CN102378669A - Model-based neuromechanical controller for a robotic leg - Google Patents

Abstract

A model-based neuromechanical controller for a robotic limb having at least one joint includes a finite state machine configured to receive feedback data relating to the state of the robotic limb and to determine the state of the robotic limb, a muscle model processor configured to receive state information from the finite state machine and, using muscle geometry and reflex architecture information and a neuromuscular model, to determine at least one desired joint torque or stiffness command to be sent to the robotic limb, and a joint command processor configured to command the biomimetric torques and stiffnesses determined by the muscle model processor at the robotic limb joint. The feedback data is preferably provided by at least one sensor mounted at each joint of the robotic limb. In a preferred embodiment, the robotic limb is a leg and the finite state machine is synchronized to the leg gait cycle.

Description

The neural mechanical control device that is used for pedipulator based on model

Related application

The application requires U.S. Provisional Application No.61/148,545 priority, and it is filed on January 30th, 2009, and its disclosed full content is herein with reference to quoting.

The application is the U.S. Patent application No.12/157 that is filed on June 12nd, 2008,727 part continuation application, the U.S. Provisional Patent Application No.60/934 that this U.S. Patent application requirement is filed on June 12nd, 2007, has stopped at present; 233 priority, and the U.S. Patent application No.11/395 that lists below being of the application, 448; 11/495,140,11/642; 993 part continuation application, its disclosed full content integral body is included this paper in as a reference.

The application also is the U.S. Patent application No.12/608 that is filed on October 29th, 2009,627 part continuation application, and this U.S. Patent application is the U.S. Provisional Patent Application No.60/751 that is filed on December 19th, 2006, has abandoned at present and required submission on December 19th, 2005, stopped at present; The U.S. Patent application No.11/642 of 680 priority, 993 continuation application, and the U.S. Patent application No.11/395 that lists below being of the application; 448,11/495,140; 11/600,291 and require the U.S. Provisional Patent Application No.60/705 be filed on August 4th, 2005, stopped at present, preferentially having authorized at present for the time being of 651 the applying date is U.S. Pat 7; 313; The part continuation application of 463 application 11/499,853, and the U.S. Patent application No.11/395 that lists below being of the application; 448 part continuation application, its disclosed full content integral body is included this paper in as a reference.

The application still name be called " the artificial human limb and the joint of using actuator, spring and adaptive damping element ", by Hugh M.herr; Daniel Joseph Paluska and Peter Dilworth are filed in the U.S. Patent application No.11/395 on March 31st, 2006,448 part continuation application.U.S. Patent application No.11/395; The U.S. Provisional Patent Application No.60/666 that 448 requirements are filed on March 31st, 2005, have stopped at present; 876 priority and the U.S. Provisional Patent Application No.60/704 that is filed on August 1st, 2005, has stopped at present, 517 priority.

The application still name be called " the artificial ankle-pedal system that has spring, variable damper and series connection resilient actuating element ", by Hugh M.Herr; Samuel K.Au; Peter Dilwort and Daniel Hosph Paluska are filed in the U.S. Patent application No.11/495 on July 29th, 2006,140 part continuation application.U.S. Patent application No.11/495, the U.S. Provisional Patent Application No.60/704 that 140 requirements are filed on August 1st, 2005, have stopped at present, 517 priority, and also be U.S. Patent application No.11/395,448 part continuation application.

The application still name be called " ectoskeleton of running and walking ", by Hugh M.Herr; Conoe Walsh; Daniel Hoseph Paluska; Andrew Valiente, Kenneth Pasch and Eilliam Grand are filed in the U.S. Patent application No.11/600 on November 15th, 2006,291 part continuation application.U.S. Patent application No.11/600, the U.S. Provisional Patent Application No.60/736 that 291 requirements are filed on November 15th, 2005, have stopped at present, 929 priority; And also be U.S. Patent application No.11/395,448,11/499; 853 and 11/495,140 part continuation application.

The application requires the priority of each aforementioned patent applications, and the disclosed full content integral body of aforementioned all patent applications is included this paper in as a reference.

Federal funding research and development statement

The present invention is authorized by Veterans' Administration by the major project exploitation that is numbered VA241-P-0026 that U.S. government supports.Government has specific rights of the present invention.

Technical field

The present invention relates to be used for the artificial joint of reparation, rectification, ectoskeleton or mechanical device (robot device) and the control of limbs, especially for control method based on the pedipulator of neuromuscular type motion model.

Background technology

The motion of animal and human's leg is by the ANN Control of complicacy.Be suggested [Brown, T.G., 1914. essence 20th century in early days about the activity of neural basis; Follow the analysis of intermittently active level conditions, and theoretical (the On the nature of the fundamental activity of the nervous centes of the evolution of nervous function; Together with an analysis of the conditioning of rhythmic activity in progression, and a theory of the evolution offunction in the nervous system) .J Physiol 48 (1), 18-46.]; Strictly set up [Orlovsky, G., Deliagina today; T.; Grillner, S., 1999.Neuronal control of locomotion:from mollusc to man.Oxford University Press; New York], central pattern generator (cpg) (CPG) forms the basis of this network.

In current viewpoint, CPG comprises neuron pool layer [Rybak, I.A., the Shevtsova in the spinal cord; N.A., Lafreniere-Roula, M.; McCrea, D.A., 2006.Modelling spinal circuitry involved in locomotor pattern generation:insights from deletions during fictive locomotion.J Physiol 577 (Pt 2); 617-639], wherein, through the coordination of other neuron pool guiding muscle; Even when lacking spinal reflex [Grillner, S., Zangger; P., 1979.On the central generation of locomotion in the low spinal cat.Exp Brain Res 34 (2), 241-261; Frigon, A., Rossignol; S., 2006.Experiments and models of sensorimotor interactions during locomotion.Biol Cybern 95 (6), 607-627]; The extensor of leg and the movable [Dietz of the rhythm and pace of moving things of musculus flexor are provided; V., 2003.Spinal cord pattern generators for locomotion.Clin Neuroph3rsiol 1 14 (8), 1379-1389; Minassian, K., Persy, I.; Rattay, F., Pinter, M.M.; Kern, H., Dimitrijevic; M.R., 2007.Human lumbar cord circuitries can be activated by extrinsic tonic input to generate locomotor-like activity.Hum Mov Sci 26 (2), 275-295] enough produce ambulatory motion.But, spinal reflex is part [Rybak, the I.A. of this complicated network; Stecina, K., Shevtsova; N.A.; McCrea, D.A., 2006.Modelling spinal circuitry involved in locomotor pattern generation:insights from the effects of afferent stimulation.J Physiol 577 (Pt 2); 641-658], very big to the output contribution of selection motor pattern, the activity time that limits extensor, musculus flexor and modulation CPG.

The combination that use maincenter pattern takes place and modulation is reflected, lamprey [Ekeberg, O., Grillner; S., 1999.Simulations of neuromuscular control in lamprey swimming.Philos Trans R Soc Lond B Biol Sci 354 (1385), 895-902], salamander [Ijspeert; A., Crespi, A., Ryczko; D..Cabelguen, J.-M., 2007.From swimming to walking with a salamander robot driven by a spinal cord model.Science 315 (5817), 1416-1420]; Cat [Ivashko, D.G., Prilutski, B.I.; Markin, S.N., Chapin, J.K.; Rybak, I.A., 2003.Modeling the spinal cord neural circuitry controlling cat hindlimb movement during locomotion.Neurocomputing 52-54,621-629; Yakovenko, S., Gritsenko, V., Prochazka, A., 2004.Contribution of stretch reflexes to locomotor control:a modeling study.Biol Cybern 90 (2), 146-155; Maufroy, C., Kimura, H.; Takase, K., 2008.Towards a general neural controller for quadrupedal locomotion.Neural Netw 21 (4), 667-681]; And people [Ogihara, N., Yamazaki; N., 2001.Generation of human bipedal locomotion by a bio-mimetic neuro-musculo-skeletal model.Biol Cybera 84 (1), 1-11:Paul; C., Bellotti, M.; Jezeraik, S., Curt; A., 2005.Development of a human neuro-musculo-skeletal model for investigation of spinal cord injury.Biol Cybern 93 (3), 153-170] the neuromuscular model developed basic instrument and be used for studying different control strategies in animal and human's motion.The emphasis of these models is to reproduce the architecture of CPG and through test potential reflection [Pearson is provided; K., Ekeberg, O.; Buschges; A., 2006.Assessing sensory function in locomotor systems using neuro.mechanical simulations.Trends Neurosci 29 (11), 625-631].Yet, seldom have notice to spend in how the such architecture of understanding is reproduced or the principle of the sport dynamics structure of encoding.

These principles disclose, and compare with the identification of the neuroid of complicacy, and leg motion needs control seldom or do not need to control.For example, walking [Alexander, R., 1976.Mechanics of bipe:dal locomotion.In:Perspectives in experimental biology (Ed.Davies, P.S.) Pergamon.Oxford; Mochon, S., McMahon; T., 1980.Ballistic walking.J.Biomech.13 (1), 49-57] and [Blickhan that runs; R., 1989.The spring-mass model for running and hopping.J.of Biomech.22,1217-1227; McMahon, T., Cheng; G.; 1990.The mechanism of running:how does stiffness couple with speed? J.of Biomech.23,65-78] two conception of species models are suggested, and are used for catching the dominant mechanism of leg motion.If dynamic system is regulated the self-stabilization ability of these models [McGeer, T., 1990.Passive dynamic walking.Int.J.Rob.Res.9 (2), 62-82 suitably; McGeer, T., 1992.Principles of walking and running.Vol.11 of Advances in Comparative and Environmental Physiology.Springer-Verlag Berlin Heidelberg, Ch.4; Seyfarth, A., Geyer, H., G ü nther, M., Blickhan, R., 2002.A movement criterion for running.J.of Biomech.35,649-655; Ghigliazza, R., Altendorfer, R., Holmes, P., Koditschek, D., 2003.A simply stabilized running model.SIAM J.Applied.Dynamical Systems 2 (2), 187-218] be unfolded research.And use the walking of this principle and the robot that runs verified its actual correlation and control action [Raibert, M., 1986.Legged robots that balance.MIT press, Cambridge; McGeer, T., 1990.Passive dynamic walking.Int.J.Rob.Res.9 (2); 62-82:Saranli, U., Buehler; M., Koditschek, D.; 2001.Rhex:A simple and highly mobile hexapod robot.Int.Jour.Rob.Res.20 (7), 616-631; Collins, S., Ruina, A., Tedrake, R., Wisse, M., 2005.Efficient bipedal robots based on passive-dynamic walkers.Science 307 (5712), 1082-1085].But, still there is an an open question, promptly how should be integrated in the middle of the human motion flesh control system with other principles what leg strength was learned.

Interactional importance between dynamics and the motor control is recognized [Pearson, K., Ekeberg by neuroscientist and biomechanists; O., Buschges, A.; 2006.Assessing sensory function in locomotor systems using neuro-mechanical simulations.Trends Neurosci 29 (11); 625-631]. for example, the movable maincenter driving of motor is widely accepted [Grillner, S. in the motion although CPG forms; Zangger; P., 1979.On the central generation of locomotion in the low spinal cat.Exp Brain Res 34 (2), 241-261; Dietz, V., 2003.Spinal cord pattern generators for locomotion.Clin Neurophy ' siol 1 14 (8), 1379-1389; Frigon, A., Rossignol, S., 2006.Experiments and models of sensorimotor interactions during locomotion.Biol Cybern 95 (6), 607-627; Ijspeert, A.J., 2008.Central pattern generators for locomotion control in animals and robots:a review.Neural Netw 21 (4); 642-653]; Lundberg proposed in 1969, and outside its simple maincenter output, the spinal reflex of propagating the information of learning about motion-promotion force can form the muscle activity [Lundberg of complicacy visible in the actual motion; A.; 1969.Reflex control of stepping.In:The Nansen memorial lecture V, Oslo:Universitetsforlaget, 5-42].Put this thought in order; Taga proposed afterwards; Because " rhythm that the center produces is caused by the transducing signal that the rhythmic movement that causes because of the motor device causes ... the output of [;] motor is the characteristic that the dynamic interaction between nervous system, skeletal muscle system and the natural environment produces " [Taga; G., 1995.A model of the neuro-musculo-skeletal system for human locomotion.I.Emergence of basic gait.Biol.Cybern.73 (2), 97-111].As support, he proposes a kind of neuromuscular model of human motion, and it is compound to have the CPG of sensory feedback, and proved basic gait how the activity of the rhythm and pace of moving things between N&M-skeletal system gets induce appearance comprehensively.

The center and the actual specific reflection input that produces motor output is continuously by contention [Pearson, K.G., 2004.Generating the walking gait:role of sensory feedback.Prog Brain Res 143,123-129; Frigon, A., Rossignol, S., 2006.Experiments and models of sensorimotor interactions during locomotion.Biol Cybem 95 (6), 607-627; Hultbom, H., 2006.Spinal reflexes, mechanisms and concepts:from Eccles to Lundberg and beyond.Prog Neurobiol 78 (3-5), 215-232; Prochazka, A., Yakovenko, S., 2007.The neuromechanical tuning hypothesis.Prog Brain Res 1 65,255-265].For example, the cat of walking be estimated as the leg extensor that has only about 30 percent taking the weight of that be observed muscle activity owing to muscle reflection [Prochazka, A.; Gritsenko; V., Yakovenko, S.; 2002.Sensory control of locomotion:reflexes versus higher-level control.Adv Exp Med Biol 508,357-367; Donelan, J.M., McVea, D.A., Pearson, K.G., 2009.Force regulation of ankle extensor muscle activity in freely walking cats.J Neurophysiol 101 (1), 360-371].

In the mankind, reflection seems more remarkable for the contribution of muscle activity in the motion.Sinkjaer and colleague estimate that through the unloading test reflection is contributed about 50 percent [Sinkjaer, T., Andersen to the activity of musculus soleus in the walking driving phase; J.B., Ladouceur, M.; Christensen, L.O., Nielsen; J.B., 2000.Major role for sensory feedback in soleus EMG activity in the stance phase of walking in man.J Physiol 523 Pt 3,817-827].In the time of closer, Grey and colleague find that the change that musculus soleus is movable and the ratio that changes over of heel string power have disclosed the direct relation [Grey between power positive feedback and this muscle activity; M.J., Nielsen, J.B.; Mazzaro, N., Sinkjaer; T., 2007.Positive force feedback in human walking.J Phyrsiol 581 (1), 99-105].Whether big like this feedback contribution all exists all leg muscles and remains unknown.Perhaps be present in motor control near terminal gradient; Wherein immediate leg muscle mainly by maincenter input control distal legs muscle because high proprioception feedback oscillator and big mechanics influence sensitivity mainly are controlled by the reflection input, as Daley and with thing by the birds exercise test infer [Daley, M.A.; Felix; G., Biewener, A.A.; 2007.Running stability is enhanced by a proximo-distal gradient in joint neuromechanical control.J Exp Biol 210 (Pt 3), 383-394].

Adaptation to the ground is an importance of walking.Commercially available now ankle-sufficient prosthese utilizes passive (passive) structure of lightweight; It is designed in the driving phase of walking, appear suitable elasticity [S.Ron, Prosthetics and Orthotics:Lower Limb and Spinal.Lippincott Williams & Wilkins 2002].The compound that is used in the advanced person in these devices allows some energy of storage in controlled dorsiflex and sufficient sole of the foot BENDING PROCESS; And in the sufficient sole of the foot BENDING PROCESS of subsequently usefulness powered, release energy; More as normal person's heel string [A.L.Hof, B.A.Geelen, Jw.Van den Berg; " Calf muscle moment, work and efficiency in level walking; Role of series elasticity, " Journal of Biomechanics, Vol.16.No.7, pp.523-537,1983; D.A.Winter, " Biomechanical motor pattern in normal walking. " Journal of Motor Behavior, Vol.15, No.4, pp.302-330,1983].

Although these passive-elastic performances are a kind of good approximate ankle function in slowly walking; Chang Su needs the outer external energy of plus with skelping; Any passive ankle-foot unit be can not pass through like this and [M.Palmer, " Sagittal plane characterization of normal human ankle function across a range of walking gait speeds, " Master ' s Thesis realized; Massachusetts Institute of Technology; Cambridge, MA, 2002; D.H.Gates; " Characterizing ankle function during stair ascent, descent, and level walking for ankle prosthesis and orthosis design. " Master ' s Thesis; Boston University, 2004; A.H.Hansen, D.S.Childress.S.C.Miff, S.A.Gard; K.P.Mesplay, " The human ankle during walking:implication for the design of biomimetic ankle prosthesis, " Journal of Biomechanics; Vol.37; Issue 10, pp.1467-1474,2004].This deficiency be reflected in below-Knee amputation use passive ankle-sufficient prosthese gait in.Than normal condition, it is from selecting leg speed slower, and stride is shorter.[D.A.Winter?and?S.E.Sienko.“Biomechanics?of?below.knee?amputee?gait，”Journal?of?Biomechanics，21，pp.361-367，1988]。In addition, its gait is obviously asymmetric, the ankle motion scope of natural side less [H.B.Skinner and D.J.Effeney, " Gait analysis in amputees, " Am J Phys Med, Vol.64, pp.82-89,1985; H.Bateni and S.Olney, " Kinematic and kinetic variations of below-knee amputee gait, " Journal of Prosthetics & Orthotics; Vol.14, No.1, pp.2-13; 2002], simultaneously, in the prosthese side; The hip stretch force moment is bigger, and less [D.A.Winter and S.E.Sienko. " Biomechanics of below-knee amputee gait, " the Journal of Biomechanics of gonocampsis moment; 21, pp.361-367,1988; H.Bateni and S.Olney, " Kinematic and kinetic variatiohs of below.knee amputee gait, " Journal of Prosthetics & Orthotics, Vol.14, No.1, pp.2-13,2002].They also consume bigger metabolic energy walking than the people who does not have amputation.[N.H.Molen，“Energy/speed?relation?of?below-knee?amputees?walking?on?motor-driven?treadmill，”Int.Z.Angew，Physio，Vol.31，p?173，1973；G.R.Colborne，S.Naumann.P.E.Longmuir，and?D.Berbrayer，“Analysis?of?mechanical?and?metabolic?factors?in?the?gait?of?congenital?below?knee?amputees，”Am.J.Phys.Med.Rehabil.，Vol.92，pp?272-278，1992；R.L.Waters，J.Perry，D.Antonelli，H.Hislop.“Energy?cost?of?walking?amputees：the?influence?of?level?of?amputation，”J?Bone?Joint?Surg.Am.，Vol.58，No.1，pp.4246，1976；E.G.Gonzalez，P.J.Corcoran，and?L.R.Rodolfo.Energy?expenditure?in?B/K?amputees：correlation?with?stump?length.Archs.Phys.Med.Rehabil.55，111-119，1974；D.J.Sanderson?and?P.E.Martin.“Lower?extremity?kinematic?and?kinetic?adaptations?in?unilateral?below-knee?amputees?during?walking.”Gait?and?Posture.6，126136，1997；A.Esquenazi，and?R.DiGiacomo.“Rehabilitation?After?Amputation，”Joum?Am?Podiatr?Med?Assoc，91(1)：13-22，2001]。These differences possibly be the result [A.D.Kuo that the amputee uses the ankle strength of bigger waist strength compensation disappearance; " Energetics of actively powered locomotion using the simplest walking model; " J Biomech Eng.; Vol.124, pp.113-120,2002; A.D.Kuo, J.M.Donelan, and A.Ruina; " Energetic consequences of walking like an inverted pendulum:Step-sto-step transitions, " Exerc.Sport Sci.Rev., Vol.33; No.2, pp.88-97,2005; A.Ruina, J.E.Bertram, and M.Srinivasan; " A collisional model of the energetic cost of support work qualitatively explains leg sequencing in walking and galloping; pseudo-elastic leg behavior in running and the walk-to-run transition. " J.Theor.Biol., Vol.237, No.2; Pp.170-192,2005].

Passive ankle-sufficient prosthese can not provide the ability of adaptation to the ground.For a kind of common economy rather than slowly fast gait is provided; Power ankle-sufficient prosthese be developed now [S.Au and H.Herr. " Initial experimental study on dynamic interaction between an amputee and a Dowered ankle-foot prosthesis; " Workshop on Dynamic Walking:Mechanics and Control of Human and Robot Locomotion; Ann Arbor; MI, May 2006; S.K.Au, J.Weber, and H.Herr; " Biomechanical design of a powered ankle-foot prosthesis, " Pmc.IEEE Int.Conf.On Rehabilitation Robotics, Noordwijk; The Netherlands, pp.298-303.June 2007; S.Au; J.Weber, E.Martinez-Villapando, and H.Herr. " Powered Ankle-Foot Prosthesis for the Improvement of Amputee Ambulation; " IEEE Engineering in Medicine and Biology International Conference.August 23-26; Lyon, France.pp.3020-3026,2007; H.Herr, J.Weber, S.Au. " Powered Ankle-Foot Prosthesis. " Biomechanics of the Lower Limb in Health; Disease and Rehabilitation.September 3-5, Manchester, England; Pp.72-74,2007; S.K.Au, " Powered Ankle-Foot Prosthesis for the Improvement of Amputee Walking Economy, " Ph.D.Thesis, Massachusetts Institute of Technology, Cambridge, MA, 2007; S.Au, J.Weber, and H.Herr. " Powered Ankle-foot Prosthesis Improves Walking Metabolic Economy, " IEEE Trans.on Robotics, Vol.25, pp.51-66,2009; J.Hitt, R.Bellman, M.Holgate; T.Sugar, and K.Hollander, " The sparky (spring ankle with regenerative kinetics) projects:Design and analysis of a robotic transtibial prosthesis with regenerative kinetics. " in Proc.IEEE Int.Conf.Robot.Autom.; Orlando; FL, pp 2939-2945, May 2006; S.K.Au; H.Herr; " On the Design of a Powered Ankle-Foot Prosthesis:The Importance of Series and Parallel Elasticity, " IEEE Robotics & Automation Magazine.pp.52-59, September 2008].Wherein some can be suitable with normal human subject ankle-foot on size and weight; And the energy content of battery [S.K.Au that has elastic energy storage, motor power and one day typical ambulatory activities can be provided; H.Herr. " On the Design of a Powered Ankle-Foot Prosthesis:The Importance of Series and Parallel Elasticity. " IEEE Robotics & Automation Magazine.pp.52-59, September 2008].

In these prostheses, use effective motor power to increase the problem of control.Before use in the work of these power set; The method of using is ankle moment state scattergram [S.K.Au, " Powered Ankle-Foot Prosthesis for the Improvement of Amputee Walking Economy, " Ph.D.Thesis of coupling normal human subject ankle expansion activity; Massachusetts Institute of Technology; Cambridge, MA, 2007; J.Hitt, R.Bellman, M.Holgate; T.Sugar, and K.Hollander, " The sparky (spring ankle with regenerative kinetics) projects:Design and analysis of a robotic transtibial prosthesis with regenerative kinetics; " In Proc.IEEE Int.Conf.Robot.Autom., Orlando, FL; Pp 2939-2945, May 2006; F.Sup, A.Bohara, and M.GoIdfarb. " Design and Control of a Powered Transfemoral Prosthesis; " The Intemational Journal of Robotics Research, Vol.27, No.2; Pp.263-273,2008].Provide work open loop that motor power means the ankle moment distribution when walking fast, can be supported, rather than only the behavior of spring-like is provided by passive device.Yet such control method demonstrates intrinsic self adaptation.Alternatively, along with their method of suitable selection, moment distribution all is essential for all activities wanted and landform variation.

Usually, existing commercially available active (active) ankle prosthese can only the reconstruct ankle-joint recovery phase, need several strides to restrain for the first time behind the kiss the earth to reach landform to adapt to the ankle position.In addition, they do not provide any normal gait required driving phase power, the pure support works state that therefore can not adaptation to the ground tilts.Particularly, power-actuated ankle-sufficient prosthese control model depends on by normal human body and passes the fixing ankle moment state relation that known topographic survey obtains with the target velocity walking.Although be that effectively these controllers do not allow the adaptation of environmental interference under leg speed of expecting and orographic condition, for example speed transients and landform change.

Have the biomethanics simulation study [H.Geyer that is applied to the leg motion as the tactful neuromuscular model of power positive feedback reflection on control basis recently; H.Herr; " A muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities, " (Submitted for publication); H.Geyer, A.Seyfarth, R.Blickhan, " Positive force feedback in bouncing gaits?, " Proc.R Society.Lond.B 270, pp.2173-2183,2003].These researchs show the hope to the adaptive demand of landform.

Summary of the invention

On the one hand, the present invention is a kind of controller and control method that is used for based on the bionic mechanical leg of human motion neuromuscular model.This control structure system is controlled the bionical moment of walking medium power leg prosthese, apparatus for correcting or ectoskeletal ankle, knee and hip joint.In a preferred embodiment, power set comprise artificial ankle and the knee joint that moment is controlled.The suitable joint moment that the feedback information that is provided by the sensor of each joint that is installed in the pedipulator device is confirmed is provided for the user.These sensors comprise; But be not limited to, use angle joint displacements and speed hall effect sensor or the analog of digital encoder, at the torque sensor at ankle and knee joint place and at least one Inertial Measurement Unit (IMU) between knee and ankle-joint.

Heat transfer agent from the joint state (position and speed) of pedipulator is used as the input to the neuromuscular model of human motion.It is definite by the spinal reflex model with rigidity to be used to confirm that from pedipulator joint state heat transfer agent the internal state of each virtual muscle, muscle activity should be given the single virtual muscular force of what level.The leg prosthese that if pedipulator is the above-knee amputee to be dressed, the angular transducer of ankle and knee is measured the joint state in these joints.For hip joint, the absolute orientation of user's thigh is confirmed with the IMU that is positioned at prosthese knee and ankle by the joint angles sensor at prosthese knee place.In order to estimate hip position and speed, the control structure system is worked under the situation of supposition gait process middle and upper part body (incompleteness) the relative vertical position of maintenance.

On the one hand; The present invention is the neural mechanical control device based on model that is used to comprise the mechanical limb at least one joint, and this controller comprises the finite state machine that is configured to receive the feedback data that relates to the mechanical limb state and definite mechanical limb state, be configured to receive from the status information of finite state machine and from the muscle geometry and the reflection system information of at least one database and the muscle model processor and being configured to of joint moment or rigidity instruction that utilizes the neuromuscular model to confirm to be sent at least one expectation of mechanical limb and control the definite bionical moment of the muscle model processor of (domination) mechanical limb joint and the joint instruction processing unit of rigidity.In a preferred embodiment, at least one sensor of each joint of feedback data through being installed in mechanical limb provides.In a further advantageous embodiment, mechanical limb is a leg, and finite state machine is synchronized with the leg gait cycle.

On the other hand; The present invention is a kind of method based on model; Be used to control the mechanical limb that comprises at least one joint; It may further comprise the steps: receive the feedback data relevant with the mechanical limb state at the finite state machine place; Utilize the feedback data of finite state machine and reception to confirm the mechanical limb state, utilize neuromuscular model, muscle geometry, reflector architecture information and confirm to be sent to the joint moment or the rigidity instruction of at least one expectation of mechanical limb, and control bionical moment and the rigidity that the muscle model processor of mechanical limb joint is confirmed from the status information of finite state machine.

Description of drawings

When accompanying drawing was considered with the following detailed description of the present invention, it is more obvious that others of the present invention, advantage and novel feature will become; Wherein:

Fig. 1 is the block diagram of the exemplary embodiment of common according to an aspect of the present invention neuromuscular model architecture;

Fig. 2 A-F shows according to an aspect of the present invention in developing six stages of common neuromuscular model architecture;

Fig. 3 illustrates according to the present invention, generate according to the model of the one side of common neuromuscular model architecture;

Fig. 4 A and 4B show the self-organizing walking and the corresponding ground reaction force of people's class model dynamic interaction between model and ground according to an aspect of the present invention respectively;

Fig. 5 A-C has compared model and people according to an aspect of the present invention respectively and has stablized walking state hip, knee and ankle;

Fig. 6 A-D shows the adaptability of going upstairs according to an aspect of the present invention, comprises model snapshot (Fig. 6 A), net work (Fig. 6 B), extensor activity pattern (Fig. 6 C) and corresponding ground reaction force (Fig. 6 D);

Fig. 7 A-D shows the adaptability that walking is according to an aspect of the present invention gone downstairs, and comprises model snapshot (Fig. 7 A), net work (Fig. 7 B), extensor activity pattern (Fig. 7 C) and corresponding ground reaction force (Fig. 6 D);

Fig. 8 is the schematic diagram of muscle-tendon model according to an aspect of the present invention;

Fig. 9 shows contact model according to an aspect of the present invention;

Figure 10 A-C shows the ankle-sufficient prosthese embodiment that is used for preferred embodiment according to an aspect of the present invention, shows real system (Figure 10 A), drive chain figure (Figure 10 B) and kinetic model (Figure 10 C) respectively;

Figure 11 be according to an aspect of the present invention with gait cycle synchronous have a sketch map at the embodiment of the finite state machine of the state switching threshold of each state and ankle-sufficient biomethanics equivalent, be used to implement the top control of ankle-sufficient prosthese of Figure 10 A-C;

Figure 12 is that ankle-sufficient prosthese is controlled the block diagram of exemplary system property embodiment according to an aspect of the present invention;

Figure 13 A-C is the interior prosthese moment example plot of a complete gait cycle of level ground (Figure 13 A), acclivity (Figure 13 B), the three kinds of walking states in decline slope (Figure 13 C) according to an aspect of the present invention;

The exemplary embodiment of the muscle skeleton model that Figure 14 A-C shows according to an aspect of the present invention as on the prosthese microcontroller, implements comprises two connection ankle-joint models (Figure 14 A), the Hill type muscle model (Figure 14 B) that is shown specifically and the geometry (Figure 14 C) that is attached to the muscle model on the bone;

Figure 15 shows the exemplary embodiment of virtual according to an aspect of the present invention plantar flexion of foot muscular reflex strategy, the relation between the plantar flexion of foot pars muscularis that comprises ankle angle, muscular force and ankle moment divides;

Weight and the moment of the biological ankle that the height main body is mated and the comparison between the angle track that Figure 16 A and 16B show the prosthese moment and the angle track of the amputee's main body that measures in the test and have complete limbs comprise ankle moment and ankle angle respectively;

Figure 17 is the comparison of moment distribution figure and biological moment distribution after the parameter optimization according to an aspect of the present invention; And

Figure 18 A-C is prosthese moment-angle geometric locus that experimental measurement obtains in the exemplary embodiment of the present invention under level ground (Figure 18 A), acclivity (Figure 18 B), decline slope (Figure 18 C) three kinds of different walking states.

The specific embodiment

A kind of hierarchy of control structure proposes to control powered leg prosthese, rectifier or ectoskeletal ankle, knee and the hip joint bionical moment when walking.In this was implemented, power set comprised artificial ankle and the knee joint that moment is controlled.The suitable moment that the feedback information that is provided by the sensor of each joint that is placed in the pedipulator device is confirmed is provided for the user.These sensors comprise; But be not limited to, use angle joint displacements and the torque sensor of speed hall effect sensor or analog, knee and ankle and at least one inertial measurement unit (IMU) between knee joint and ankle-joint of digital encoder.

Be used as the input of the neuromuscular model of human motion from the heat transfer agent of the joint state (position and speed) of pedipulator (hip, knee and ankle).This model uses confirms the internal state of its each virtual muscle from the joint state heat transfer agent of pedipulator, and sets up the power and the rigidity of each virtual muscle of the definite muscle activity that should give what specified level of spinal reflex model.If pedipulator is the leg prosthese that the amputee dresses on the leg, measure the joint state in these joints at the angular transducer at ankle and knee place.For hip joint, the absolute orientation of user's thigh is confirmed by the angle joint sensors and the IMU between prosthetic knee joints and ankle-joint of prosthese knee.In order to estimate hip joint position and speed, the supposition of control structure system is worked under the condition of gait process middle and upper part body (incompleteness) the relative vertical position of maintenance.

As use here and herein with reference in the application of quoting, following term clearly includes, but are not limited to:

A kind of motor of " actuator " expression as giving a definition.

" agonist " expression contraction elements, it is by another element, Opposing muscle opposing or obstruction.

" agonist-Opposing muscle actuator " represented a kind of mechanism, comprises that (at least) two actuators oppose each other to work: the agonist actuator, when it makes two elements of time spent pulling to together, and the Opposing muscle actuator when it does the time spent, impels two elements to separate.

" Opposing muscle " represented a kind of expansion member, and it is by another kind of element, shrinker opposing or obstruction.

" bionical " expression man-made structures or mechanism, its mimic biology structure or mechanism, like the characteristic or the behavior of joint or limbs.

" dorsiflex " crooked ankle-joint of expression so that sufficient end move up.

" elasticity " expression is owing to can recover original-shape after stretching or the compression.

" stretching, extension " expression centers on the bending motion in joint in the limbs, and it increases the angle between joint limbs bone.

" bending " expression centers on the bending motion in joint in the limbs, and it reduces the angle between the joint limbs bone.

" motor " expression is (active) element initiatively, and it comprises electronic, pneumatic or hydraulic motor and actuator through being the mechanical energy generation with the power conversion that provides or moving.

The crooked ankle-joint of " sole of the foot is bent " expression makes sufficient end move down.

" spring " expression elastic device, for example wire coil or platy structure, its recovery original-shape being compressed or stretching after.

According to an aspect of the present invention, a kind of exemplary embodiment of the control scheme based on the neuromuscular model is shown in the block diagram of Fig. 1.Among Fig. 1, neuromuscular model according to the present invention comprises the tore of reflection 110 of each model muscle unit 120.Power and the rigidity calculated by all model muscle are used to use from 140 calculating 130 joint moments and the rigidity of the muscle arm of force value in the document.Then model assessment is sent to useful moment and the rigidity value of controller as 150 expectations of bionic mechanical leg joint.Controller 160 is the moment and the rigidity value at each mechanical joint 150 place of track record then.

In order to make each virtual muscle produce the power that it needs, muscular irritation parameter S TIM (t) needs.This parameter can be confirmed by outside input or LOCAL FEEDBACK ring.In the bionic leg control strategy of embodiment, STIM (t) calculates based on the LOCAL FEEDBACK ring.This architecture is based on the reflection feedback framework [H.Geyer by Geyer and Herr exploitation; H.Herr; " A muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities; " (submitted to and be used for publishing), whole herein the merging introduced reference]. in this framework, neural control is designed to imitate the stretching reflection of normal human's muscle.Should allow the bionic mechanical leg to reproduce type people's joint mechanism based on the control method of neuromuscular reflex.

Neural mechanism model.Have leg strength and learn the human model reckoning people's walking dynamics and the muscle activity of principle coding reflection control.When the neuroscientist discerned the neutral net of complicated day by day control animal and human gait, biomechanist was found, learns principle if be careful leg strength, and motion needs motor to control hardly.Show here and how to make that the muscle reflex behavior is vital to getting in touch two kinds of observed results.Developed a kind of model of people's motion, its muscle reflex behavior of learning principle by the coding leg strength drives.Equipped based on the reflection of this principle control because the dynamic interaction on itself and ground, this model stability gait, allow the ground disturbance, but and self adaptation stair environment.In addition; This models show goes out and the qualitative uniformity of joint angles, joint moment and muscle activity that test draws, has hinted that the output of human motion flesh to a great extent can be by leg strength being learned the muscle reflex behavior shaping that principle is linked to the neutral net of being responsible for motion.

The human walking model that has motor (motor) control reflects based on muscle, and it is designed to include leg strength and learns principle.These principles comprise dependence [Blickhan, R., 1989.The spring-mass model for running and hopping.J.of Biomech.22, the 1217-1227 of driving phase to being obedient to leg from the simple concept model of leg motion; Ghigliazza, R., Altendorfer, R., Holmes, P., Koditschek, D., 2003.A simply stabilized running model.SIAM J.Applied.Dynamical Systems 2 (2), 187-218; Geyer, H., Seyfarth, A.; Blickhan, R., 2006.Compliant leg behaviour explains the basic dynamics of walking and running.Proc.R.Soc.Lond.B 273,2861-2867]; Stability [Seyfarth, A., G ü nther, M. based on the segmentation leg of static joint moment balance; Blickhan, R., 2001.Stable operation of an elastic three.segmented leg.Biol.Cybern.84,365-382; G ü nther, M., Keppler, V., Seyfarth, A.; Blickhan, R., 2004.Human leg design:optimal axial alignment under constraints.J.Math.Biol.48,623-646], impact [the Mochon that safeguards that swings leg mechanism; S., McMahon, T., 1980.Ballistic walking.J.Biomech.13 (1), 49-57]; And use the withdrawal of leading leg to strengthen leg speed stability [Seyfarth, A., Geyer, H., G ü nther; M., Blickhan, R., 2002.A movement criterion for running.J.of Biomech.35,649-655; Seyfarth, A., Geyer, H., Herr, H.M., 2003.Swing-leg retraction:a simple control model for stable running.J.Exp.Biol.206,2547-2555].Hill type muscle is also comprising the use that combines of the positive feedback scheme of power and length with spinal reflex, with these mechanical characteristics of encoding effectively.

Than from the model behavior dynamic (dynamical), kinematic and the myoelectricity phenomenon that has, demonstrated and have the leg strength that is designed to encode and learn the motor control neuromuscular model of principle and can generate biological walking mechanics and muscle activity about the works of human walking.This reflection control allows model to allow the flip-flop of ground level, adapts to stair and rises and descend, and interferes and need not variable.

Reflect neuromuscular that two, three of control save the biped of legs to having upper torso and every joint by seven muscle drivings and by muscle from the conceptual point quality model, the development of human structure of models and control is divided into six stages.According to this aspect of the present invention, Fig. 2 A-F shows six stages that common neuromuscular model structure system develops.First three stage is integrated is obedient to the behavior (Fig. 2 A-C) of leg at driving phase with having stablized.Stage has increased upper torso and its Balance Control (Fig. 2 D).Latter two stage prepare and guaranteed in swing process leg forward with regain action (Fig. 2 E and Fig. 2 F).

In Fig. 2 A-F, describe in more detail in the subsequent paragraph, from supporting leg configuration (Fig. 2 A), be obedient to the leg behavior and develop through having positive feedback power F+ driving musculus soleus (SOL) and block femur muscle group (VAS) generation walking and run (Fig. 2 B) as key factor.In order to prevent that knee from excessively stretching, the additional two joint gastrocnemiuses (GAS) (Fig. 2 C) that use F+, and if knee joint stretch and surpass 170 degree threshold value VAS and begin to be suppressed.In order to prevent that ankle from excessively stretching, add the muscle (TA) of shin bone front, it holds ankle-joint to the flexing position through length positive feedback L+, and is suppressed at normal condition through the power negative-feedback F-from musculus soleus.In order to allow the leg swing, added upper torso (Fig. 2 D).(GLU HAM) is driven to reference gradient with respect to vertical direction, and two joint HAM prevent excessively to be stretched by the knee of hip extensor moment generation here for its hip musculus flexor (HFL) through supporting leg and interactional hip extensor.Another of landing (dominating) leg passes through increase/minimizing respectively to the continuous pump of HFL/GLU, and swings with respect to ratio (Fig. 2 E) startup of the load of another leg carrying through suppressing VAS.Actual leg swing promotes to be suppressed (Fig. 2 F) up to it by the L-of HAM through the HFL that uses L+.The inclination that the excitation of HFL depends on upper torso especially when standing up.And, use F+ to regain and stretch leg to GLU and HAM towards oscillation end.At last, the L+ of the TA that does not suppress now drives ankle to bending position (Fig. 2 G).

The biddability of supporting leg and stability.Spring-mass two sufficient models are used as the starting point (Fig. 2 A) of human motion conceptual foundation.Although this model only is made up of point-quality 205, it moves ahead on two no quality springs 210,215; It has reproduced in people's walking and observed mass centre dynamics when running, and under the conceptual framework of being obedient to the leg behavior based on holding state, has unified two kinds of gaits [Geyer, H.; Seyfarth; A., Blickhan, R.; 2006.Compliant leg behaviour explains the basic dynamics of walking and running.Proc.R.Soc.Lond.B 273,2861-2867].

In order to realize submissive behavior at the neuromuscular leg; Each spring 210,215 is by the additional thigh 220 of displacement; Shin bone 225 and foot 230; And musculus soleus (SOL) 235 and block femur muscle group (VAS) 240, the both generates their muscle activity through the positive feedback (F+) of localized forces at the driving phase (Fig. 2 B) of gait.The reflection of this power with Geyer, H., Seyfarth, A., Blickhan, R., 2003 at Positive force feedback in bouncing gaits? Proc.R.Soc.Lond.B 270, same mode modeling among the 2173-2183.Under the positive feedback condition of power, the excitation Sm (t) of muscle m is predrive S0, the time lag of m and muscle (Δ t) and gain (G) power Fm with Sm (t)=SO, m+GmFm (t-Δ tm).

When to be obedient to the leg behavior be basic, it also threatened stability [Seyfarth, the A. of segmental appendage leg joint; G ü nther, M., Blickhan; R., 2001.Stable operaion of an elastic three.segmented leg.Biol.Cybern.84,365-382; Giinther, M., Keppler, V., Seyfarth, A., Blickhan, R., 2004.Human leg design:optimal axial alignment under constraints.J.Math.Biol.48,623-646].In the segmental appendage leg, knee and ankle moment, τ _aAnd τ _b, observe standing balance τ a/ τ b=hk/ha, wherein hk and ha are respectively the vertical ranges from knee and ankle to sufficient force vector Fleg.Effectively, when big stretching moment of torsion on another joint is ordered about in a joint near Fleg, the generation of the leg of its spring-like behavior of threat is excessively stretched [in detail referring to Seyfarth; A., G ü nther, M.; Blickhan; R., 2001.Stable operaion of an elastic three-segmented leg.Biol.Cybem.84,365-382].

Through (TA) 250 muscle (Fig. 2 C) of additional gastrocnemius (GAS) 245 and shin bone front side, this trend that excessively stretches on knee and ankle by converse.Be similar to SOL and VAS, use positive localized forces feedback (F+) at the driving phase two joint GAS of gait.This muscle reflection prevents that not only knee from causing excessive stretching because of the big stretching moment of torsion of ankle, also contributes to generating all leg behaviors of being obedient to.On the contrary, simple joint TA use have STA (t) apart from local positive feedback (L+), STA (t)=S _{0, TA}+ G _TA(l _{CE, TA}-l _{Off, TA}) (t-Δ _{T, TA}), wherein, l _{CE, TA}Be the TA fibre length, l _{Off, TA}It is length offset.Club foot, the L+ of TA prevents that ankle from excessively stretching when producing big knee moment of torsion.If yet the enough active equalising torque that can keep knee and ankle of ankle extensor, this muscle reflection is unwanted.For fear of the unnecessary antagonism of TA and SOL under this state, the TA excitation is suppressed by the negative force feedback (F-) from SOL, causes S _TA(t)=S _{0, TA}+ G _TA(l _{CE, TA}-l _{Off, TA}) (t-Δ _{T, TA})-G _SOLTAF _SOL(t-Δ t _SOL).Avoid excessively pulling in order further to protect knee; If stretching, knee surpasses 170 degree threshold values; VAS is suppressed; Wherein,

is proportional gain,

be the knee angle.This reflection only suppresses under condition and knee works when being actually extended state.

Upper torso and its balance.In the next stage of the sufficient neuromuscular model evolution of spring-quality concept model to two, the point mass representation is abandoned, and has introduced lead leg (Fig. 2 D) around upper torso 255.This upper torso 255 has made up head, arm and trunk (HAT).For balance HAT 255 at the volley, every leg has all added buttocks muscles crowd (GLU) 260 and hip flexor muscles crowd (HFL) 265.GLU260 and HFL265 be by the proportion differential signal excitation of HAT255, with respect to the gravity θ angle that turns forward, S _GLU/HFL～± [k _p(θ-θ _Ref)+k _dD θ/dt], k wherein _pAnd k _dBe the ratio and the differential gain, and θ _RefBe reference tilt angle [comparison of similarity method, for example, G ü nther, M., Ruder, H., 2003.Synthesis of two-dimensional human walking:a test of the λ-model.Biol.Cybem.89,89-106].Also comprise and use S _HAM～S _GLuReverse when the heavy HAT255 of pulled backwards because the excessive diarticular hamstring muscle group (HAM) 270 of stretching of the knee that the big hip moment that GLU260 produces causes.Because if leg bears enough weight, hip moment can only balance HAT255, and the excitation of GLU260, HAM270 and HFL265 is modulated into the gross weight that every leg bears health pro rata.As a result, the hip muscle of every leg only works in the Balance Control of driving phase to HAT.

Lead leg and stretch out and regain.Except the muscle reflection control that generation is led leg and stretched out and regain, human structure of models is complete.Suppose the functional importance of supporting leg and the contralateral leg of dual support and the proportional minimizing of health gross weight (bw) (Fig. 2 E) that the elongation of initial swing leg is born.Which leg people's class model detects and gets into support latter end (contralateral leg); And the ratio of the weight of bearing in contralateral leg suppresses the F+ of the VAS240 of health corresponding leg;

stretches and promotes the leg built on stilts forward the time when ankle; Allow knee to interrupt its function spring behavior to lateral inhibition, and crooked.If although have only ankle to promote effectively, this catapult mechanism can start swing, in the dual-gripper stage, this model can also be through starting the elongation of leading leg with the excitation of fixed amount Δ S increase HFL265 and the excitation of minimizing GLU260.

In the recovery phase of reality, mainly rely on the leg ballistic motion, but it is influenced (Fig. 2 F) by two aspects.The elongation of leading leg on the one hand, is pushed.Through in the swinging in the cross rest transfer process, using HAT255 pitch angle θ forward _RefThe length positive feedback (L+) of setovering, HFL265 is energized S _{HFL (t)}=S _{0, HFL}+ k _Lean(θ-θ _Ref) _TO+ G _HFL(l _{CE, HFL}-l _{Off, HFL}) (t-Δ _{T, HFL}).Make in this way, guaranteed that the ballistic motion of leading leg in time obtains proal moment [Mochon, S., McMahon, T., 1980.Ballistic walking.J.Biomech.13 (1), 49-57].

In addition, lead leg and also be prevented from excessive stretching, and guaranteed to regain.If recovery phase leg to arrive and keep suitable location, leg system self-stabilization be a gait cycle [McGeer, T., 1990.Passive dynamic walking.Int.J.Rob.Res.9 (2), 62-82; Seyfarth, A., Geyer, H., Giinther, M., Blickhan.R., 2002.A movement criterion for running.J.of Biomech.35,649-655; Ghigliazza, R., Altendorfer, R., Holmes, P., Koditschek, D., 2003.A simply stabilized running model.SIAM J.Applied.Dynamical Systems 2 (2), 187-218; Geyer, H., Seyfarth, A., Blickhan, R., 2006.Compliant leg behaviour explains the basic dynamics of walking and running.Proc.R.Soc.Lond.B 273,2861-2867].If lead leg and landing preceding withdrawal; The permission of this mechanical self-stabilization to disturbance rejection can be increased substantially [Seyfarth; A.; Geyer; H., 2002.Natural control of spring.like running-optimized self-stabilization.In:Proceedings of the 5th international conference on climbing and walking robots.Professional Engineering Publishing Limited, pp.81-85; Seyfarth, A., Geyer, H., Herr, H.M., 2003.Swing.leg retraction:a simple control model for stable running.J.Exp.Biol.206.2547-2555].In order to realize this stopping-regaining strategy, three kinds of muscle reflections are contained in people's class model.When knee reached full extension in the elongation process, the stretching transition of leading leg that the motive force forward that is received by leg produces was prevented from.About this, the L+ quilt of HFL suppresses S pro rata with the tension force that HAM received in recovery phase _HFL(t)=k _Lean(θ-θ _Ref) TO+G _HFL(l _{CE, HFL}-l _{Off, HFL}) (t-Δ _{T, HFL})-G _HAMHFZ1(l _{CE, HAM}-l _{Off, HAM}) (t-Δ _{T, HAM}).In addition, F+ is used to GLU, S _GLU(t)=S _{0, GLU}+ G _GLUF _GLU(t-Δ t _GLU), and HAM, S _HAM(t)=S _{0, HAM}+ G _HAMF _HAM(t-Δ t _HAM), guarantee to depend on actual elongation moment, lead leg and not only can hover, can also transmitting portions moment align and regain to leg.At last, introducing guarantees that the TA L+ of sufficient error was held in whole recovery phase.Inoperative at this stage SOL, GAS, VAS.

Reflection control parameter.Reflections affect the function management through in model use different to muscle excitation Sm (t).There is not optimum parameters to be employed.Parameter from before reflex behavior (F+, L+) knowledge or get through specious estimation.Be converted into muscle activity Am (t) before, all muscle excitations are defined between the scope 0.01 to 1.Table 1 has been represented the supporting reflex function that uses in this preferred embodiment.

Table 1

S _SOL(t)＝S _0，SOL+G _SOLF _SOL(t _l)

＝0.01+1.2/F _max，SOLF _SOL(t _l)

S _TA(t)＝S _0，TA+G _TA[l _CE，TA(t _l)-l _off，TA)]-G _SOL，TAF _SOL(t _l)

＝0.01+1.1[l _CE，TA(t _l)-0.71l _opt，TA)]-0.3/F _max，SOLF _SOL(t _l)

S _GAS(t)＝S _0，GAS+G _GASF _GAS(t _l)

＝0.01+1.1/F _max，GASF _GAS(t _l)

$-k_{\mathrm{bw}}\left|F_{\mathrm{leg}}^{\mathrm{contra}}\left(t_s\right)\right|*\mathrm{DSup}$

$-0.00167\left|F_{\mathrm{leg}}^{\mathrm{contra}}\left(t_s\right)\right|*\mathrm{DSup}$

$S_{\mathrm{HAM}}\left(t\right)={\mathrm{<figure-callout\; id="0"\; label="S"\; filenames="HPA00001462734000033.png,HPA00001462734000042.png"\; state="\{\{state\}\}">S</figure-callout>}}_{0,\mathrm{HAM}}+\left\{k_p\right[\mathrm{\&theta;}\left(t_s)-{\mathrm{\&theta;}}_{\mathrm{ref}}]+k_d\mathrm{d\&theta;}/\mathrm{dt}\left(t_s\right)\right\}+k_{\mathrm{bw}}\left|F_{\mathrm{leg}}^{\mathrm{ipsi}}\left(t_s\right)\right|$

$=0.05+\left\{1.9\right[\mathrm{\&theta;}\left(t_s\right)-0.105]+0.25\mathrm{d\&theta;}/\mathrm{dt}\left(t_s\right)\}+0.00167\left|F_{\mathrm{leg}}^{\mathrm{ipsi}}\left(t_s\right)\right|$

$S_{\mathrm{GLU}}\left(t\right)={\mathrm{<figure-callout\; id="0"\; label="S"\; filenames="HPA00001462734000033.png,HPA00001462734000042.png"\; state="\{\{state\}\}">S</figure-callout>}}_{0,\mathrm{GLU}}+\left\{k_p\right[\mathrm{\&theta;}\left(t_s\right)-{\mathrm{\&theta;}}_{\mathrm{ref}}]+k_d\mathrm{d\&theta;}/\mathrm{dt}\left(t_s\right)\}+k_{\mathrm{bw}}\left|F_{\mathrm{leg}}^{\mathrm{ipsi}}\left(t_s\right)\right|$

$-\mathrm{\&Delta;S}*\mathrm{DSup}$

$=0.05+\left\{1.3\right[\mathrm{\&theta;}\left(t_s\right)-0.105]+0.25\mathrm{d\&theta;}/\mathrm{dt}\left(t_s\right)\}+0.00167\left|F_{\mathrm{leg}}^{\mathrm{ipsi}}\left(t_s\right)\right|$

$-0.25*\mathrm{Dsup}$

$S_{\mathrm{HFL}}\left(t\right)={\mathrm{<figure-callout\; id="0"\; label="S"\; filenames="HPA00001462734000033.png,HPA00001462734000042.png"\; state="\{\{state\}\}">S</figure-callout>}}_{0,\mathrm{HFL}}+\left\{k_p\right[\mathrm{\&theta;}\left(t_s\right)-{\mathrm{\&theta;}}_{\mathrm{ref}}]+k_d\mathrm{d\&theta;}/\mathrm{dt}\left(t_s\right)\}-k_{\mathrm{bw}}\left|F_{\mathrm{leg}}^{\mathrm{ipsi}}\left(t_s\right)\right|$

$+\mathrm{\&Delta;S}*\mathrm{DSup}$

$=0.05+\left\{1.9\right[\mathrm{\&theta;}\left(t_s\right)-0.105]+0.25\mathrm{d\&theta;}/\mathrm{dt}\left(t_s\right)\}-0.00167\left|F_{\mathrm{leg}}^{\mathrm{ipsi}}\left(t_s\right)\right|$

$+0.25*\mathrm{DSup}$

(tl=t-20ms, tm=t-10ms, ts=t-5ms, trailing leg DSup is 1 if leg is the dual-gripper state, otherwise 0, { }+/-just only quoting/negative value)

Table 2 has been represented the pendular reflex function that uses in this preferred embodiment.

Table 2

S _SOL(t)＝S _0，SOL

＝0.01

S _TA(t)＝S _0，TA+G _TA[l _CE，TA(t _l)-l _off，TA)]

＝0.01+1.1[l _CE，TA(t _l)-0.71l _opt，TA)]

S _GAS(t)＝S _0，GAS

＝0.01

S _VAS(t)＝S _0，VAS

＝0.01

S _HAM(t)＝S _0，HAM+G _HAMF _HAM(t _s)

＝0.01+0.65/F _max，HAMF _HAM(t _s)

S _GLU(t)＝S _0，GLU+G _GLUF _GLU(t _s)

＝0.01+0.4/F _max，GLUF _GLU(t _s)

S _HFL(t)＝S _0，HFL+GH _FL[l _CE，HFL(t _s)-l _off，HFL]-G _HAM，HFL[l _CE，HAM(t _s)-l _off，HAM]

+{k _lean[θ(t _s)-θ _ref]}PTO

＝0.01+0.35[l _CE，HFL(t _s)-0.6l _opt，HFL]-4[l _CE，HAM(t _s)-0.85l _opt，HAM]

+{1.15[θ(t _s)-0.105]}PTO

({ } PTO: get normal value when liftoff preceding.)

The result.Do not send the central pattern generator (cpg) (CPG) that acts on its muscle forward to although people's class model does not have, it all is positioned at the sensor that sole and heel place on every pin detect ground through use to every leg, is supporting and is swinging between the different reflections and change.As a result, the dynamic interaction of model and exterior mechanical environment becomes the pith that generates muscle activity.Fig. 3 illustrates mode generator according to an aspect of the present invention.In Fig. 3, be not the maincenter pattern, but reflection generate muscle excitation, Sm305,310.A left side (L) 320 has the support 340,350 that separates and swings 350,355 reflections with right (R) 330 legs, and it is selected based on the contact sensing 360,365 from sole and heel sensor 370,375.Reflection output depends on the mechanics input, Mi380,385, and mechanics and motor control are interlaced.

The walking gait.How important in order to study having interdepended of concerning human motion mechanics and motor control, this model is starting point with common walking speed v0=1.3m/s from the left leg of its holding state and the right leg of swing state.Because the reflection of the muscle of simulation comprises the time lag of maximum 20ms, all muscle is not have response at first.According to an aspect of the present invention, Fig. 4 A and 4B show the human walking model of self-organizing of dynamic interaction between model and the ground respectively, and corresponding ground reaction force.In Fig. 4 A and 4B, the every 250ms of people's class model snapshot takes once (Fig. 4 A), and left leg 405,410 is illustrated (30Hz LPF) with right leg 415,420 corresponding model GRF (Fig. 4 B) with the curve that separates.Horizontal velocity with 1.3m/s begins, and this model slowed down in two step, returned to identical walking speed then rapidly.Leg muscle only shows right leg 415, indication muscle activity＞10%.The primary condition of every leg (pa; K; H (ankle, knee and calcaneal definition)) be: 175 degree; 175 degree (left leg) and

, 175 degree, 140 degree (right leg).

Because the disturbance primary condition, this model lost efficacy slightly and slowed down (Fig. 4 A) in the first step.Yet suitable if parameter is selected, this model promptly recovers in step subsequently, because the dynamic interaction self-organizing walking between model and the ground.Here, the vertical ground reaction force of the leg of driving phase (GRF) shows the graphic feature (Fig. 4 B) of M type walking gait, has shown in the stable state walking integral power characteristic like the model and physiognomy.

The steady-state mode of angle, moment and muscle activity.This similitude can also continue further to observe; This models show angle, moment and muscle activity pattern aspect qualitative, be consistent with known human walking data.According to an aspect of the present invention, Fig. 5 A-C has compared the walking with 1.3m/s respectively, the stable state of model and people's hip (Fig. 5 A), knee (Fig. 5 B) and ankle (Fig. 5 C).In Fig. 5 A-C, with knocking behind the pin on the one leg with being standardized as a stride to calcaneal knocking, the muscle activity of the hip of this model, knee and ankle, moment and angle steady-state mode and human walking data compare (by Perry, 1992 rewrite).Vertically dotted line 510 demonstrates about 60% stride toe built on stilts.By muscle relatively is long adductor muscle (HFL) 520, big flesh (GLU) 530 on the stern, semimembranosus (HAM) 540, and musculus vastus lateralis (VAS) 550.

Model prediction and the maximum consistent ankle (Fig. 5 C) that is found in of walking data.The observed ankle motion of human ankle when reflection model not only generates walking

With moment τ _a, and calculate that SOL, TA and GAS are movable, it is similar to test SOL, TA and the GAS activity of inferring from its surperficial flesh streaming current describer.For SOL and GAS, except use local F+ reflection at driving phase, all producing should activity.For TA, the plantar flexion of foot flesh of active response foot when its L+ is reflected in early stage driving phase, but suppressed by F-at the other times in this stage from SOL.Have only when (60% stride) SOL in supporting to the transfer process of swing is movable and reduce, the L+ of TA recovers, pulling foot antagonism plantar flexion of foot flesh.

Above-mentioned knee and the relatively poor uniformity of hip of relatively demonstrating.For example; Although the track of common human knee

is caught by model, the knee jerk of this model is approximately Duoed than human early stage driving phase and 10 is spent or 30% (Fig. 5 B).Relate to this a large amount of knee jerk, it is movable that this model lacks the VAS that observes the swing in late period that enters into early stage driving phase subsequently.Only after heel strike, the F+ of VAS participates in and can act on the load of musculature response leg.The movable time lag of extensor not only causes early stage driving phase more weak knee relatively, and heavy HAT turns forward after impacting.Because the Balance Control of HAT is little by little participated in the weight that supporting leg bears, the balance reflection will be inoperative up to heel strike always, must produce a large amount of GLU and HAM activity then artificially and make HAT return it with reference to gradient (Fig. 5 c).Therefore, the hip track of model

With moment figure τ _hAt least be similar to the mankind, human hip extensor GLU and HAM were acting before impacting, and prevented trunk over-tilting like this.

The adaptivity that ground changes.Although reflection control is limited, but people's class model allows and self adaptation is unexpected, permanent ground level is changed.Fig. 6 A-D shows model and has met with the example that rises the stair of 4cm on a series of every rank.According to an aspect of the present invention, Fig. 6 A-D shows the adaptability of going upstairs, and comprises the snapshot (Fig. 6 A) of model, net work (Fig. 6 B), and extensor activity pattern (Fig. 6 C), and to the response (Fig. 6 D) of ground reaction force.In Fig. 6 A-D, with the method that 1.3m/s walks steadily, eight strides of people's class model are illustrated, and it has comprised five steps on per step 4cm slope.This model has recovered steady walking in the 8th step.Each step is defined as knocks heel from the heel of right leg and knocks.Being shown in Fig. 6 A, is the snapshot of model when the right crus of diaphragm heel knocks with the toe built on stilts.For this leg, further be illustrated in Fig. 6 B, the net work that driving phase produces at hip, knee and ankle is owing to stretch the positive work that produces; In Fig. 6 C, the operational factors binding mode of five extensors in per step; In Fig. 6 D, with respect to the normalized corresponding ground reaction force 650 of body wt (bw), the ground reaction force 660 of left leg also is wherein, is used for comparison.

From stable state walking (first step) beginning, stair (Fig. 6 A) are run at the second step latter end by the foot of the right leg that this model stretches out.Impact early makes model slow down, the upper body that turns forward, and its big hip moment that is produced by GLU and HAM is resisted (the 3rd step, Fig. 6 B and 6C).Tend to the knee that stretches because hip stretches moment, even if roughly the same in support maintenance stage and stabilization sub stage at the net work of knee, VAS does not feel that in so big power of stabilization sub stage its force feedback control has reduced muscular irritation (Fig. 6 C) yet.On the contrary, model is slack-off to have reduced the power that driving phase ankle extensor GAS and SOL feel, their force feedback reflection has produced slight a small amount of muscular irritation, has reduced the net work (Fig. 6 B and 6C) of ankle.In the 4th step and the 5th step, this model has been accustomed to the walking of going upstairs of about 1m/s, wherein forward and impact backward mainly by the generation of hip and knee.After the 6th step arrived the stationary stage of rising, this model recovered original stable walking speed 1.3m/s in the 8th step.

According to an aspect of the present invention, Fig. 7 A-D shows the adaptability that walking is gone downstairs, and comprises model snapshot (Fig. 7 A), net work (Fig. 7 B), extensor activity pattern (Fig. 7 C), and corresponding ground reaction force (Fig. 7 D).In Fig. 7 A-D, begin from the walking of 1.3m/s stable state, eight steps of people's class model are illustrated, and have comprised five steps on per step 4cm slope.This model has recovered steady walking in the 8th step.Each step is defined as knocks heel from the heel of right leg and knocks.Being shown in Fig. 7 A, is the snapshot of model when the right crus of diaphragm heel knocks with the toe built on stilts.For this leg, further be illustrated in Fig. 7 B, the net work that driving phase produces at hip, knee and ankle is owing to stretch the positive work that produces; In Fig. 7 C, the operational factors binding mode of five extensors in per step; And in Fig. 7 D, with respect to the normalized corresponding ground reaction force 750 of body wt (bw), the ground reaction force 760 of left leg also is wherein, is used for comparison.This model recovers stable state walking 1.3m/s in the 14th step, after accomplishing five steps of per step downslope 4cm.

Fig. 7 A-D sequence that walks on, model suffers from downward stair.At the latter end in the 9th step, this model knocks downward first step (Fig. 7 A) with right crus of diaphragm.Model sport has been quickened in action downwards, causes bigger impact first time in whole process in the 10th step of right leg, and most extensors respond this impact (Fig. 7 C and 7D) strongerly.Have only GAS to produce less power, because knee keeps more crooked than the common state of this gait.The positive net work of ankle significantly increases (Fig. 7 B) as a result.This increase and big HFL excitation (not shown) cause that by upper torso forward inclination when it starts motion (Fig. 7 A) promotes right leg and increase step-length in recovery phase forward.The 10th step after conversion; This model keeps big step-length to move downward (the 11st and 12 step); Wherein the downward accelerated motion of this model is followed closely by GLU, HAM and VAS and is impacted the effect opposing (Fig. 7 C and 7D) that increases; Its clean positive work that reduces hip increases the clean positive work (Fig. 7 B) of knee, and stablizes this model walking speed and drop to about 1.5m/s.Get back to the level ground, lack the slowed down speed of model of downward acceleration, it has automatically reduced step-length (Fig. 7 A), and within the 13rd, 14 steps, drives it and get back to 1.3m/s and stablize walking.

For the upper and lower stair of walking, there is not independent control signal response.This model admissibility and adaptive key are its dynamic muscle reflex responses.The bounce-back of supporting leg depends on leg extensor SOL, GAS and VAS and feels how many loads, and it guarantees that leg enough causes when upwards walking, advancing forward, and significantly brakes when walking downwards.On the other hand, lead leg forward thrust along with the model dynamic change.Suddenly slow down after contralateral leg receives and impacting, upper torso turns forward, and ankle extensor speed is near support tip---leg thrust has all been given in all effects in recovery phase.These compound characteristics have guaranteed to lead leg and when going upstairs, extended enough, and are more remarkable when going downstairs.For the latter, the force feedback of GLU and HAM suppresses leg excessively rotates, and instead orders about it and regains rapidly and stretch.

Tendon units.The biological tendon units (MTU) 14 of all two foots has the structure of same pattern.According to an aspect of the present invention, Fig. 8 is the schematic diagram of tendon model.In Fig. 8, effectively contraction elements (CE) 810 forms the tendon units (MTU) of running usually together with series connection flexible member (SE) 820.If the CE810 elongation surpasses its optimization length l _CE830 (l _CE＞l _Opt840), parallelly connected flexible member (PE) 850 participates in working.On the contrary, if SE820 lost efficacy, cushion element (BE) 860 prevents the acting CE810 (l that relaxes _MTU870-l _CE830＜l _Lax880).

Be shown in Fig. 8, an effective Hill type contraction elements (CE) is according to series connection elasticity (SE) generation power.Although MTU is installed on this skeleton, so independently CE mainly acts on the limbs of rising of their power-distance relation, and said MTU model comprises parallel elasticity (PE), if the CE elongation surpasses its optimization length l _OptThen PE works.In addition, cushion (BE) is to such an extent as to guarantee so muchly to become that CE did not lose efficacy when lax when what the leg geometrical model shortened MTU.It should be noted that BE only is a digital tool that allows MTU to describe relaxed muscle, for example, the GAS that when the knee overbending, relaxes.Yet BE does not cause the outside power at MTU.

Table 3 has presented single MTU parameter.All parameters are by estimation [Yamaguchi, G.T., Sawa, A.G.-U., Moran such as Yamaguchi; D.W., Fessler, M.J., Winters; J.M., 1990.A survey of human musculotendon actuator parameters.In:Winters, J., Woo; S.-Y. (Eds.), Multiple Muscle Systems:Biomechanics and Movement Organization.Springer-Verlag, New York, pp.717-778]. maximum tension F _MaxBy supposition 25N/cm ²Power according to independently or muscle in groups-estimation of physiology cross section.Maximal velocity of contraction v _MaxMuscle at a slow speed is arranged to 6l _Opts ^-1Medium speed's muscle is arranged to 12l _Opts ^-1Optimum CE length l _OptWith SE relaxed length l _LaxMeat fiber and tendon length have been reflected.

Table 3

Are CE and the SE how details of modeling found in [Geyer, H., Seyfarth, A., Blickhan, R., 2003.Positive force feedback in bouncing gaits in the article of Geyer etc.? Proc.R.Soc.Lond.B 270,2173-2183].The power F of CE _CE=AF _Maxf _l(l _CE) f _v(V _CE), be muscle activity A, CE force-length relationship f _l(l _CE) and CE power-length velocity relation f _v(v _CE) product.Based on this product method, MTU dynamics is through integration CE speed V _CECalculate, it passes through f _v(v _CE) inverse operation set up.Given F _SE=F _CE+ F _PE-F _BE, f _v(v _CE)=(F _SE-F _PE+ F _BE)/(AF _Maxf _l(l _CE)).Work as F in the muscle elongation stage _PE-F _BENear 0 o'clock, this function had digital critical point.In order to quicken excitation, through introducing f _v(v _CE) to parallelly connected elastic force product F _PE～(l _CE-l _Opt) ²f _v(V _CE) avoid this critical point.Attention PE outside the common scope of model running works, and picture BE, and play the part of secondary part to muscular motivation in moving usually.Yet, make in this way, obtain f _v(v _CE)=(F _SE+ F _BE)/(AF _Maxf _l(l _CE)+FPE), it can be by approx with the digit time of integration progressively.Although this method is easily for acceleration model excitation, when muscular motivation is learned on the PC circuit board to fix and it also has the limit during limited temporal resolution emulation.

MTU has shared and independent parameter.Shared parameter comprises that the excitation contraction connects time constant, t _Ecc=0.01; CE force-length relationship width, w=0.56l _OptWith the residual force factor, c=0.05; The enhancing eccentric force of CE power-length velocity relation, N=1.5; Form factor, K=5; And the SE Reference Stress, ε _RefDoes=0.04 [see in Geyer H., Seyfarth, A., Blickhan, R., 2003.Positive force feedback in bouncing gaits for details? Proc.R.Soc.Lond.B 270,2173-2183].Common parameter also is PE Reference Stress ε _PE=w, wherein F _PE=F _Max(l _CE/ l _Opt-1) ²/ ε _PE ²f _v(v _CE), and the non-loaded length l of BE _Min=l _Opt-w, it is with reference to compression ε _BE=w/2, wherein F _BE=F _Max[(l _Min-l _CE)/l _Opt] ²/ ε _PE ²Independently the MTU additional parameter is obtained by document easily, is used to distinguish each muscle or muscle group.Their numerical value is listed in the table 4.

Table 4:MTU connects parameter

Muscle is connected and mass distribution with bone.Said MTU is connected to bone through striding across one or two joint.Be used variable lever arm from the transmission of muscular force Fm moment of torsion τ m to the joint Knee and ankle are carried out modeling, wherein

Be joint angle,

Be r _mAngle when reaching maximum, and

For hip, supposition simply

On the other hand, to knee and ankle, MTU length Δ l _mChange be modeled as And do to hip

With reference to angle Be joint angles, l wherein _m=l _Opt+ l _SlackFactor ρ represents muscle muscle fibre and tendon angle and guarantees that the MTU fibre length remains on physical endurance intrinsic articulation full operating range.The special parameter that is used for each muscle and joint is listed in table 4.These numerical value are perhaps by experimental evidence support [Muraoka, T., Kawakami; Y., Tachi, M.; Fukunaga; T., 2001.Muscle fiber and tendon length changes in the human vastus lateralis during slow pedaling.J.Appl.Physi01.91,2035-2040; Maganaris, C., 2001.Force.length characteristics of in vivo human skeletal muscle.Acta Physi01.Scand.172,279-285; Maganaris, C., 2003.Force-length characteristics of the in vivo human gastrocnemius muscle.Clin.Anat.16,215-223; Oda, T., Kanehisa, H., Chino; K., Kurihara, T., Nagayoshi, T.; Kato, E., Fukunaga, T., Kawakami; Y., 2005.In vivo lenth-force relationships on muscle fiver and muscle tendon complex in the tibialis anterior muscle.Int.J.Sport and Health Sciences 3,245-252], perhaps estimate to obtain through anatomy roughly.

Seven segmentations of manikin are simple rigid bodies, and its parameter is listed in the table 5.These numerical value with in other scale-model investigation, use similar, for example, G ü nther and Ruder [G ü nther; M., Ruder, H.; 2003.Synthesis of two-dimensional human walking:a test of the n model.Biol.Cybem.89,89-106].These segmentations are connected by revolute joint.In human body; These joints have free operant scope (

and

); Outside this scope, worked by mechanism's soft limiting, it orders the same the mode modeling to impact with ground.The segmentation of this model has different quality m _SAnd length l _SAnd the characteristic distance d of their local mass centres _{G, S}And joint position d _{J, S}(beginning to measure) and inertia Θ from end _S

Table 5

The ground contact and the joint limit.The sufficient segmentation of each of two sufficient models has contact point at its toe and heel.When knocking ground, contact point (CP) is by vertical reaction force F _y=-F _Reff _lf _vTo pusher, this power is similar to muscular force, is by force-length relationship f _l(Δ y _CP)=Δ y _CP/ Δ y _RefAnd power-length velocity relation f _v(dy _CP/ dt)=1-dy _CP/ dt/v _MaxProduct (Fig. 9).This product method is simulated vertical reaction force and is similar to that to describe vertical force be the existing method [Scott of spring and nonlinear spring-shock absorber sum; S.; Winter; D., 1993.Biomechanical model of the human foot:kinematics and kinetics during the stance phase of walking.J.Biomech.26 (9), 1091-1104; Gerritsen, K., van den Bogert, A., Nigg, B., 1995.Direct dynamics simulation of the impact phase in heel-toe running.J.Biomech.28 (6), 661-668; Giinther, M., Ruder, H..2003.Synthesis of two-dimensional human walking:a test of the λ-model.Biol.Cybem.89,89-106].Yet through cut spring and shock absorber, the parameter of contact model can be considered to two kinds of base substance attributes: ground rigidity k=F _Ref/ Δ y _RefWith maximum relaxation velocity v _Max, how soon it can recover its shape after having characterized ground deformation.For example, v _Max=∞ has described the perfect elasticity ground shock, and wherein ground is always pushed back by CP, v _Max=0 has described complete non-resilient impact, and wherein ground is as sand ground, and CP pushes back with the speed that descends, and can not push back with the speed that makes progress.Notice that same impulsive model is used to describe has soft limiting rigidity 0.3N m deg ^-1With maximum relaxation velocity 1deg ^-1Mechanism's soft limiting (referring to leading portion) in model joint.

Fig. 9 shows contact model according to an aspect of the present invention.In Fig. 9,, contact 910 drops to y if appearing at contact point 920 ₀Below.Vertical ground reaction force F _yBe, be similar to muscular force, be modeled as power-length (f _i) and power-speed (f _v) product of relation, Δ y wherein _RefBe F when dy/dt=0 _y=F _RefGround decrement under the condition, dy _Ref/ dt is the maximum (little figure) of the relaxation velocity on ground.At first, level ground reaction force F _XBe modeled as and use slide coefficient μ _SlF _yProportional sliding friction.In case if contact point 920 slows down 930 to being lower than minimum speed v _Lim, the level simulation is transformed into stiction 930.At 930 stages of stiction, F _XAlso simulated as the product of power-length and power-length velocity relation, itself and those in early days in order to allow about static reference points x ₀A little some is different in both direction and ground reciprocation.If F _XSurpass static limit power μ _StF _y, this simulation rotates back into sliding friction.Parameter: F _Ref=815N, Δ y _Ref=0.01mm, dy _Ref/ dt=0.03ms ^-1, Δ x _Ref=0.1m, dx _Ref/ dt=0.3ms ^-1, v _Lim=0.01ms ^-1, μ _Sl=0.8, μ _St=0.9.

In the contact process of ground, except vertical reaction force, horizontal reacting force also is used to CP.At first, this power is modeled as kinetic force of friction, the power F that itself and ground CP motion have _x=μ _SlF _yOn the contrary.When the CP speed v that slows down _LimWhen following, horizontal reacting force is modeled as stiction, calculates (Fig. 9) with the mode that is similar to the calculated level impulsive force.The power F if stiction oversteps the extreme limit _Lim=μ _StF _y, stiction is changeed back dynamic friction.Like this, depend on transient process, two types horizontal reacting force exchanges the built on stilts up to CP each other.

This result's hint, the control of mechanics and motor can not be examined closely separatedly in human motion.According to an aspect of the present invention, the neuromuscular model of human motion organizes themselves into the walking gait after initial the promotion, allow the change that ground level is unexpected, and is fit to not have the walking of the ground of interference in stair.This model permission and adaptive key are to depend on muscle reflection, and it integrates in leg muscle activity the heat transfer agent about sport dynamics.This model does not have CPG, demonstrates thus, does not have the center input to need in principle for generating walking movement, has hinted in common human motion control between the nervous system and mechanism's environment the reflection input of modulation continuously even possibly import by the replacement maincenter.

In addition, this model result hint, these continuous reflection input coding leg strengths are learned principle.Current test and scale-model investigation about the role of spinal reflex in moving process focus on its contribution [Pang for the power of the muscle of swing and the selection of time in attitude stage and generation load bearing extensor muscle; M.Y.; Yang; J.F., 2000.The initiation of the swing phase in human infant stepping:importance of hip position and leg loading.J Physiol 528 Pt 2,389-404; Dietz, V., 2002.Proprioception and locomotor disorders.Nat Rev Neurosci 3 (10), 781-790; Ivashko, D.G., Prilutski; B.I., Markin, S.N.; Chapin, J.K., Rybak; I.A., 2003.Modeling the spinal cord neural circuitry controlling cat hindlimb movement during locomotion.Neurocomputing 52-54,621-629; Yakovenko, S., Gritsenko, V., Prochazka, A., 2004.Contribution of stretch reflexes to locomotor control:a modeling study.Biol Cybem 90 (2), 146-155; Ekeberg; O.; Pearson; K., 2005.Computer simulation of stepping in the hind legs of the cat:an examination of mechanisms regulating the stance-to-swing transition.J Neurophysiol 94 (6), 4256-4268; Maufroy, C., Kimura, H., Takase, K., 2008.Towards a general neural controller for quadrupedal locomotion.Neural Netw 21 (4), 667-681; Donelan; J.M., Pearson, K.G.; 2004.Contribution of sensory feedback to ongoing ankle extensor activity during the stance phase of walking.Can J Physiol Pharmacol 82 (8-9), 589-598; Frigon, A., Rossignol, S., 2006.Experiments and models of sensorimotor interactions during locomotion.Biol Cybern 95 (6), 607-627; Grey, M.J., Nielsen, J.B., Mazzaro, N., Sinkjaer, T., 2007.Positive force feedback in human walking.J Physiol 581 (1), 99-105].Reflection has begun positive and negative dynamics [Prochazka, A., the Gillard that is fed to inherence in the kinematic system of the power of relating to for the contribution of load-bearing; D., Bennea, D.; 1997.Positive force feedback control of muscles.J.of Neurophys.77,3226-3236; Geyer, H., Seyfarth, A., Blickhan, R., 2003.Positive force feedback in bouncing gaits? Proc.R.Soc.Lond.B 270,2173-2183].Here seem the prerequisite work that the idea of leg dynamics coding principle in the motor control system systematically is not detailed.Yet; The muscle reflection that realizes in some manikin is for making that it is easy to be favourable getting into do action (trunk balance, leg swing beginning), mainly aforementioned leg dynamics and control driving phase reflection code principle; Comprise and comply with attitude leg behavior [Blickhan; R., 1989.The spring-mass model for running and hopping.J.of Biomech.22,1217-1227; McMahon, T., Cheng, G., 1990.The mechanism of running:how does stiffness couple with speed? J.of Biomech.23,65-78; Geyer, H., Seyfarth, A.; Blickhan, R., 2006.Compliant leg behaviour explains the basic dynamics of walking and running.Proc.R.Soc.Lend.B 273,2861-2867]; The stability of segmentation chain [Seyfarth, A., Giinther, M.; Blickhan, R., 2001.Stable operation of an elastic three-segmented leg.Biol.Cybern.84,365-382; G ü nther, M., Keppler, V.; Seyfarth, A., Blickhan, R.; 2004.Human leg design:optimal axial alignment under constraints.J.Math.Biol.48,623-646], and lead leg retraction [Herr, H.; McMahon, T., 2000.A trotting horse model.Int.J.Robotics Res.19,566-581; Herr, H., McMahon, T., 2001.A galloping horse model.Int.J.Robotics Res.20,26-37; Herr, H.M., Huang, G.T., McMahon, T.A., Apr 2002.A model of scale effects in mammalian quadrupedal running.J Exp Biol 205 (Pt 7), 959-967; Seyfarth; A.; Geyer; H., 2002.Natural control of spring-like running-optimized self-stabilization.In:Proceedings of the 5th international conference on climbing and walking robots.Professional Engineering Publishing Limited, pp.81-85; Seyfarth, A., Geyer, H., Herr, H.M., 2003.Swing-leg retraction:a simple control model for stable running.J.Exp.Biol.206,2547-2555].Based on these functional reflections, this model has not only been gathered the joint angles and the zmp trajectory of known human walking, has also calculated the individual muscle activity pattern of observing in some gait tests.The coupling of the muscle activity of calculating and observing has hinted that leg dynamics can play the part of important role than what infer in the past in motion control, and these principles of muscle reflection contact are to the neutral net of being responsible for motion.

In a preferred embodiment, neuromechanism model of the present invention is implemented as the muscle reflection controller to power-actuated ankle-sufficient prosthese.This embodiment is a kind of adaptability muscle reflection controller, comprises the simulation study of the Hill type muscle ankle plantar flexion of foot flesh of the positive feedback reflection with power based on utilization.The parameter of this model is mated the moment of torsion-angle distribution figure of human ankle suitably, and this distribution map is taken under the normal main body walking condition of 1m/sec weight and matched the level ground walking and measures.Use so single parameter setting, the clinical testing operation of below-knee amputee's walking under the slope condition of the slope of level ground, rising, decline implemented.In these tests, observe the adaptability of ankle prosthesis structure response ground slope change, with the mode of comparing, there is not the difficulty of tangible perception landform with normal main body condition.Significantly, the energy that provides of prosthese is directly related with the ground line gradient angle.This research has stressed that the neuromuscular controller improves its adaptive importance when the power drive prosthetic appliance passes the topographical surface of variation.

In order to process the controller with adaptive capacity, the neuromuscular model that has a power positive feedback reflection strategy is used as the part of power-actuated ankle-sufficient prosthese control system as the basis of the present invention's control.This controller that here proposes uses ankle-sufficient composite model to confirm physics Torque Control ankle-joint.In this model, ankle-joint has two virtual actuators.For sufficient sole of the foot bending moment, this actuator is the Hill type muscle that has positive force feedback reflection strategy.This policy-simulative is because the muscle reflex response of some composite signal of importing into from muscle-spindle and Golgi tendon organ.For dorsiflex moment, impedance is provided by virtual rotation spring cushion.

The parameter of this neuromuscular model through the optimizing process match or optimize suitable, to provide ankle under normal person's target velocity 1.0m/sec walking condition to measure moment and to be given the optimum Match between the output torque of this model of the motion input under the normal person's situation that measures.Prosthese controller based on the neuromuscular model is used to provide moment to control power ankle-sufficient prosthese that the amputee dresses.This control strategy uses two criterion evaluations.The first, controller generates prosthese ankle moment and ankle angle distribution and comparison other, the test of normal person's these index consistency matching capacities under target level ground running velocity conditions.Second performance standard be controller appear do not change controller parameter and increase the ability of the biological consistent trend of net work in the gait cycle along with the increase of the walking gradient.It is difficult using typical sensors to detect the ground line gradient variable, so the ability of intrinsic these changes of adaptation of controller is valuable especially.

Figure 10 A-C shows the physical system (Figure 10 A) of the ankle-sufficient prosthese exemplary embodiment that is used for a preferred embodiment, the figure of drive chain (Figure 10 B), and mechanical model (Figure 10 C).Said ankle-sufficient the prosthese that is used for this research is iWalk, a prosthese in the LLC exploitation.This prosthese is the succession of a series of prototypes of having developed of the biological electromechanical integration of MIT Media Lab seminar; It is described in U.S. Patent application No.12/157; 727, be filed on June 12nd, 2008, its disclosed full content integral body is included in here as a reference.This prosthese is a complete independent device, has the weight (1.8kg) and the size of normal biological ankle-foot complex.Referring to Figure 10 A, can see the housing 1005 of motor, transmission device and electronic equipment, ankle-joint 1010, foot 1015, unidirectional parallelly connected leaf spring 1020 and series connection leaf spring 1025.Figure 10 B shows and is with 1030 synchronously, articulated joint main shell 1035, motor 1040, ball-screw 1045, ankle-joint 1010, feed screw nut 1050 articulated joints (serial spring) 1055, and sufficient motion indicator 1060.Basic connecting rod 1065 has been shown, motor 1040, transmission device 1070, serial spring 1025, unidirectional parallelly connected spring 1020, foot 1015, serial spring transmission arm r in the mechanical model of Figure 10 C _S1075, length 1080 that spring is non-loaded and SEA 1085.For clarity sake, the spinner member in this physical arrangement is shown as linear equivalent in illustraton of model.

Ankle-joint is rolling bearing (rotation is supported) design, and the foot structure that connects the bottom is to the upper leg bar linkage structure, and its top has the connecting portion that prosthese pyramid fixture is connected to the amputee.Foot comprises passive low profile Flex-Foot ^TM(Osur ^TM), to minimize ground contact impact to the amputee.Unidirectional leaf spring, parallelly connected spring action be in ankle-joint, when ankle and foot are vertical each other, works.It acts on the power drive chain abreast, and the passive functions of heel string is provided.The power drive chain is for the motor-driven link through ankle-joint, shown in Figure 10 B.Terminal from the leg connecting rod on top, its successively by the brushless electric machine of 24V operation (Powermax EC-30,200Watt, 48V, Maxon), have the band drive transmission of 40/15 speed reducing ratio and the linear ball-screw of 3mm pitch is formed.At this working voltage, this motor is approximately 340Nm through the theoretical maximum moment that this drive chain can produce.

At sufficient place, serial spring, the compound leaf spring of Kevlar connect foot to the ball nut, have the arm of force r relevant with direction _STherefore; Effective rotational stiffness of this serial spring estimates that through the lock drive chain with to the ankle-joint applied moment size is positive moment 533Nm/rad, negative moment 1200Nm/rad; Wherein positive moment (or plantar flexion of foot moment of muscle power) trends towards compressing serial spring, shown in Figure 10 C.Drive chain and serial spring comprise series connection resilient device (SEA) [G.A.Prart and M.M.Williamson together; " Series elastic actuators; " Proceedings on IEEE/RSJ International Conference on Intelligent Robots and Systems; Pittsburgh, pp.399-406,1995].The layout drawing of these elements is illustrated in Figure 10 C.

Sensor.The Hall effect angular transducer of ankle is elementary control input, has the scope of-0.19 to 0.19 radian, and wherein 0 corresponding foot is perpendicular to shin bone.Joint angles is calculated through the linear hall effect sensor (Allegro A 1395) that is installed on the main casing.This sensor is near being rigidly connected to the structural magnet of foot, and magnetic axis is the tangent line of magnet movement radian like this.As the result of this layout, along with the magnet rotating tee is crossed sensor, the sensing station magnetic field intensity changes.Deformeter is positioned at prosthese tapering connecting portion, allows the moment estimation of ankle-joint.Deformeter is positioned on the serial spring, allows the moment output of perception motorization drive chain, therefore, allows the closed loop power control of SEA.Motor itself comprises Hall effect rectification sensor, and adapts to the optical shaft encoder that can use advanced brushless motor control device.

Microcontroller.The overall control of ankle-sufficient prosthese and communication are by 16 monolithic DSP microcontrollers, and Microchip Technology Incorporated dsPIC33FJ128MC706 provides.4,000 ten thousand instructions of this microcontroller per second computing have 128K byte program flash memory, the 16384RAM byte.Control provides enough computing capabilitys in order to support in real time for it.

Electric Machine Control.16 dsPIC33FJ128MC706 of per second are used to the motor special controller.The high-speed computation requirement of the heavy load of modern brushless motor control strategy has promoted this architecture development together with the independent task of host microcontroller real-time command.In fact high-speed figure connection between host microcontroller and the motor microcontroller is provided by the real-time command of motor.

Wave point.In order to develop and collect data, the high speed serial port of microcontroller offers external communication specially and uses.This port can be used to directly maybe can have multiple wireless telecommunications system through cable and connect.To current research, 500Hz sensor and internal state information are passed through serial port with 460 kilobaud speed remote measurements, and transmit through IEEE 802.11 Wireless LAN devices (Lantronix Wiport).

Battery.Prosthese dynamic lithium polymer battery by 0.22kg provide, this battery has 165 watt-hours of/kilogram energy densities.This battery can provide the power walking of one day 5000 step required power.

Best mechanical elements are selected.The requirement of satisfying quality, size, moment, speed, energy efficiency, impact tolerance level and approaching silent running is not common task.It it is especially important the structure and optimization drive chain that produces biological moment and walking movement.Some effect of selection of Motor, whole gearratio, serial spring and parallelly connected spring is described in S.K.Au; H.Herr; " " IEEE Robotics&Automation Magazine.pp.52-59, September 2008 for On the Design of a Powered Ankle-Foot Prosthesis:The Importance of Series and Parallel Elasticity..

The control structure system.The purpose of control structure system is the ankle moment of control by the definite suitable amputee's gait cycle of the reversible transducer information of prosthese ankle state.Use the neuromuscular model of human ankle-foot complex, controller is confirmed suitable moment.In this model; Hinge joint; Represent human ankle-joint; Activate by two virtual actuators of vying each other: the unidirectional plantar flexion of foot flesh of a Hill muscle model, and one depend on gait phase or as twocouese proportion differential positioner or as the ankle flexor lift ridge musculus flexor of unidirectional virtual rotation spring-shock absorber.Limited state machine keeps amputee's gait phase to estimate.Depend on the gait phase of estimation, one of virtual brake or another or two Virtual Controllers produce moment in ankle.Along with ankle torque command prosthese hardware, so virtual net torque is used.Actual moment at ankle-joint is produced by motorization drive chain and parallelly connected spring.The ankle angular transducer is used to confirm the moment by parallelly connected spring generation, and remaining expectation moment is passed through motor controller controls.

Top state machine control.Through a limited state machine that gait cycle is synchronous, the top State Control of prosthese is implemented.In gait processes, need the identification two states: recovery phase and driving phase.In order to confirm state-transition, prosthese sensor input (by the ankle moment of pyramid deformeter, ankle angle and motor speed estimation) is to observe continuously.The situation of these state conversions is confirmed through test.Figure 11 shows the operation and the switch condition of state machine.Ankle flexor lift ridge musculus flexor and the virtual actuator of plantar flexion of foot flesh depend on the gait state estimation of estimating from state machine and produce moment.

In Figure 11, recovery phase 1110 is illustrated as SW1120, and driving phase 1130 is divided into crooked (CP) 1140 of the controlled sufficient sole of the foot, controlled dorsiflex (CD) 1150, and crooked (PP) 1160 of the power foot sole of the foot.State conversion 1170,1180 is confirmed through prosthese ankle moment Tp and the prosthese ankle angle θ from the cone strain-ga(u)ge measurement.

When foot's built on stilts, to the conversion of recovery phase through the cone strain-ga(u)ge measurement to the ankle resultant couple drop to the ankle angle θ that is lower than 5Nm or measures drop to be lower than-0.19rad is detected, and is saturated to prevent angular transducer.Trend towards the crooked ankle of the sufficient sole of the foot, positive-angle is corresponding to dorsiflex, and positive moment is defined as actuator torque.In order to prevent too early state conversion, having proposed angle moment at driving phase must could begin above these conversions of 20Nm.In addition, the buffer time of 200ms provides driving phase minimum time framework.Based on heel knock to the conversion of driving phase through use the taper strain-ga(u)ge measurement obtain moment reduce to be lower than-7Nm is detected.

The block diagram of the exemplary embodiment of the control system of ankle-sufficient prosthese is illustrated in Figure 12 according to an aspect of the present invention.Neuromuscular model 12010 has been shown, parallelly connected spring model 1220, advancer 1230, friciton compensation device 1240, electric machine controller 1250 and prosthese 1260 (showing mechanical model) according to Figure 10 C in Figure 12.

Prosthese is measured the ankle state

Be used to generate torque command τ from the neuromuscular model _dThis expectation ankle moment is carried through the Torque Control system to obtain the current-order to the prosthese actuator.Three major parts of this Torque Control system are feed-forward gain K _Ff, advancer and friciton compensation item.The parallel connection spring action is in prosthese ankle moment τ _p, deduct from the ankle moment of expectation, to obtain the actuator torque τ of expectation _{D, SEA}The actuator torque τ that this moment, the closed loop torque controller measured through use _SEAStrengthen the actuator torque of expectation.At last, the friciton compensation device produces additional moment value τ _f, it is added in the output of closed loop Torque Control.

Figure 13 A-C is the curve map of prosthese moment in the next complete gait cycle of three kinds of situation (heel strike of same pin is to the heel heel strike): level ground (Figure 13 A), acclivity (Figure 13 B), decline slope (Figure 13 C).Shown in each is to control moment mean value 1305,1310,1315 (fine rule) ± standard deviations (short setting-out), prosthese moment, as use the SEA moment loading measured and based on the parallelly connected spring torque effect 1320,1325,1330 (thick line) that angle is estimated estimate.Vertically (strokes and dots) line 1335,1340,1345 is indicated the latter end of driving phases.

Ankle flexor lift ridge musculus flexor model.Figure 14 A-C shows the flesh skeleton model in prosthese controller embodiment realizes; Comprise Hill type muscle model (Figure 14 B) and be connected to the two-way spring buffer that is connected ankle-joint model (Figure 14 A); The geometry of the muscle model that adheres on the skeleton (Figure 14 C) comprises the realization and the angle coordinate framework of the variable arm of force of muscle model.Shown in Figure 14 A and the 14C is ankle flexor lift ridge musculus flexor (spring buffer) 1405, plane musculus flexor (MTC) 1410, and foot 1415, trunk 1420, and the mechanics of heel 1425 is expressed.

The ankle flexor lift ridge musculus flexor of Figure 14 A is an ankle flexor lift ridge musculus flexor actuator.It represents tibialis anterior and other biological ankle flexor lift ridge musculus flexor.This ankle flexor lift ridge musculus flexor is implemented as virtual rotation spring buffer, has adjustment point

and relation:

$T_{\mathrm{DORSI}}=K_P\mathrm{\&theta;}+K_V\stackrel{\&CenterDot;}{\mathrm{\&theta;}}---\left(1\right)$

Here, K _pBe spring constant, K _vBe damped coefficient, θ is the ankle angle, Be ankle angular speed.For driving phase, K _pValue optimized with other muscle model parameters, to reach and the biological ankle driving phase behavior optimum Match of on the light water plane earth, walking.Damping term K _vDriving phase is coordinated 5Nm-s/rad through test, to prevent forward foot in a step bounce-back built on stilts when foot is set level.At driving phase, ankle flexor lift ridge musculus flexor is only as providing ankle flexor lift ridge musculus flexor moment, with the unidirectional characteristic for mimic biology muscle equally.In addition, become vertical the result with body as foot, when the moment that is produced by ankle flexor lift ridge musculus flexor at driving phase drops to zero, inoperative at the other times ankle flexor lift ridge musculus flexor of driving phase.Therefore; Ankle flexor lift ridge musculus flexor only contributes to driving phase and generates moment in early days, this moment human ankle flexor lift ridge musculus flexor to play an important role be known [J.Perry, Gait Analysis:Normal and Pathological Function; New Jersey:SLACK Inc.; 1992, Chapter 4, pp.55-57].In recovery phase, ankle flexor lift ridge musculus flexor drives foot to the adjustment point as positioner

For this purpose, gain K _p=220Nm/rad and buffering constant K _v=7Nm/rad offers the quick ground clearance of early stage foot recovery phase.

The plantar flexion of foot muscle model.The virtual plantar flexion of foot flesh of Figure 14 A-C comprises muscle-tendon complex (MTC), and it represents the complex of human plantar flexion of foot flesh muscle.MTC is based on S.K.Au, J.Weber, and H.Herr; " Biomechanical design of a powered ankle-foot prosthesis, " Proc.IEEE Int.Conf.On Rehabilitation Robotics, Noordwijk; The Netherlands, pp.298-303, June 2007; Wherein, it is discussed in more detail.It is made up of the telescopic element (CE) of simulate muscular fiber and the series element (SE) of simulation tendon.Contraction elements is made up of three unidirectional elements: have the Hill type muscle of power positive feedback reflection strategy, parallelly connected elastomer and Min. or buffer parallel connection elastomer to greatest extent.What connect with contraction elements is series element, and it is nonlinear a, unidirectional spring, represents heel string.The additional geometry of the muscle on the ankle-joint model-tendon complex is non-linear, makes the calculating of the moment that is produced by driving force become complicated.

Plantar flexion of foot flesh series connection flexible member.This series connection flexible member (SE) conduct with as [H.Geyer, A.Seyfarth, R.Blickhan; " Positive force feedback in bouncing gaits?; " Proc.R Society.Lond.B 270, pp.2173-2183,2003] the tendon action of contraction of muscle element connected in series.With ε is tendon stress, is defined as:

L wherein _SEBe the length of series element, and l _LaxBe its non-loaded length, series element is designated as H.Geyer, A.Seyfarth; R.Blickhan, " Positive force feedback in bouncing gaits?, " Proc.R Society.Lond.B 270; Pp.2173-2183, the nonlinear spring of describing in 2003:

$F_{\mathrm{SE}}=\left\{\begin{array}{cc}{F}_{\max }{\left(\frac{\mathrm{\&epsiv;}}{{\mathrm{\&epsiv;}}_{\mathrm{ref}}}\right)}^2,& \mathrm{\&epsiv;}>0\\ 0,& \mathrm{\&epsiv;}\&le;0\end{array}\right.---\left(3\right)$

Wherein, F _MaxIt is the maximum tension that muscle can apply.Following H.Geyer; A.Seyfarth, R.Blickhan, " Positive force feedback in bouncing gaits '?; " Proc.R Society.Lond.B 270; Pp.2173-2183,2003, this quadratic form is used to approach the segmentation index of general models--linear tendon stiffness curve.This number that is used for reducing model parameter that approaches.

Plantar flexion of foot flesh contraction elements.The contraction elements (CE) of the virtual actuator of plantar flexion of foot flesh, Figure 14 B is the Hill type muscle model with positive force feedback reflection strategy.It comprises the fiber that flexes a muscle of generative power; Flexible member with two parallel connections; As H.Geyer; H.Herr is described in " A muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities, " (submission is waited to publish). and this Hill type meat fiber applies unidirectional force.This power is the meat fiber length l _CE, speed v _CEFunction with muscle activity A.F makes a concerted effort _MFBe, as H.Geyer, A.Seyfarth, R.Blickhan. " Positive force feedback in bouncing gaits?, " Proc.R Society.Lond.B 270, pp.2173-2183 provides in 2003:

F _MF(l _CE，v _CE，A)＝F _maxf _L(l _CE)f _V(v _CE)A (4)

This force-length relationship of Hill type muscle, f _L(l _CE), be the mitriform curve, provide by following formula:

$f_L\left(l_{\mathrm{CE}}\right)=\exp \left(c{\left|\frac{l_{\mathrm{CE}}-l_{\mathrm{opt}}}{l_{\mathrm{opt}}w}\right|}^3\right)---\left(5\right)$

Wherein, l _OptIt is the contraction elements length l _CEThis moment, muscle can provide maximum tension force (waiting a power), F when this length _MaxParameter w is the width of mitriform curve, and parameter c has been described the subapical magnitude of mitriform curve, wherein:

f _L(l _CE＝(1+w)l _opt)＝exp(c) (6)

This power-length velocity relation of CE, fv (v _CE) be the Hill equation:

$f_v\left(v_{\mathrm{CE}}\right)=\left\{\begin{array}{cc}\frac{v_{\max }-v_{\mathrm{CE}}}{v_{\max }+K{v}_{\mathrm{CE}}},& v_{\mathrm{CE}}<0\\ N+(N-1)\frac{v_{\max }+v_{\mathrm{CE}}}{7.56K{v}_{\mathrm{CE}}-v_{\max }},& v_{\mathrm{CE}}\&GreaterEqual;0\end{array}\right.,---\left(7\right)$

Wherein, V _MaxThe＜0th, the maximal velocity of contraction of muscle, V _CEBe filament contraction speed, K is a bending Constant, and N is defined as nondimensional muscular force (by F _MaxNormalization) like this,

N＝f _v(v _CE＝-v _max) (8)

According to H.Geyer; H.Herr; " A muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities, " (submission is waited to publish), the said force-length relationship that is used for parallelly connected to greatest extent elastomer (HPE); Be set to parallelly connectedly, provide by following formula with CE:

Comprise that also lower bound cushions parallelly connected elasticity (LPE); Based on H.Geyer; H.Herr. " A muscle.reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities, " (submission is waited to publish).This is provided by the form with nonlinear spring:

So total plantar flexion of foot muscular strength is described to:

F _CE＝F _MF(l _CE，v _CE，A)+F _HPE-F _PE (11)

F wherein _CEBe the power that produces by contraction elements.Because CE and SE connect, following equation keeps: F _CE=F _SE=F _MTC

The reflection strategy.The contraction elements activity; A; Use shown in Figure 15; As [H.Geyer, H.Hellr.ccA muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities, " (submission is waited to publish); H.Geyer.A.Seyfarth, R.Blickhan, " Positive force feedback in bouncing gaits?, " Proc.R Society.Lond.B 270, pp.2173-2183,2003] described in positive feedback power reflection strategy generate.Figure 15 shows an exemplary embodiment of virtual plantar flexion of foot flesh being used the reflection strategy, comprises the relation between the plantar flexion of foot pars muscularis spare of ankle angle, muscular force and ankle moment.

Shown in figure 15, this feedback loop comprises the driving phase switch, is used to make do not produce the plantar flexion of foot muscular strength in recovery phase.At driving phase, plantar flexion of foot muscular strength, F _MTC, multiply by reflection gain Gain _RF, by Delay _RFPostpone, and be added to the skew excitation, PRESTIM is to obtain the neural activation signal.This excitation is limited in the scope of 0-1, makes free constant T LPF, is used for simulating or the emulation muscular irritation--shrink coupling.The signal that obtains is used as the actuating in the equation (4) with initial value PreA.In addition, inhibition gain G ain _SUPPAccording to H.Geyer; H.Herr, " A muscle ' reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities, " (submission is waited to publish); Be implemented, fight each other at driving phase to help prevent two actuators.Here, the moment that is produced by ankle flexor lift ridge musculus flexor is passed through Gain _SUPPF _MTCReduce or drop to zero up to its value.

Plantar flexion of foot caryolytes structure and realization.In the muscle model framework, ankle angle θ _Foot, be defined as shown in Figure 14 C.Use of the input of this angle as model; The length quilt of muscle-tendon complex is like H.Geyer; H.Herr; Such calculating in " A muscle.reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities, " (submission is waited to publish):

l _MTC=r _Footρ (sin (φ _Ref-φ _Max)-sin (θ _Foot-φ _Max))+l _Lax+ l _Opt(12)

Wherein ρ represents the middle muscle fibre of meat fiber and the scale factor of tendon angulation, φ _RefBe l when non-loaded _CE=l _OptAnkle angle under the condition.

Fibre length l _CECan be used l _CE=l _MTC-l _SECalculate, wherein l _SEFrom learning at given existing value F by muscular motivation _CE=F _SE=F _MTCObtain through the inverse operation of (3) under the condition.The filament contraction speed v _CECan obtain through differential.Produced the differential equation of first order of controlling by neuromuscular model dynamics like this.Given θ _FootTime dependent curve and primary condition, this equation can be found the solution and obtain F _MTCYet because integration is more stable than differential calculation, the integrated form of this embodiment is used to find the solution F _MTCAs H.Geyer; H.Herr is as described in " A muscle.reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities, " (submission is waited to publish).

Given connection radius r _FootWith angle φ _Max, under this condition, will realize maximum muscle-tendon arm of force, F _MTCWith the consequent plantar flexion of foot flesh effect T that influences ankle moment _{The foot sole of the foot}Between relation provide through following formula:

T _{The foot sole of the foot}=F _MTCCos (θ _Foot-φ _Max) r _Foot=F _TMCR (θ _Foot) (13)

Wherein, R (θ _Foot) be the variable arm of force that obtains by the muscle that is connected to the ankle-joint model.This relation is illustrated by Fig. 6.Therefore, the plantar flexion of foot muscle model finally is used as the single input of coupling θ _FootWith single output T _{The foot sole of the foot}Dynamic system handle.

The neuromuscular model parameter is confirmed.The plantar flexion of foot muscle model is the lump representative of all biology plantar flexion of foot fleshes.Likewise, ankle flexor lift ridge musculus flexor has been represented all biology ankle flexor lift ridge musculus flexors.In this work, only carry out joint and torgue measurement in ankle.As a result, multi-joint muscle state like gastrocnemius, can not accurately be estimated.Therefore, plantar flexion of foot flesh is based on the simple joint plantar flexion of foot flesh that human body plays a major role, musculus soleus.Therefore; Most plantar flexion of foot flesh parameter values are to be reported in H.Geyer; H.Herr; Be applied to those of musculus soleus in " A muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities, " (submission is waited to publish).Yet some plantar flexion of foot flesh parameter, those parameters of ankle flexor lift ridge musculus flexor also are supposed to or significantly by the lumped model influence, perhaps are not that biology is known.These six parameters use the composite algorism of genetic algorithms and gradient descent algorithm to make it suitable, make walking data optimum Match under neuromuscular model and the normal person's condition.

The optimum parameters value is not shown in table 6.

Table 6

l opt[m]	0.04	w	0.56
				l Lax[m]	0.26	c	ln(0.05)
V max[lopt/S]	6.0	N	1.5
				ε ref	0.04	K	5
PreA	0.01	ρ	0.5

T[s]	0.01	?r Foot[m]	0.05
				PreSTIM	0.01	Postpone RF[s]	0.02

Not collected by the people's that amputation influences data.The kinematics of walking data and dynamic (dynamical) data are at Gait Laboratory of Spaulding Rehabilitation Hospital; Be collected [H.Herr in the research of Harvard Medical School by Spaulding committee on the Use of Humans as Experimental Subjects approval; M.Popovic; " Angular momentum in human walking, " The Journal of Experimental Biology, Vol.211; Pp 487-481,2008].After being apprised of and obtaining to agree, a healthy adult male (81.9kg) is asked to pass 10m pavement in the motion-captured laboratory with slower speed walking.

Use has VICON 512 motion capture system of eight thermal cameras and carries out motion-captured.In order in experiment, to allow thermal camera to follow the tracks of the specific region, reflective marker is set at 33 specific regions of main body health.Video camera is taken with 120Hz, can be in the scope of about 1mm the sign of tracing preset.In order to follow the tracks of the lower part of the body; Said sign is set at following bone mark: bilateral anterior superior spine (bilateral anterior superior iliac spines); Posterior superior iliac spine (posterior superior iliac spines); Thigh lateral condyle (lateral femoral condyles), external malleolus point (lateral malleoli), preceding toe (forefeet) and heel (heels).The identification pen is set on shin bone and the thigh, and said mark is attached to the identification pen at shin bone middle part and the axle middle part more than the thigh middle part.Mark also is set at the following point of upper torso: breastbone, clavicle, the shoulder of C7 and T10 vertebra, head and bilateral, elbow, oar vertebral articulations.

Ground reaction force use 2 staggered power plates (model no.2222 or OR6-5-1, Advanced Mechanical Technology Inc., Wateno, IV n, MA USA) measures, and it is combined on the pavement by one.The precision of measuring these power plates of ground reaction force and Center of Pressure is approximately respectively 0.1N and 2mm.Power plate data are collected with 1080Hz, and are consistent with the VICON movement capturing data.Joint moment is got by the amendment scheme calculating of ground reaction force and articular kinesiology application standard inverse dynamics model.Vicon Bodybuilder, Oxford Metrics, UK are used to accomplish inverse kinematics and calculate.

Obtained six tests of (average 1.0m/s) walking at a slow speed on the level ground, once independent test is used to represent target ankle and the zmp trajectory under this walking state.The driving phase latter end is defined as the time point when in joint moment is reaching gait cycle, dropping to zero for the first time after the maximum moment.To optional test, this incident occurred in for 67% gait cycle time.

Figure 16 A and 16B show the moment and the angle track of the prosthese that amputee's main body measures in the experiment, and with these data of the weight and the biological ankle of height and the complete main body of limbs coupling relatively.Shown in Figure 16 A and the 16B is that heel at same pin knocks (0% cycle) and knocks the cycle ankle moment (Figure 16 A) and the ankle angle (Figure 16 B) of the level ground walking gait in (100% cycle) to heel.The curve of Figure 16 A and Figure 16 B is the prosthese moment that measures that causes of neuromuscular model control and average 1610,1620 (the fine rule) ± standard deviations (lacking setting-out) of angle graphics, weight and matched with the complete main body of limbs with the ankle biomethanics under the gait cycle situation of same walking speed (1m/sec) 1630,1640 (thick line).Vertical curve indication be that the prosthese gait cycle begins recovery phase 1650,1660 (choice refreshments setting-out) and begin recovery phase 1670,1680 with biological ankle the average time of (slightly dotted line).

Make model parameter adequacy test data through optimization.Following parameter is selected adjustment: F _Max, Gain _FB, Gain _SUPP, φ _Ref, and φ _MaxThe target of parameter adjustment is to find out, and under specific walking condition, given corresponding biological ankle track is imported as model, can be so that the parameter setting of the biological ankle zmp trajectory of neuromuscular model optimum Match.The evaluation function of optimizing is decided to be at the given biological ankle angle track of driving phase, the variance of biology and model ankle figure, just:

$\mathrm{Cost}=\underset{t\&Element;\mathrm{STANCE}}{\mathrm{\&Sigma;}}{(T_m\left(t\right)-T_{\mathrm{bio}}\left(t\right))}^2---\left(14\right)$

Wherein, T _mBe the moment output of model, and T _BioBe biological ankle moment.

Genetic Optimization Algorithm is selected the initial ranging of implementing the optimized parameter value, and direct search also is introduced into searches the optimized parameter setting.Matlab genetic algorithm instrument is used to implement this two kinds of optimization methods.Be used to provide the reference behavior of optimization with the level ground human walking data of the 1.0m/s walking speed selected.The scope that each parameters optimization is allowed is shown in table 7.

Table 7: parameters optimization scope

Parameter (unit)	Minimum of a value	Maximum
			F max(N)	3000	7000
Gain FB	0.6	1.5
			K P(N·m/rad)	20	250
Gain SUPP	0	5
			φ ref(rad)	0.52	2.09
φ max(rad)	1.40	2.44

Initial data crowd is selected by optimizer.The parameter value that parameter optimisation procedure draws is illustrated in table 8.

Table 8: the neuromuscular model parameter value of match

F max(N)	3377
		Gain FB	1.22
K P(N·m/rad)	72.9
		Gain SUPP	0
φ ref(rad)	1.49
		φ max(rad)	1.95

Parameter optimization result.As the checking optimization effect, use biological ankle angle distribution to be optimized and obtain final argument as the input of stretching the neuromuscular model.Moment distribution figure result who obtains and biological moment distribution figure relatively are illustrated in Figure 17.

Shown in figure 17; The ankle moment distribution figure of complete biological ankle and biological ankle angle distribution figure are as comparison input and the ankle moment distribution figure neuromuscular model of the parameter value of optimizing, and the parameter value of optimizing is respectively: plantar flexion of foot sarcous element (fine rule) 1730 and whole neuromuscular model (plantar flexion of foot flesh and ankle flexor lift ridge musculus flexor) moment (dotted line) 1740 that the ankle flexor lift ridge musculus flexor composition (dotted line) 1720 that biological ankle moment (grey lines) 1710, quilt are simulated, quilt are simulated.To the gait cycle of the overwhelming majority, the almost strict coupling of neuromuscular model moment and biological ankle moment.

Elementary Torque Control.The actual moment that produces in the driving phase ankle is from the acting in conjunction of parallelly connected spring and motorization drive chain.Rotating parallelly connected spring rate is approximately linear in the scope of application, and its spring rate is 500Nm/rad.Use this spring constant, the effect of parallelly connected spring is contemplated to from the ankle moment of expectation and deducts.Excess torque must be produced by the motorization drive chain.

The performance of motorization drive chain is improved through using advancer, friciton compensation device and feed-forward technology, and is shown in figure 12.Open loop drive chain dynamic test research is unfolded and is used to implement these improvement [M.Eilenberg; " A Neuromuscular.Model Based Control Strategy for Powered Ankle-Foot Prostheses; " Master ' s Thesis.Massachusetts Institute of Technology; Cambridge, MA, 2009].Different Figure 13 A-C that are illustrated in of output torque and demand moment during walking on level ground, acclivity and the decline slope.The prosthese output torque uses the deformeter on the serial spring to be estimated to the effect of SEA moment, and estimates the effect of parallelly connected spring torque based on the ankle angle of parallelly connected spring torque.

Clinical assessment.Prosthese is installed on below-knee amputee's the right leg of health, active, 75kg.This main body was allowed in the up time of walking to carry out the nature adjustment of prosthese.The wireless connections of prosthese are used to write down the walking data from these tests.In the gait test of level ground, this main body is asked to the long road of walking 10m.The target setting walking speed is 1.0m/s, with the walking speed of matching complete main body.This main body begins walking from the place apart from the about 5m of path starting point, and the about 3m of the path termination of passing by stops walking.The mark on ground is used to point out the starting point and the terminal point in 10 meters paths.Use stopwatch to confirm the average walking speed of each test through the time of each mark through the mass centre of record main body.10 tests have been carried out altogether.The test of walking speed in 5% target velocity is used to handle, and obtains 45 gait cycles.This main body is asked to self-selected speed walking 11 degree, the long slope of 2m again.This main body begins apart from the place of the about 2m of slope starting point on the level ground, stops on the platform of 1m after approximately through the slope, has carried out the test of climbing of 10 slopes.Same then path quilt is through carried out the slope dropping test 12 times on the contrary.

Data analysis.First three of level ground test is assumed that with back three gait cycles and makes transient process, therefore be left in the basket.Remaining each gait cycle is generated 1000 data points by resampling.Draw mean value and standard deviation track by these data computation that obtain.To rising on the slope and two kinds of situation of slope decline, final step is used to represent gait cycle on the slope.To test identical mode with the level ground of describing, each selecteed gait cycle is resampled and is averaged.

Independently gait cycle is liftoff about ankle angle digital integration ankle Calculating Torque during Rotary net work through knock toe from heel to each.Here, to the calculating of net work, be left in the basket recovery phase.Calculate the net work mean value under every kind of walking condition by gait cycle net work value independently then.

The result.Moment is followed the tracks of.The prerequisite of current test is the ability of the speed that can be controlled by the neuromuscular controller of the actual generation of ankle-sufficient prosthese moment.This ability is showed in Figure 13 A-C, illustrates rising on level ground walking, the slope and slope decline, and the output torque of controlling moment and measurement is opposite.

The adaptability of ground line gradient.The prosthese of neuromuscular model control is proved by the clinical testing data of Fig. 9 a-9c the ground line gradient Adaptability Evaluation.The digital integration data of these tests provide following net work value (building ring area):

5.4 ± 0.5 joules of level grounds

12.5 ± 0.6 joules of acclivities

0.1 ± 1.7 joule on decline slope

Comparison with biological ankle.The purpose of this neuromuscular model is to reproduce the intrinsic dynamics of human ankle-foot complex with useful mode.Therefore, based on the ability of its mimic human behavior, people can estimate the prosthese controller that obtains.At Figure 16 of preceding discussion A and 16B, show with weight and matched and have the level ground walking moment and the angle distribution figure of the main body prosthese together of complete limbs.

Figure 18 A-C is the prosthese moment-angle geometric locus of the prosthese that measures under level ground (Figure 18 A), acclivity (Figure 18 B), decline slope (Figure 18 C) these three kinds of situation.Shown in Figure 18 A-C is 1810,1820,1830 ± one standard deviations of average.The arrow indication propagated forward time.Along with ground line gradient increases, average prosthese net work increases.Human ankle data consistent [A.S.McIntosh, K.T.Beatty, L.N.Dwan in this result and the document; And D.R.Vickers, " Gait dynamics on an inclined walkway, " Joumal of Biomechanics; Vol.39, pp 2491-2502,2006].

The ankle moment of the prosthese that measures and ankle angle distribution are mated as walking ankle moment and the ankle angle distribution of complete individuality relatively in the level ground preferably.The observed low level that is not both, it can be suitably owing to o lot of reasons, comprise since clinical main body because perhaps leg muscle atrophy that amputation causes and/or hypertrophy, limbs length difference be the functional deficiencies of two joint gastrocnemiuses.In addition, the limited range of prosthese angular transducer stops prosthese to arrive the gamut of complete ankle motion.

Ground line gradient adaptability.The neuromuscular control that here proposes has been showed and need not perception landform significantly and to the inherent adaptability of ground line gradient.According to data from [A.S.McIntosh, K.T.Beatty, L.N.Dwan, and D.R.Vickers, " Gait dynamics on an inclined walkway." Journal of Biomechanics; Vol.39, pp 2491-2502,2006] ankle net work that in the uphill process of ramp, increases and the ankle net work that in the process of decline ramp, reduces; according to the ankle net work of level ground walking, consistent with the behavior under the same conditions of normal human subject ankle.This variation of the positive net work of driving phase demonstrates the slope adaptability performance that the neuromuscular model produces under the walking condition.The ability that the neuromuscular model produces these bionical variation behaviors has hinted that model has embodied the key character of human plantar flexion of foot flesh.In addition, model is had the velocity adaptive begetting power by expection.In the trial of fast moving, wearer possibly more enforced prosthese.This additional power can cause that the model reflection controls more virtual muscular force, causes bigger energy output, therefore improves walking speed.

Although disclose preferred embodiment, those skilled in the art expect many other embodiments that do not break away from the scope of the invention easily.Each of above-mentioned various embodiment can combine other embodiment that describe that various characteristic is provided.In addition, although the apparatus and method of a large amount of independent embodiment among the present invention have been described in the front, described herein only is the example that the principle of the invention is used.Other arrangements that therefore those of ordinary skills expect, method, modification and replacement will fall into scope of the present invention, and it is not limited to appended claim.

Claims (9)

1. neural mechanical control device that is used to comprise the mechanical limb at least one joint based on model, this controller comprises:

Finite state machine, this finite state machine are configured to receive the feedback data that the feedback data that relates to the mechanical limb state and utilization receive and confirm the mechanical limb state;

The muscle model processor; This muscle model processor is configured to receive from the status information of finite state machine with from the muscle geometry and the reflector architecture information of at least one database, and utilizes the neuromuscular model to confirm to be sent to the joint moment or the rigidity instruction of at least one expectation of mechanical limb; And

Joint instruction processing unit, this joint instruction processing unit are configured to control the definite bionical moment and the rigidity of muscle model processor of mechanical limb joint.

2. controller as claimed in claim 1 is characterized in that, at least one sensor of each joint of feedback data through being installed in mechanical limb provides.

3. controller as claimed in claim 1 is characterized in that mechanical limb is a leg, and finite state machine is synchronized with the leg gait cycle.

4. controller as claimed in claim 3 is characterized in that, leg comprises power ankle-sufficient prosthese.

5. controller as claimed in claim 3 is characterized in that leg comprises knee joint.

6. controller as claimed in claim 4 is characterized in that leg also comprises knee joint.

7. controller as claimed in claim 6 is characterized in that leg also comprises hip joint.

8. controller as claimed in claim 2 is characterized in that, at least one sensor is angle joint displacements and velocity sensor, torque sensor or Inertial Measurement Unit.

9. a control comprises the method for the mechanical limb at least one joint, and this method may further comprise the steps:

Receive the feedback data that relates to the mechanical limb state at the finite state machine place;

Utilize finite state machine and the feedback data that receives to confirm the mechanical limb state;

The joint moment or the rigidity instruction that utilize neuromuscular model, muscle geometry and catoptric arrangement system information and confirm to be sent at least one expectation of mechanical limb from the status information of finite state machine; And

Control the definite bionical moment and the rigidity of muscle model processor of mechanical limb joint.

Images