] Hengst B., Ibbotson D., Son Bao Pham, et al. Omnidirectional locomotion for quadrupedrobots[M]. Lecture Notes in Artificial Intelligence. Seattle, WA, USA: Springer-Verlag. 2002:368-373.
[70] Kohl Nate, Stone Peter. Policy gradient reinforcement learning for fast quadrupedal loco-motion[M]. IEEE International Conference on Robotics and Automation. New Orleans, LA,USA: IEEE. 2004: 2619-2624.
[71] Kohl Nate, Stone Peter. Machine learning for fast quadrupedal locomotion[M]. National Con-ference on Artificial Intelligence. San Francisco, CA, USA: AAAI Press. 2004: 611-616.
[72] Du¨ffert Uwe, Hoffmann Jan. Reliable and precise gait modeling for a quadruped robot[A].Proceedings of RoboCup Symposium[C] Osaka, Japan: Springer-Verlag, 2005.
[73] Chalup Stephan K., Murch Craig L., Quinlan Michael J. Machine learning with AIBO robotsin the four-legged league of RoboCup[J]. IEEE Transactions on Systems, Man and Cybernet-ics, 2006, 37(3): 297-310.
[74] Hornby Gregory S., Takamura Seichi, Yamamoto Takashi, et al. Autonomous evolution ofdynamic gaits with two quadruped robots[J]. IEEE Transactions on Robotics, 2005, 21(3):402-410.
[75] Stronger Daniel, Stone Peter. Simultaneous calibration of action and sensor models on a mo-bile robot[M]. IEEE International Conference on Robotics and Automation. Barcelona, Spain:IEEE. 2005: 4563-4568.
[76] Stronger Daniel, Stone Peter. Towards autonomous sensor and actuator model induction on amobile robot[J]. Connection Science (Special Issue on Developmental Robotics), 2006, 18(2):97-119.
[77] He Haitao, Chen Xiaoping. A model-based approach to calculating and calibrating the odom-etry for quadruped robots[M]. Proceedings of RoboCup Symposium. Atlanta, USA: Springer-Verlag. 2007.
[78] Rofer T, Brunn R, Dahm I, et al. GermanTeam 2004-the german national Robocup team[J].RoboCup 2004: Robot Soccer World Cup VIII, 2005, 3276.
[79] Hildebrand M. Symmetrical gaits of horses[J]. Science, 1965, 150(3697): 701-708.
[80]蔡自兴.智能控制[M].北京:电子工业出版社, 2004.
[81]蔡自兴,周翔.基于功能/行为集成的自主式移动机器人进化控制体系结构[J].机器人,2000, 22(3): 169-175.
[82]张楫,赵明国,董浩.基于圆弧模型的四足机器人步态规划[J].机器人, 2006, 28(5):536-539.
[83]许涛,陈启军.四腿机器人步态参数自动进化研究与实现[J].机器人, 2009, 31(1): 72-77.
[84]徐凯,陈小平.一种有腿机器人步态轨迹生成算法[J].小型微型计算机系统, 2008, 29(5):854-858.
[85] Hirose S., Kan Yoneda, Furuya R., et al. Dynamic and static fusion control of quadrupedwalking vehicle[M]. Intelligent Robots and Systems. Tsukuba, Japan: IEEE. 1989: 199-204.
[86] Kurazume R., Yoneda K., Hirose S. Feedforward and feedback dynamic trot gait control forquadruped walking vehicle[J]. Autonomous Robots, 2002, 12(2): 157-172.
[87] Liu Jinsu, Veloso Manuela. Online ZMP sampling search for biped walking planning[M].IEEE/RSJ International Conference on Intelligent Robots and Systems. Nice, France: IEEE.2008: 185-190.
[88]陈国良,王东生,王煦法, et al.遗传算法及其应用[M].北京:人民邮电出版社, 1996.
[89]韩丹夫,吴庆标.数值计算方法[M].杭州:浙江大学出版社, 2006.
[90] Futami S, Kyura N, Hara S. Vibration absorption control of industrial robots by accelerationfeedback[J]. IEEE Transactions on Industrial Electronics, 1983: 299-305.
[91]常西畅,赵力行.频谱分析仪及其在故障诊断中的应用[M].北京:中国宇航出版社, 2006.
[92] Kajita S., Kanehiro F., Kaneko K., et al. Biped walking pattern generation by using previewcontrol of zero-moment point[M]. IEEE International Conference on Robotics and Automa-tion. Taipei, Taiwan: IEEE. 2003: 1620-1626.
[93] Kajita S., Kanehiro F., Kaneko K., et al. The 3D linear inverted pendulum mode: a simplemodeling for a bipedwalking pattern generation[M]. IEEE/RSJ International Conference onIntelligent Robots and Systems. Maui, Hawaii, USA: IEEE. 2001: 239-246.
[94] Hofmann Ag. Robust execution of bipedal walking tasks from biomechanical principles[D].Cambridge, Massachusetts: Massachusetts Institute of Technology, 2005.
[95] Rebula J, Canas F, Pratt J, et al. Learning capture points for humanoid push recovery[M].IEEE International Conference on Robotics and Automation. IEEE. 2007: 1774-1781.
[96]周纪芗.回归分析[M].上海:华东师范大学出版社, 1993.
[2]蒋新松.未来机器人技术发展方向的探讨[J].机器人, 1996, 18(5): 285-291.
[3]陈学东,孙翊,贾文川.多足步行机器人运动规划与控制[M].武汉:华中科技大学出版社,2006.
[4] Raibert Marc H. Legged robots that balance[J]. Communications of the ACM, 1986, 29(6):499-514.
[5] Raibert Mh, Brown Hb, Murthy Ss. 3D balance using 2D algorithms[M]. International Sym-posiun of Robotics Research. MIT Press. 1984: 279-301.
[6] Raibert M. H., Brown H. B., Chepponis M., et al. Dynamically stable legged locomotion[EB/OL]. 1989. http://hdl.handle.net/1721.1/6820.
[7] Raibert Mh, Brown Jr Hb, Chepponis M. Experiments in balance with a 3D one-legged hop-ping machine[J]. The International Journal of Robotics Research, 1984, 3(2): 75.
[8] Pratt Ga, Williamson Mm. Series elastic actuators[M]. IEEE/RSJ International Conference onIntelligent Robots and Systems. Pittsburgh, PA: IEEE. 1995.
[9] Pratt G, Williamson M, Dillworth P, et al. Stiffness isn’t everything[M]. Proceedings ofISER’95. Stanford CA: Springer-Verlag. 1995: 253-262.
[10] Pratt Jerry, Dilworth Peter, Pratt Gill. Virtual model control of a biped walking robot[M].International Conference on Robotics and Automation. Albuquerque, New Mexico: IEEE.1997.
[11] Pratt J, Torres A, Dilworth P, et al. Virtual actuator control[M]. Intelligent Robots and Systems.IEEE. 1996.
[12] Pratt Jerry, Chew Chee-Meng, Torres Ann, et al. Virtual model control: An intuitive approachfor bipedal locomotion[J]. The International Journal of Robotics Research, 2001, 20(2): 129-143.
[13] Chestnutt J., Lau M., Cheung G., et al. Footstep planning for the Honda ASIMO hu-manoid[M]. Proceedings of the IEEE International Conference on Robotics and Automation.IEEE. 2005: 629-634.
[14] Sakagami Y., Watanabe R., Aoyama C., et al. The intelligent ASIMO: System overview andintegration[M]. Intelligent Robots and Systems. IEEE. 2002.
[15] Hirose R., Takenaka T. Development of the humanoid robot ASIMO[R], 2001.
[16] Fujita M. AIBO: Toward the era of digital creatures[J]. International Journal of Robotics Re-search, 2001, 20(10): 781-791.
[17] Pransky J. AIBO-the No. 1 selling service robot[J]. Industrial Robot, 2001, 28(1): 24-26.
[18] Hirose S, Arikawa K. Development of quadruped walking robot TITAN-VIII for commerciallyavailable research platform[J]. Journal of the Robotics Society of Japan, 1999, 17(8): 1191-1197.
[19] Kato K, Hirose S. Development of the quadruped walking robot, TITAN-IX mechanical designconcept and application for the humanitarian de-mining robot[J]. Advanced Robotics, 2001,15(2): 191-204.
[20] Fukuoka Y, Kimura H, Cohen Ah. Adaptive dynamic walking of a quadruped robot on irreg-ular terrain based on biological concepts[J]. The International Journal of Robotics Research,2003, 22(3-4): 187-202.
[21] Buehler M., Playter R., Raibert M. Robots step outside[A]. International symposium on Adap-tive Motion of Animals and Machines[C] IImenau, Germany: EPFL, 2005.
[22] Raibert M., Blankespoor K., Nelson G., et al. Bigdog, the rough-terrain quaduped robot[J].International Federation of Automation Control, 2008.
[23] Altendorfer R., Moore N., Komsuoglu H., et al. Rhex: A biologically inspired hexapod run-ner[J]. Autonomous Robots, 2001, 11(3): 207-213.
[24] Saranli U, Buehler M, Koditschek De. Rhex: A simple and highly mobile hexapod robot[J].The International Journal of Robotics Research, 2001, 20(7): 616.
[25]梶田秀司.仿人机器人[M].北京:清华大学出版社, 2007.
[26] Kuffner James, Nishiwaki Koichi, Kagami Satoshi, et al. Motion planning for humanoidrobots[J]. Transactions in Advanced Robotics, 2005, 15: 365-374.
[27] Latombe Jean-Claude. Robot motion planning[M]. Boston: Kluwer Academic Publishers,1991.
[28] Chevallereau C., Aoustin Y. Optimal reference trajectories for walking and running of a bipedrobot[J]. Robotica, 2001, 19: 557-569.
[29] Huang Qiang, Yokoi Kazuhito, Kajita Shuuji, et al. Planning walking patterns for a bipedrobot[J]. IEEE Transactions on Robotics and Automation 2001, 17: 280-289.
[30] Kajita Shuuji, Yamaura Tomio, Kobayashi Akira. Dynamic walking control of a biped robotalong a potential energy conserving orbit[J]. IEEE Transactions on Automation, 1992, 8(4):84-101.
[31] Park Jong H., Rhee Yong K. ZMP trajectory generation for reduced trunk motions of bipedrobots[M]. IEEE/ RSJ International Conference on Intelligent Robots and Systems. IEEE.1998: 90-95.
[32] Bay John S., Hemami Hooshang. Modeling of a neural pattern generator with coupled nonlin-ear oscillators[J]. IEEE Transactions on Biomedical Engineering, 1987, 34(4): 297-306.
[33] Kun Andrew L., W. Thomas Miller Ill. Control of variable-speed gaits for a biped robot[J].IEEE Robotics & Automation Magazine, 1999, 6(3): 19-29.
[34] Matsuoka Kiyotoshi. Mechanisms of frequency and pattern control in the neural rhythm gen-erators[J]. Biological Cybernetics, 1987, 413(56): 345-353.
[35] Ogino M, Katoh Y, Aono M, et al. Reinforcement learning of humanoid rhythmic walkingparameters based on visual information[J]. Advanced Robotics, 2004, 18(7): 677-697.
[36] Taga G., Yamaguchi Y., Shimizu H. Self-organized control of bipedal locomotion by neuraloscillators in unpredictable environment[J]. Biological Cybernetics, 1991, 65: 147-159.
[37] Zhen Yuanfang. Reinforcement learning for a biped robot to climb sloping surface[J]. Journalof Robotics System, 1997, 14(4): 283-296.
[38] Kajita S. , Kanehiro F. , Kaneko K. , et al. A realtime pattern generator for biped walking[M].IEEE international Conference on Robotics and Automation. Washington, D.C.: IEEE Com-puter Society Press. 2002: 31-37.
[39] Lohmeier S. , Lof?er K. , Gienger M. , et al. Computer system and control of biped“John-nie”[M]. IEEE International Conference on Robotics and Automation. New Orleans, LA,USA: IEEE. 2004: 4222-4227.
[40] Mcgeer Tad. Passive dynamic walking[J]. The International Journal of Robotics Research,1990, 9(2): 62-82.
[41] Park Jong Hyeon. Fuzzy-logic zero-moment-point trajectory generation for reduced trunk mo-tions of biped robots[J]. Journal of Fuzzy Sets and Systems 2003, 134(1): 189-203.
[42] Salatian A.W. , Zheng Y.F. . Gait synthesis for a biped robot climbing sloping surfaces usingneural networks part II: Dynamic learning[M]. IEEE International Conference on Roboticsand Automation. IEEE. 1992: 2601-2606.
[43] Capi Genci, Nasu Yasuo, Barolli Leonard, et al. Application of genetic algorithms for bipedrobot gait synthesis optimization during walking and going up-stairs[J]. Advanced Robotics,2001, 15(6): 675-694.
[44] Cheng M.-Y., Lin C.-S. Genetic algorithm for control design of biped locomotion[J]. Journalof Robotic System, 1997, 14(5): 365-373.
[45] Russell S., Norvig P.人工智能——一种现代方法[M].北京:人民邮电出版社, 2004.
[46] Siegwart R., Nourbakhsh I. R.自主移动机器人导论[M].西安:西安交通大学出版社,2006.
[47] Mcghee Rb, Frank Aa. On the stability properties of quadruped creeping gaits[J]. Mathemati-cal Biosciences, 1968, 3(3-4): 331-351.
[48] Yoneda K, Hirose S. Tumble stability criterion of integrated locomotion andmanipulation[M].Intelligent Robots and Systems. IEEE. 1996.
[49] Hirose S, Tsukagoshi H, Yoneda K. Static stability criterion for walking robots on irregularterrains[J]. Journal of the Robotics Society of Japan, 1998, 16: 48-54.
[50] Papadopoulos Eg, Rey Da. A new measure of tipover stability margin for mobile manipula-tors[M]. IEEE International Conference on Robotics and Automation. IEEE. 1996.
[51] Hirose S, Tsukagoshi H, Yoneda K. Normalized energy stability margin and its contour ofwalking vehicles on rough terrain[M]. IEEE International Conference on Robotics and Au-tomation. Seoul, Korea: IEEE. 2001: 181-186.
[52] Ghasempoor A., Sepehri N. A measure of machine stability for moving base manipulators[M].IEEE International Conference on Robotics and Automation. Nagoya-shi, Japan: IEEE. 1995.
[53] Kajita S, Espiau B. Springer handbook of robotics[M]. Springer, 2008.
[54] Vukobratovic M, Borovac B. Zero-moment point: thirty five years of its life[J]. InternationalJournal of Humanoid Robotics, 2004, 1(1): 157-173.
[55] Vukobratovic M, Borovac B, Surdilovic D. Zero moment point proper interpretation and newapplication[M]. IEEE/ RAS International Conference on Humanoid Robots. IEEE. 2001: 237-244.
[56] Maki Be, Mcilroy We. The role of limb movements in maintaining upright stance: the“change-in-support”strategy[J]. Physical Therapy, 1997, 77(5): 488.
[57] Atkeson Christopher G., Stephens B. Multiple balance strategies from one optimization crite-rion[A]. IEEE-RAS International Conference on Humanoid Robots[C] Daejeon, Korea: IEEE,2007.
[58] Pratt J., Carff J., Drakunov S., et al. Capture point: A step toward humanoid push recovery[M].IEEE-RAS International Conference on Humanoid Robots. IEEE. 2006: 200–207.
[59] Kudoh S, Komura T, Ikeuchi K. Stepping motion for a humanlike character to maintain bal-ance against large perturbations[M]. IEEE International Conference Robotics and Automation.Orlando, Florida, USA: IEEE. 2006.
[60] Yin K., Pai D. K., Van De Panne M. Data-driven interactive balancing behaviors[M]. PacificConference on Computer Graphics and Applications. Macao, China: Springer. 2005: 118-121.
[61] Chestnutt Joel, Kuffner James, Nishiwaki Koichi, et al. Planning biped navigation strategies incomplex environments[M]. IEEE International Conference on Humanoid Robotics. Munich,Germany: IEEE. 2003: 322-329.
[62] Jean Fre′de′ric. Complexity of nonholonomic motion planning[J]. International Journal of Con-trol, 2001, 74(8): 776-782.
[63] Lavalle Steven M. Planning algorithms[M]. Chicago: Cambridge University Press, 2004.
[64] Mackworth Ak, Vancouver Bc. On seeing robots[J]. Mind Readings: Introductory Selectionson Cognitive Science, 1998: 271.
[65] Burkhard Hd, Duhaut D, Fujita M, et al. The road to RoboCup 2050[J]. IEEE Robotics &Automation Magazine, 2002, 9(2): 31-38.
[66] Kitano H, Asada M, Kuniyoshi Y, et al. Robocup: A challenge problem for ai and robotics[M].RoboCup-97: Robot Soccer World Cup I. Springer. 1998: 1-19.
[67] Kitano H, Asada M, Noda I, et al. RoboCup: Robot world cup[J]. IEEE Robotics & Automa-tion Magazine, 1998, 5(3): 30-36.
[68]多智能体系统实验室. http://ai.ustc.edu.cn/[EB/OL].
[69] Hengst B., Ibbotson D., Son Bao Pham, et al. Omnidirectional locomotion for quadrupedrobots[M]. Lecture Notes in Artificial Intelligence. Seattle, WA, USA: Springer-Verlag. 2002:368-373.
[70] Kohl Nate, Stone Peter. Policy gradient reinforcement learning for fast quadrupedal loco-motion[M]. IEEE International Conference on Robotics and Automation. New Orleans, LA,USA: IEEE. 2004: 2619-2624.
[71] Kohl Nate, Stone Peter. Machine learning for fast quadrupedal locomotion[M]. National Con-ference on Artificial Intelligence. San Francisco, CA, USA: AAAI Press. 2004: 611-616.
[72] Du¨ffert Uwe, Hoffmann Jan. Reliable and precise gait modeling for a quadruped robot[A].Proceedings of RoboCup Symposium[C] Osaka, Japan: Springer-Verlag, 2005.
[73] Chalup Stephan K., Murch Craig L., Quinlan Michael J. Machine learning with AIBO robotsin the four-legged league of RoboCup[J]. IEEE Transactions on Systems, Man and Cybernet-ics, 2006, 37(3): 297-310.
[74] Hornby Gregory S., Takamura Seichi, Yamamoto Takashi, et al. Autonomous evolution ofdynamic gaits with two quadruped robots[J]. IEEE Transactions on Robotics, 2005, 21(3):402-410.
[75] Stronger Daniel, Stone Peter. Simultaneous calibration of action and sensor models on a mo-bile robot[M]. IEEE International Conference on Robotics and Automation. Barcelona, Spain:IEEE. 2005: 4563-4568.
[76] Stronger Daniel, Stone Peter. Towards autonomous sensor and actuator model induction on amobile robot[J]. Connection Science (Special Issue on Developmental Robotics), 2006, 18(2):97-119.
[77] He Haitao, Chen Xiaoping. A model-based approach to calculating and calibrating the odom-etry for quadruped robots[M]. Proceedings of RoboCup Symposium. Atlanta, USA: Springer-Verlag. 2007.
[78] Rofer T, Brunn R, Dahm I, et al. GermanTeam 2004-the german national Robocup team[J].RoboCup 2004: Robot Soccer World Cup VIII, 2005, 3276.
[79] Hildebrand M. Symmetrical gaits of horses[J]. Science, 1965, 150(3697): 701-708.
[80]蔡自兴.智能控制[M].北京:电子工业出版社, 2004.
[81]蔡自兴,周翔.基于功能/行为集成的自主式移动机器人进化控制体系结构[J].机器人,2000, 22(3): 169-175.
[82]张楫,赵明国,董浩.基于圆弧模型的四足机器人步态规划[J].机器人, 2006, 28(5):536-539.
[83]许涛,陈启军.四腿机器人步态参数自动进化研究与实现[J].机器人, 2009, 31(1): 72-77.
[84]徐凯,陈小平.一种有腿机器人步态轨迹生成算法[J].小型微型计算机系统, 2008, 29(5):854-858.
[85] Hirose S., Kan Yoneda, Furuya R., et al. Dynamic and static fusion control of quadrupedwalking vehicle[M]. Intelligent Robots and Systems. Tsukuba, Japan: IEEE. 1989: 199-204.
[86] Kurazume R., Yoneda K., Hirose S. Feedforward and feedback dynamic trot gait control forquadruped walking vehicle[J]. Autonomous Robots, 2002, 12(2): 157-172.
[87] Liu Jinsu, Veloso Manuela. Online ZMP sampling search for biped walking planning[M].IEEE/RSJ International Conference on Intelligent Robots and Systems. Nice, France: IEEE.2008: 185-190.
[88]陈国良,王东生,王煦法, et al.遗传算法及其应用[M].北京:人民邮电出版社, 1996.
[89]韩丹夫,吴庆标.数值计算方法[M].杭州:浙江大学出版社, 2006.
[90] Futami S, Kyura N, Hara S. Vibration absorption control of industrial robots by accelerationfeedback[J]. IEEE Transactions on Industrial Electronics, 1983: 299-305.
[91]常西畅,赵力行.频谱分析仪及其在故障诊断中的应用[M].北京:中国宇航出版社, 2006.
[92] Kajita S., Kanehiro F., Kaneko K., et al. Biped walking pattern generation by using previewcontrol of zero-moment point[M]. IEEE International Conference on Robotics and Automa-tion. Taipei, Taiwan: IEEE. 2003: 1620-1626.
[93] Kajita S., Kanehiro F., Kaneko K., et al. The 3D linear inverted pendulum mode: a simplemodeling for a bipedwalking pattern generation[M]. IEEE/RSJ International Conference onIntelligent Robots and Systems. Maui, Hawaii, USA: IEEE. 2001: 239-246.
[94] Hofmann Ag. Robust execution of bipedal walking tasks from biomechanical principles[D].Cambridge, Massachusetts: Massachusetts Institute of Technology, 2005.
[95] Rebula J, Canas F, Pratt J, et al. Learning capture points for humanoid push recovery[M].IEEE International Conference on Robotics and Automation. IEEE. 2007: 1774-1781.
[96]周纪芗.回归分析[M].上海:华东师范大学出版社, 1993.