KR20110122150A - Model-based neuromechanical controller for a robotic leg - Google Patents

Abstract

A model-based neuromachine controller for a robotic limb with one or more joints 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 robot limb, and receive state information from the finite state machine and the muscles. A muscle model processor configured to determine one or more desired joint torque or stiffness commands to be sent to the robot limb using shape and reflex architecture information and a neuromuscular model, and to command the biomimetic torque and stiffness determined by the muscle model processor at the robotic limb joint. A configured articulation command processor. The feedback data is preferably provided by one or more sensors mounted at each joint of the robot limb. In a preferred embodiment, the robotic limb is a leg and the finite state machine is synchronized to the leg walking cycle.

Description

MODEL-BASED NEUROMECHANICAL CONTROLLER FOR A ROBOTIC LEG}

Related application

This application claims the benefit of US Provisional Patent Application No. 61 / 148,545, filed January 30, 2009, all of which is incorporated herein by reference.

This application is now extinct but in part consecutive application of US patent application Ser. No. 12 / 157,727, filed Jun. 12, 2008, claiming the benefit of U.S. Provisional Patent Application No. 60 / 934,223, filed Jun. 12, 2007. And partial serial applications of US Patent Application Nos. 11 / 395,448, 11 / 495,140, and 11 / 642,993, the disclosures of which are incorporated herein by reference in their entirety.

In addition, US patent application Ser. No. 11 / 642,993, filed Dec. 19, 2006, which is now extinct but now disclaims the claim of US patent application 60 / 751,680, filed December 19, 2005. US Patent Application Nos. 12 / 608,627, filed Oct. 29, 2009, which is the serial application of US Patent Application Nos. 11 / 395,448, 11 / 495,140, and 11 / 600,291, Partial continuation of US Patent Application No. 11 / 499,853 (currently US Patent No. 7,313,463), which is now extinguished but claims the benefit of the filing date of US Provisional Patent Application No. 60 / 705,651, filed August 4, 2005. Application, which is a partial serial application of US Patent Application Ser. No. 11 / 395,448, below, the entire disclosure of which is incorporated herein by reference in its entirety.

The application is also filed on March 31, 2006 by Hugh M. Herr, Daniel Joseph Paluska and Peter Dilworth, "Artificial human limbs and joints employing actuators employing actuators, springs and variable damper elements. , springs, and Variable-Damper Elements), a partial continuation of US patent application Ser. No. 11 / 395,448, filed under the name of the invention, and US patent application Ser. No. 11 / 395,448, which is now extinguished, Claim the benefit of the filing date of U.S. Provisional Patent Application 60 / 666,876 and the filing date of U.S. Provisional Patent Application No. 60 / 704,517, filed August 1, 2005, which is now extinguished.

In addition, this application was issued on July 29, 2006 by Hugh M. Herr, Samuel K. Au, Peter Dilworth, and Daniel Joseph Paluska, "Ankle-foot system with spring, variable damping and a series of elastic actuator components. (An Artificial Ankle-Foot System with Spring, Variable- Damping, and Series-Elastic Actuator Components) is a partial serial application of US patent application Ser. No. 11 / 495,140 filed.

U.S. Patent Application No. 11 / 495,140 claims the benefit of the filing date of U.S. Provisional Patent Application No. 60 / 704,517, which is now extinguished but filed August 1, 2005, and also a partial consecutive application of U.S. Patent Application No. 11 / 395,448. to be.

The application is also filed November 15, 2006 by Hugh M. Herr, Conor Walsh, Daniel Joseph Paluska, Andrew Valiente, Kenneth Pasch, and William Grand, entitled "Exoskeletons for running and walking." US Patent Application No. 11 / 600,291 filed under the name of the invention. U.S. Patent Application No. 11 / 600,291 claims the benefit of the filing date of U.S. Provisional Patent Application No. 60 / 736,929, filed November 15, 2005, but now extinguished, and U.S. Patent Application Nos. 11 / 395,448, 11 / 499,853, and 11 / 495,140.

The present invention claims the benefit of the filing date of each of the aforementioned patent applications, and the disclosure of each of the aforementioned patent applications is incorporated herein by reference.

Reference to government-funded research or development

The present invention was made with US government support under license number VA241-P-0026, licensed by the United States Veterans Administration. The government has certain rights in the invention.

Technical field

FIELD OF THE INVENTION The present invention relates to the control of artificial joints and limbs for use in prosthetics, correction, exoskeleton, or robotic devices, and in particular to control of control methodologies for robotic legs based on neuromuscular models of motion.

Animal and human leg movements are controlled by complex neural networks. It was proposed in the early 20th century [Brown, TG, 1914, On the nature of the fundamental activity, with the analysis of the state of continuous regular activity in the characterization of the basic activity of the nerve center. of the nervous centers; together with an analysis of the conditioning of rhythmic activity in progression, and a theory of the evolution of function in the nervous system (J Physiol 48 (1), 18-46)], Orlovsky, G., Deliagina, T., Grillner, S., 1999, Neuronal control of locomotion: from mollusc to man (Oxford University Press, New York), pivotal pattern The generator (CPG) forms the basis of this network.

In the present sense, the Central Pattern Generator (CPG) consists of layers of neuron pools in the spinal cord [Rybak, IA, Shevtsova, NA, Lafreniere-Roula, M., McCrea, DA, 2006, Associated modeling spinal circuitry involved in locomotor pattern generation: insights from deletions during fictive locomotion (J Physiol 577 (Pt 2), 617-639)] Other neuronal pools that deliver genes provide regular activity to the leg extensor and flexor muscles (Dietz, V., 2003, Mobile Spinal Pattern Generator (Clin Neurophysiol 114 (8), 1379-1389)); Minassian, K., Persy, L, Rattay, F., Pinter, MM, Kern, H., Dimitrijevic, MR, 2007, Human ureter circuits may be activated by external tonic injection to generate motor activity. (Human lumbar cord circuitries can be activated by extrinsic tonic input to generate locomotor-like activity) (Hum Mov Sci 26 (2), 275-295)], even without spinal reflexes, sufficient to generate stepping movements [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). Nevertheless, spinal cord reflexes are part of this complex network [Rybak, IA, Stecina, K., Shevtsova, NA, McCrea, DA, 2006. Spinal cord circuit modeling involved in generating movement patterns: understanding from the effects of centripetal simulation Modeling spinal circuitry involved in locomotor pattern generation: insights from the effects of afferent stimulation (J Physiol 577 (Pt 2), 641-658)], selection of motor patterns, timing of extensor and flexor activity, central pattern generator (CPG) contributes to the adjustment of the output.

By using a combination of central pattern generation and adjusting the reflexes, a neuromuscular model of 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)], neuromuscular model of 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)], the neuromuscular model in cats Ivashko, DG, Prilutski, B. L, Markin, SN, Chapin, JK, Rybak, IA, 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: 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 generation of human bipedal locomotion by a bio-mimetic neuro-musculo-skeletal by Ogihara, N., Yamazaki, N., 2001. model) (Biol Cybern 84 (1), 1-11); Paul, C, Bellotti, M., Jezernik, S., Curt, A., 2005. Development of a human neuro-musculo-skeletal model for investigation of spinal cord injury (Biol Cybern 93 (3), 153-170) has been developed as a fundamental tool for studying various control strategies of animal and human movement. The focus of these models was to reproduce the configuration of the lower reflection and pivot pattern generator (CPG) proposed by the experiment [Pearson, K., Ekeberg, O., Buschges, A., 2006. Using neuro-mechanical simulation Assessing sensory function in locomotor systems using neuro-mechanical simulations (Trends Neurosci 29 (11), 625-631). However, little attention has been given to understanding how this configuration can represent or encode the principles of kinetic mechanics.

This principle suggests that, unlike the complexity of the identified neural networks, little or no control is required for leg movement. For example, walking (Alexander, R., 1976. Mechanics of bipedal locomotion. In: Perspectives in experimental biology (Ed. Davies, PS) (Pergamon, Oxford; Mochon, S., McMahon, T., 1980. Ballistic walking.J. Biomech. 13 (1) , 49-57) and driving [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) Two conceptual models were published to account for the dominant mechanism of leg motion, and the researchers demonstrated the capacity of these models to self-stabilize if the mechanical systems were properly adjusted [McGeer, T., 1990. Passive. Passive dynamic walking (Int. J. Rob. Res. 9 (2), 62-82; McGeer, T., 1992), Principles of walking and running (Vol. 11 o f Advances in Comparative and Environmental Physiology.Springer-Verlag Berlin Heidelberg, Ch. 4; Seyfarth, A., Geyer, H., Guenther, M., Blickhan, R., 2002. 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). Walking and traveling robots have also demonstrated control gains and substantial relevance derived from this principle. [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. Rex: A simple and highly mobile hexapod robot (Int. Jour. Rob. Res. 20 (7), 616-631); Collins, S., Ruina, A., Tedrake, R., Wisse, M., 2005. Effective bipedal robots based on passive-dynamic walkers (Science 307 (5712), 1082- 1085)]. However, how these and other principles of leg dynamics are integrated into the human motion control system remains an open question.

The importance of the interaction between dynamics and motor control has been recognized by neuroscientists and biomechanics [Pearson, K., Ekeberg, O., Buschges, A., 2006. Sensory functions in mobility using neuro-mechanical simulations]. Assessing sensory function in locomotor systems using neuro-mechanical simulations (Trends Neurosci 29 (11), 625-631). For example, although central pattern generators (CPGs) are generally allowed to form central drives for athletic activity in athletics [Grillner, S., 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 Neurophysiol 114 (8), 1379-1389; Frigon, A., Rossignol, S., 2006), sensory motor interaction during exercise 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, in 1969, even without a rather simple central input, spinal reflexes that relay information about kinematics can form the complex muscle activity found in real exercise. It is suggested that it can be [Lundberg, A., 1969. Reflex control of stepping. In: The Nansen memorial lecture V (Oslo: Universitetsforlaget, 5-42)]. By refining this idea, Taga later said that the rhythm generated in the center is accompanied by sensory signals induced by the regular movement of the exercise device, so that the output of the exercise is the nervous system, the muscle-skeletal system, It is proposed to be an anomalous characteristic of dynamic interactions between environments [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). In favor, a combination of central pattern generator (CPG) with feedback and a neuromuscular model of human migration has been presented demonstrating how basic gait can emerge from a comprehensive incorporation between the regular activity of the neuronal and muscular-skeletal systems.

The actual proportions of the central and reflex inputs that produce motor output are discussed continuously [Pearson, KG, 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 Cybern 95 (6), 607-627); Hultborn, 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 165, 255-265). For example, in walking cats, it was estimated that only 30% of the muscle activity observed in weight-bearing leg extensors can contribute to muscle reflexes [Prochazka, A., Gritsenko, V., Yakovenko, S., 2002. Sensory control of locomotion: reflexes versus higher-level control (Adv Exp Med Biol 508, 357-367); Donelan, JM, McVea, DA, Pearson, KG, 2009. Force regulation of ankle extensor muscle activity in freely walking cats (J Neurophysiol 101 (1), 360- 371)].

In humans, the contribution of reflexes to muscle activity in movement appears to be more pronounced. Sinkjaer and colleagues estimate that reflexes from unloading experiments contribute about 50% to halibut muscle activity during the standing phase of walking [Sinkjaer, T., Andersen, JB, Ladouceur, M., Christensen] , LO, Nielsen, JB, 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). More recently, Gray and colleagues have found that halibut activity changes in proportion to changes in Achilles tendon force, suggesting a direct link between activity for this muscle and positive force feedback [Grey, MJ , Nielsen, JB, Mazzaro, N., Sinkjaer, T., 2007. Positive force feedback in human walking (J Physiol 581 (1), 99-105). Whether such large reflex contributions are present in all leg muscles remains unresolved. As Daley and his colleagues concluded from the exercise experiments with birds, perhaps a proximo-distal gradient exists in motor control, where the close leg muscles are primarily controlled by central input, Far leg muscles are dominated by reflex inputs due to higher self-acceptance feedback gain and greater sensitivity to mechanical effects [Daley, MA, Felix, G., Biewener, AA, 2007. Driving stability is controlled by joint nervous system control. Running stability is enhanced by a proximo-distal gradient in joint neuromechanical control (J Exp Biol 210 (Pt 3), 383-394)].

Adaptation to the terrain is an important aspect of walking. Today's commercially available ankle-foot prostheses utilize lightweight passive structures that are configured to provide adequate elasticity during the standing phase of the walk [S. Ron, Prosthetics and Oralology: Prosthetics and Orthotics: Lower Limb and Spinal (Lippincott Williams & Wilkins 2002). Advanced composites used in these devices allow for a certain energy storage during controlled dorsiflexion and plantar flexion and subsequent energy release during energized lower flexion, such as in the Achilles tendon of a healthy person. [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 passive-elastic behavior is a good approximation of the function of the ankle during low speed walking, normal and high speed walking speeds require the addition of external energy and thus cannot be performed by any passive ankle-foot device [M . Palmer, "Sagittal plane characterization of normal human ankle function across a range of walking gait speeds (Master's Thesis, Massachusetts Institute of Technology, Cambridge, MA, 2002); DH Gates, "Characterizing ankle function during stair ascent, descent, and level walking for ankle prosthesis and orthosis design," (Master's Thesis, Boston University, 2004); AH Hansen, DS Childress, SC Miff, SA Gard, KP 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 defect is reflected in the gait of the lower leg cutter using passive ankle-foot prostheses. Their self-selected walking speed is slower than normal and the stride length 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, the gait is clearly asymmetrical, with a smaller range of ankle movements on the unaffected side [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), on the other hand, on the affected side, the hip extension moment is greater and the knee flexion moment is smaller [D. A. Winter and S. E. Sienko. "Biomechanics of below-knee amputee gait" (Journal of Biomechanics, 21, pp. 361-367, 1988); 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). In addition, they consume more metabolic walking energy than non-cutters [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, 126 136, 1997); A. Esquenazi, and R. DiGiacomo. "Rehabilitation After Amputation" (Journ Am Podiatr Med Assoc, 91 (1): 13-22, 2001). This difference may be the result of the cleaver using more heap power to compensate for the lack of ankle power [A. D. Kuo, "Energetics of actively powered locomotion using the simplest walking model" (J Biomech Eng., Vol. 124, pp. 113-120 , 2002); AD Kuo, JM 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, JE Bertram, and M. Srinivasan, "Effective cost-impact models of support walks qualitatively explain rapid leg resilient leg behavior in walking and walking and walking-to-travel transitions. (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-foot prosthetics cannot provide adaptation to the terrain. To provide a normal economic gait beyond slow walking speed, a power ankle-foot prosthesis has now been developed [S. Au and H. Herr. "Initial experimental study on dynamic interaction between an amputee and a powered ankle-foot prosthesis" (Workshop on Dynamic Walking: Mechanics and Control of Human and Robot Locomotion, Ann Arbor, MI, May 2006); SK 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); 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 and 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-Foot) Project with Regenerating Kinematics: Design and Analysis of Robot Tibial Prosthetics with Regenerating Kinematics. 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); SK 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). Some of these are of size and weight comparable to complete human ankles and have elastic energy storage, motor power and battery energy to provide daily walking activity [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).

The use of active motor power in these prostheses creates a problem of control. In past studies of these powered devices, attempts have been made to match the torque-ankle state profile of the complete human ankle for the activity being performed [S. K. Au, "Powered Ankle-Foot Prosthesis for the Improvement of Amputee Walking Economy" (Ph. D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, 2007); J. Hitt, R. Bellman, M. Holgate, T. Sugar, and K. Hollander, "Sparky (Spring Ankle-Foot) Project with Regenerating Kinematics: Design and Analysis of Robot Tibial Prosthetics with Regenerating Kinematics (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. Goldfarb, "Design and Control of a Powered Transfemoral Prosthesis" (The International Journal of Robotics Research, Vol. 27, No. 2, pp. 263-273, 2008). The provision of motor power means that in addition to the spring-like behavior provided by the manual device, an open work loop of faster walking angle-torque profiles can be supported. However, this control approach did not show any inherent adaptation. Instead, for all intended activities and terrain changes, a torque profile is required with appropriate means for choosing between them.

In general, existing commercially available active ankle prostheses can only reconstruct the ankle joint angle during the phase phase and require multiple steps to converge to the ankle position appropriate for the terrain upon initial ground contact. In addition, they do not provide any power in the stance phase needed for general gait and thus cannot adapt pure standing work to terrain gradients. In particular, the control scheme for power ankle-foot prosthetics relies on a fixed torque-ankle state relationship obtained from complete human measurements walking at a target speed across a known terrain. Although they may be effective at their target gait speed and terrain, these controllers are unable to adapt to environmental disturbances such as speed transitions and terrain variations.

A neuromuscular model with a positive force feedback reflex system as the basis of control has recently been employed in a biomechanical simulation study of leg motion [H. Geyer, 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 studies show the prospect of the need for terrain adaptation.

In one aspect, the invention is a controller and control method for a biomimetic robotic leg based on a neuromuscular model of human motion. The control architecture commands biomimetic torque at the ankle, knee and hip joints of the powered leg prosthesis, braces or exoskeleton during walking. In a preferred embodiment, the powered device includes an artificial ankle and knee joint, which are torque controllable. Appropriate joint torque is provided to the user as determined by feedback information provided by sensors mounted on each joint of the robotic leg device. These sensors include, but are not limited to, angular joint displacement and speed, torque sensors at the ankle and knee joint, and at least one inertial measurement unit (IMU) located between the knee and ankle joint using digital encoders, hall effect sensors, and the like. It doesn't work.

Sensor information of joint state (position and velocity) from the robotic leg is used as input to the neuromuscular model of human motion. Joint state sensor information from the robotic leg is used to determine the internal state for each of its virtual muscles, and it is determined from the spinal reflex model which individual virtual muscle force and stiffness should be provided for a particular level of muscle activity. If the robotic leg is a leg prosthesis worn by the femoral cutter, the ankle and knee angle sensors measure the joint condition for these joints. For hip joints, the absolute orientation of the user's thigh is determined using both the IMU located between the prosthetic knee and ankle joint and the angular joint sensor of the prosthetic knee. To estimate heap position and speed, the control architecture operates under the assumption that the upper body (upper body) maintains a relatively vertical position during gait.

In one aspect, the present invention is a model-based neuromachine controller for a robotic limb that includes at least one joint, the controller configured to receive feedback data about the status of the robotic limb and to determine the status of the robotic limb. And receive state information from a finite state machine and muscle shape, reflection architecture information from at least one database, and using the neuromuscular model to determine at least one desired joint torque or stiffness command to be sent to the robot limb. A muscle model processor and a joint command processor configured to command the biomimetic torque and stiffness determined by the muscle model processor at the robotic limb joint. In a preferred embodiment, the feedback data is provided by at least one sensor mounted to each joint of the robot limb. In another preferred embodiment, the robotic limb is a leg and the finite state machine is synchronized to the gait cycle.

In another aspect, the present invention is a model-based method for controlling a robot limb including at least one joint, the method comprising receiving feedback data relating to a state of the robot limb in a finite state machine, finite state machine and Determining the state of the robot limb using the received feedback data, and at least one desired joint torque or stiffness to be transmitted to the robot limb using neuromuscular model, muscle shape and reflex architecture information, and state information from the finite state machine. Determining a command, and commanding the biomimetic torque and stiffness determined by the muscle model processor at the robotic limb joint.

Other aspects, advantages and novel features of the invention will become more apparent from the following detailed description of the invention when considered in conjunction with the accompanying drawings.
1 is a block diagram of an exemplary embodiment of a general neuromuscular model architecture, in accordance with an aspect of the present invention.
2A-2F illustrate six stages of development of a general neuromuscular model architecture, in accordance with an aspect of the present invention.
3 diagrammatically illustrates pattern generation according to one aspect of a general neuromuscular model architecture in accordance with the present invention.
4A and 4B illustrate the walking of a self-organized human model and the corresponding ground reaction forces, respectively, from the dynamic interaction between the model and the ground, in accordance with an aspect of the present invention.
5A-5C compare static state gait for each hip, knee and ankle for the model and human subject, in accordance with an aspect of the present invention.
6A-6D are snapshots of the model (FIG. 6A), net work (FIG. 6B), extensor muscle activation pattern (FIG. 6C) and corresponding ground reaction forces (FIG. 6D), according to one aspect of the present invention. It shows the adaptation to walking walking up stairs.
7A-7D illustrate a staircase including a snapshot of a model (FIG. 7A), a pure work (FIG. 7B), an extensor muscle activation pattern (FIG. 7C), and a corresponding ground reaction force (FIG. 7D), in accordance with an aspect of the present invention. It shows the adaptation to walking.
8 is a schematic of a muscle-tendon model, in accordance with an aspect of the present invention.
9 illustrates a contact model in accordance with an aspect of the present invention.
10A-10C show an exemplary embodiment of ankle-foot prosthesis used in the preferred embodiment, according to one aspect of the present invention, which is a physical system (FIG. 10A), a view of the drive train (FIG. 10B), and The mechanical model (FIG. 10C) is shown respectively.
11 is a step with a state transition threshold and an equivalent ankle-foot biomechanical device during each state used to implement top level control of the ankle-foot prosthesis of FIGS. 10A-10C, in accordance with an aspect of the present invention. A diagram of an exemplary embodiment of a finite state machine synchronized to a hang cycle.
12 is a block diagram of an exemplary embodiment of a control system for ankle-foot prosthetics in accordance with an aspect of the present invention.
13A-13C show one complete gait cycle for three walking conditions, namely flat ground (FIG. 13A), rising ramp (FIG. 13B) and falling ramp (FIG. 13C), according to one aspect of the present invention. An example plot of prosthetic torque over.
14A-14C show a prosthesis comprising two link ankle joint models (FIG. 14A), a detailed heel-shaped muscle model (FIG. 14B) and a muscle model skeletal attachment (FIG. 14C), according to one aspect of the present invention. An exemplary embodiment of a musculoskeletal model as implemented on a microcontroller is shown.
15 illustrates an exemplary embodiment of a reflex system for a virtual plantar flexion muscle, including the relationship between plantar flexion components of ankle angle, muscle force, and ankle torque, in accordance with an aspect of the present invention.
16A and 16B show prosthetic measurement torques during an experiment of a subject undergoing amputation compared to those of a biological ankle of a subject having a weight and height with complete limbs, including ankle torque and ankle angle, respectively; The angle trajectory is shown.
17 is a comparison of torque profiles after parameter optimization for biological torque profiles, in accordance with an aspect of the present invention.
18A-18C are experimentally measured for an exemplary embodiment of the present invention for three different walking conditions: flat ground (FIG. 18A), rising ramp (FIG. 18B) and falling ramp (FIG. 18C). Plots of prosthetic torque angle trajectories.

A control architecture is provided for commanding biomimetic torques at the ankle, knee and hip joints of powered leg prosthetics, orthopedic or exoskeleton during walking. In this embodiment, the powered device includes a torque controllable artificial ankle and knee joint. Appropriate joint torque is provided to the user as determined by feedback information provided by sensors mounted on each joint of the robotic leg device. These sensors include, but are not limited to, angular joint displacement and speed using digital encoders, Hal-effect sensors, and the like, torque sensors in the knee and ankle joints and at least one inertial measurement unit (IMU) located between the knee and ankle joints. It doesn't work.

Sensor information of joint status (position and velocity) from the robotic legs (hip, knee and ankle) is used as input to the neuromuscular model of human motion. This model uses the joint state sensor information from the robotic leg to determine the internal state for each of its virtual muscles, and to form individual virtual muscle forces and stiffnesses that should be provided for specific levels of muscle activation determined from the spinal reflex model. do. If the robotic leg is a leg prosthesis worn by a femoral cutter, the ankle and knee angle sensors measure the joint condition for these joints. With respect to the hip joint, the absolute orientation of the user's thigh is determined using both the IMU located between the prosthetic knee and the ankle joint and the angular joint sensor of the prosthetic knee. To estimate heap position and speed, the control architecture works on the assumption that the upper body (upper body) remains in a relatively vertical position during the gait.

As used in this specification and the application incorporated herein by reference, the following terms expressly include but are not limited to the following meanings.

"Operator" means one type of motor as described below.

"Apical muscle" means a contractile element that is arrested or counteracted by another element, ie the antagonist.

"Apical-antagonist" means a mechanism that includes (at least) two actuators operating opposite each other, namely an agonist that pulls two elements together here and an antagonist actuator that presses two elements apart here. it means.

"Antagonist" means another element, an expansion element that is arrested or canceled by the main root muscle.

"Bioimitation" means a human made structure or mechanism that mimics the properties and behavior of a biological structure or mechanism, such as a joint or limb.

"Odor" refers to bending the ankle joint so that the end of the foot moves upward.

"Elastic" means that it can return to the original phenomenon after deformation by stretching or compression.

"Elongation" refers to the bending movement around the joint in the limb, which increases the angle between the bones of the limb at the joint.

"Flexion" means the bending motion around the joint of the limb, in which the angle between the bones of the limb in the joint decreases.

"Motor" means an active element that generates or imparts movement by converting supplied energy, including electric, pneumatic or hydraulic motors and actuators, into mechanical energy.

"Foot plantar flexion" means bending the ankle joint so that the end of the foot moves downward.

"Spring" means an elastic device, such as a metal coil or plate structure, that is returned to its original shape after compression or stretching.

An exemplary embodiment of a neuromuscular model based control scheme in accordance with this aspect of the present invention is shown as a block diagram in FIG. 1. In FIG. 1, the neuromuscular model according to the present invention includes a reflex loop 110 for each modeled muscle unit 120. Predicted force and stiffness from all modeled muscles are used to calculate 130 joint torque and stiffness using muscle moment arm values 140 from the literature. The model estimate is then sent to the controller as the desired pure torque and stiffness value for the biomimetic robot leg joint 150. The controller 160 then tracks the torque and stiffness values at each robot joint 150.

In order for each virtual muscle to generate its required force, muscle stimulation parameters [STIM (t)] are needed. This parameter may be determined from either an external input or a local feedback loop. In the control method for an exemplary biomimetic leg, STIM (t) is calculated based on a local feedback loop. This architecture is based on the reflex feedback framework developed by Geyer and Herr [H. Geyer, H. Herr, "A muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities", (Submitted for initiation), the entirety of which is incorporated herein by reference]. In this framework, neural control is designed to mimic renal reflexes of intact human muscles. This neuromuscular reflection-based control method allows a biomimetic robot leg to simulate human-like joint dynamics.

Nervous system model. Predict human walking kinetics and muscle activity with a human model with reflex control encoding the principles of leg dynamics. While neuroscientists have uncovered an increasingly complex neural network that controls animal and human walking, biomechanics have found that, if the principles of leg dynamics are taken into account, the motion requires small motor control. Here we show how muscle reflex behavior can be important in linking these two observations. A human kinetic model developed by muscle reflex behavior that encodes the principles of leg dynamics has been developed. With reflection control based on this principle, the model stabilizes from gait interaction with the ground, withstands ground obstacles, and adapts itself to stairs. The model also shows a qualitative agreement with joint angles, joint torques, and muscle activity known in the experiment, and human motor output is largely driven by muscle reflex behavior that links the principles of leg dynamics to the neural networks responsible for movement. It can be formed.

Human walking models with motor control are based on muscle reflexes and are designed to incorporate the principles of such leg dynamics. This principle derives from a simple conceptual model of legged locomotion and is dependent on the compliant leg behavior in stance [Blickhan, R., 1989. Spring for driving and hopping. The spring-mass model for running and hopping. J. of Biomech. 22, 1217-1227; 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 behavior explains the basic dynamics of walking and running. Proc. R. Soc. Lond. B 273, 2861-2867] and stabilization of articulated legs based on static joint torque equilibrium [Seyfarth, A., Guenther, M., Blickhan, R., 2001. of an elastic three-segmented leg). Biol. Cybern. 84, 365-382; Guenther, M., Keppler, V., Seyfarth, A., Blickhan, R., 2004. Human leg design: optimal axial alignment under constraints. J. Math. Biol. 48, 623-646 and exploitation of ballistic swing-leg mechanics [Mochon, S., McMahon, T., 1980. Ballistic walking. J. Biomech. 13 (1), 49-57], and improved gait stability using angle-leg rims [Seyfarth, A., Geyer, H., Guenther, M., Blickhan, R., 2002. Movement for Driving Standard (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 muscles combined with spinal reflexes are employed, and positive force and length feedback schemes are included to effectively encode these mechanical features.

Comparing the behavior of the model with the kinetic, kinematic, and electromechanical evidence from the literature on human gait, a neuromuscular model with motor control designed to encode the principles of leg dynamics can produce biomechanical walking dynamics and muscle activity. It turns out that you can. This reflection control allows the model to withstand drastic changes in ground level and adapt to the rise and fall of stairs without adjusting parameters.

The structure and control of the human model are six stages, from the conceptual point-mass model, to a neuromuscular bipedal animal with two 3-segment legs, each operated by the upper body and seven muscles and controlled by muscle reflexes. Develops. 2A-2F show six stages of development of a general neuromuscular model architecture in accordance with this aspect of the invention. The first three stages integrate and stabilize the conformal leg behavior of the stator (FIGS. 2A-2C). The fourth stage adds an upper body and its balance control (FIG. 2D). The last two stages prepare and assure the straightening and pinching of the legs during the instep (FIGS. 2E-2F).

In FIGS. 2A-2F, which are described in more detail in the following paragraphs, lumped vasti group muscle (VAS) and soleus muscle (SOL) with positive force feedback (F +) By driving, it develops from the standing leg structure (FIG. 2A), resulting in a compliant leg behavior that enables walking and running (FIG. 2B). Biarticular gastrocnemius muscle (GAS) is added using F + to prevent excessive elongation of the knee (FIG. 2C) and VAS is inhibited if the knee is stretched beyond the 170 ° threshold. Tibialis anterior muscle (TA) is added to prevent excessive extension of the ankle, and pulling the ankle joint to a position bent by positive length feedback (L +) is caused by negative force feedback (F-) from the sphincter. Suppressed under normal standing conditions. An upper body is added to allow leg elevation (FIG. 2D). The hip flexor (HFL) and the co-activated hip extensor muscles (GLU, HAM) of the leg legs are driven into the vertically inclined reference, and the articulated HAM is the hip extensor moment Prevent excessive elongation of the knee by Landing of the other (preceding) leg initiates the burn by adding a constant stimulus to the HFL / GLU, respectively, and suppressing the VAS in proportion to the load held by the other leg (FIG. 2E). The actual leg elevation is promoted by HFL using L + until suppressed by L− of HAM (FIG. 2F). The stimulus of the HFL is deflected according to the tilt of the upper body taking off. In addition, using F + for GLU and HAM causes the legs to be lifted and stretched towards the ends of the glaze. Finally, now unsuppressed L + of TA drives the ankle to the bent position (FIG. 2G).

Plinth leg compliance and stability. The two-spring spring-mass model is used as a starting point for the conceptual basis for human motion (FIG. 2A). Although this model consists of only point-mass 205 running on two massless spring legs 201 and 215, it unifies both steps into one conceptual framework based on the conformable leg behavior of the stance. Geyer, H., Seyfarth, A., Blickhan, R., 2006. Compliant leg behavior explains the basic dynamics of walking and running. Proc. R. Soc. Lond. B 273, 2861-2867.

In order to achieve adaptive behavior of the nerve root legs, each spring 210, 215 is replaced with a thigh 220, a shin 225, and a foot 230, and the sphincter muscle (SOL) 235 and the basti muscles. A group (VAS) 240 is added to generate their muscle activity both via local positive force feedback (F +) during the gait phase of the gait (FIG. 2B). These force reflections are described by Geyer, H., Seyfarth, A., Blickhan, R., 2003. Positive force feedback in bouncing gaits? Proc. R. Soc. Lond. Modeled in the same manner as in B 270, 2173-2183. Under positive force feedback, the stimulus [Sm (t)] of muscle (m) is the sum of the prestimulus (S0, m) and the time delay (Δt) and the obtained (G) force (Fm) of the muscle: Sm (t ) = S0, m + GmFm (t-Δtm).

While compliant leg behavior is fundamental, it also threatens joint stability of the articulated leg [Seyfarth, A., Guenther, M., Blickhan, R., 2001. Stable operation of an elastic articulated leg elastic three-segmented leg). Biol. Cybern. 84, 365-382; Guenther, M., Keppler, V., Seyfarth, A., Blickhan, R., 2004. Human leg design: optimal axial alignment under constraints. J. Math. Biol. 48, 623-646]. In the articulated leg, the knee and ankle torques (τ _k and τ _a ) follow the static equilibrium (τ _k / τ _a = h _k / h _a ), where h _k and h _a are the leg force vectors from the knee and ankle, respectively ( Right angle to Fleg). In effect, a large stretch torque in one joint forces the other joint close to the Fleg, resulting in excessive stretch concerns for the leg behaving like a spring [see Seyfarth, A., Guenher, M., Blickhan, R., 2001. Stable operation of an elastic three-segmented leg. Biol. Cybern. 84, 365-382].

This overstretching tendency in the knee or ankle is countered by the addition of gastrocnemius (GAS) 245 and tibial anterior (TA) 250 muscles (FIG. 2C). Like SOL and VAS, articulated GAS uses local positive force feedback (F +) during the gait phase of the gait. This muscle reflex not only prevents excessive elongation of the knee resulting from large stretch torque at the ankle, but also contributes to generating comprehensive compliant leg behavior. In contrast, the mono-articular TA uses local positive length feedback (L +), where S _TA (t) = S _{0 , TA} + G _TA (l _{CE , TA} -l _{off , TA} ) (t-Δ _{t, TA} ), Where l _{CE and TA} are TA fiber lengths, and l _{off and TA} are length offsets. When bending, TA's L + prevents the ankle from overextending when large knee torque occurs. However, this muscle reflex is not necessary if sufficient ankle extensor activity preserves the knee and ankle torque balance. In this situation, TA stimulation is suppressed with negative force feedback (F-) from SOL to avoid TA from unnecessarily opposing SOL, resulting in S _TA (t) = S _{0 , TA} + G _TA (l _{CE , TA} -l _{off , TA} ) (t-Δ _{t, TA} ) -G _SOLTA F _SOL (t-Δt _SOL ). To further prevent the knee from being stretched too much, if the knee is stretched beyond the 170 degree threshold, VAS is suppressed and S _VAS (t) = S _{0 , VAS} + G _VAS F _VAS (t-Δt _VAS )-k _Φ ΔΦ _k (t-Δt _k ), with k _Φ being proportional gain, ΔΦ _k = Φ _k -170 degrees, and Φ _k being the knee angle. This reflex suppression is ΔΦ> 0 and is only activated when the knee is actually stretched.

Upper body and its balance. In the next stage of development from a conceptual spring-mass model to a neuromuscular biped, the point mass indication is eliminated and an upper body 255 is introduced in which a leg can be angled around (FIG. 2D). This upper body 255 combines the head, arms and trunk HAT. In order to balance the HAT 255 during exercise, each leg is added with a gluteus muscle group (GLU) 260 and a hip flexor muscle group (HFL) 265. GLU 260 and HFL 265 are stimulated with a proportional-derivative signal 255 of the HAT towards the tilt angle θ with respect to gravity, and S _{GLU / HFL} ˜ ± [k _P ( θ-θ _ref ) + k _d dθ / dt], provided that k _P and k _d are proportional and differential gain, and θ _ref is the reference tilt angle [as a similar approach, for example, Guenther, M., Ruder , H., 2003. Synthesis of two-dimensional human walking: a test of the λ-model. Biol. Cybern. 89, 89-106]. In addition, a biarticular hamstring muscle group (HAM) 270 is included and S _HAM -S _GLU results from the large hip torque generated by the GLU 260 when pulling the heavy HAT 255. Against excessive elongation of the knee; Since hip torque can only balance the HAT 255 if the legs endure enough weight, the stimulation of the GLU 260, HAM 270, and HFL 265 is dependent on the amount of body weight it bears for each leg. Modularly proportional. As a result, the hip muscles of each leg contribute to the balance control of the HAT only during the standing phase.

이다. 대측성 억제는 무릎이 그의 기능적 스프링 거동을 파괴하고, 발목이 신장하는 동안 굽혀지며, 지면으로부터 다리를 떨어지게 하고 전방으로 가압하게 한다. 이러한 캐터펄트 기구(catapult mechanism)는 발목이 충분히 가압될 때에만 유각을 시작할 수 있는 반면, 본 모델은 또한 HFL(265)의 자극을 증가시키고 GLU(260)의 자극을 감소시킴으로써, 이중 지지부의 고정된 양 ΔS만큼 유각 다리 펴기를 준비한다.Stretch and pinch your bead. The structure of the human model was completed except for the muscle-reflective control, which produced straightening and pinching of the leg legs. The functional importance of the standing leg is reduced in proportion to the amount of body weight (bw) to be tolerated by the contralateral leg, and it is believed that the angled leg extension already begins at the double support (FIG. 2E). The human model detects which leg enters the standing phase last (the contralateral leg), suppresses the F + of the VAS 240 of the ipsilateral leg in proportion to the weight that the contralateral leg bears, and S _VAS (t ) = S _{0 , VAS} + G _VAS F _VAS (t-Δt _VAS )-k _Φ ΔΦ _k (t-Δt _k )-k _bw

to be. The contralateral suppression causes the knee to break its functional spring behavior, bend while the ankle is stretched, to lift the leg from the ground and press forward. While this catapult mechanism can only start the burn when the ankle is fully pressurized, the model also increases the stimulation of the HFL 265 and decreases the stimulation of the GLU 260, thereby fixing the double support. Prepare the leg stretch for both ΔS.

It depends mainly on the ballistic behavior of the legs during actual burns, but is affected in two ways (FIG. 2F). On the one hand, it is easier to straighten the legs. HFL 265 is stimulated using positive length feedback (L +) biased by the forward pitch angle (θ _ref ) of HAT 255 in the transition from standing to angular, where S _HFL (t) = S _{0 , HFL} + k _lean (θ-θ _ref ) _TO + G _HFL (l _{CE , HFL} -l _{off , HFL} ) (t- _{Δt, HFL} ). Using this approach, ballistic movement of the ridge bridge ensures that momentum is gained and sent forward in time (Mochon, S., McMahon, T., 1980. Ballistic walking). J. Biomech. 13 (1), 49-57].

In addition, the legs are also prevented from overreaching and are guaranteed to be pinched. If the leg reaches and remains in proper orientation during the angle, the leg system self-stabilizes with a gait cycle [McGeer, T., 1990. Passive dynamic walking. Int. J. Rob. Res. 9 (2), 62-82; Seyfarth, A., Geyer, H., Guenther, 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 behavior explains the basic dynamics of walking and running. Proc. R. Soc. Lond. B 273, 2861-2867. If the legs are additionally retracted before landing, the tolerance of this mechanical self-stabilization to disturbances can be greatly improved [Seyfarth, A., Geyer, H., 2002. Natural control of spring-type driving-optimized self 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, HM, 2003. Swing-leg retraction: a simple control model for stable running. J. Exp. Biol. 206, 2547-2555. To implement this halt-and-retract strategy, three muscle reflexes are included in the human model. While the knees are prevented from being stretched, overstretched legs are prevented from being overstretched, which may be caused by the anterior impact the legs receive when they reach full extension. At this point, L + in HFL is suppressed in proportion to the height HAM receives from the flap, and S _HFL (t) = k _lean (θ-θ _ref ) TO + G _HFL (l _{CE , HFL} -l _{off , HFL} ) (t -Δ _{t, HFL} ) -G _HAMHFL (l _{CE , HAM} -l _{off , HAM} ) (t- _{Δt, HAM} ). Also, F + is used for _GLU , S _GLU (t) = S _{0 , GLU} + G _GLU F _GLU (t-Δt _GLU ), also used for _HAM , S _HAM (t) = S _{0 , HAM} + G _HAM F _HAM (t-Δt _{H AM} ), which ensures that the leg is not only stopped according to the actual amount of stretching, but also passes that portion of the momentum to straightening and pinching. Finally, TA L + introduced to ensure foot spacing is maintained throughout the well. SOL, GAS, and VAS remain inactive during this phase.

Reflection control parameters. The equations used in this model control different reflection contributions to muscle stimulation [Sm (t)]. Parameter optimization was not performed. The parameters were derived from prior knowledge of reflection behaviors (F +, L +), or by making valid estimates. All muscle stimuli are limited in the range of 0.01 to 1 before being transformed into muscle activity [A _m (t)]. Table 1 shows the standing reflection equations used in the preferred embodiment.

및

다리가 이중 지지부 내의 끌려가는 다리(trailing leg)이면 DSup는 1이고, 그렇지 않으면 0이다. {}_+/- 는 포지티브/네거티브 값만을 지시한다.

And

DSup is 1 if the leg is a trailing leg in the double support, otherwise 0. {} _+/- indicates only positive / negative values.

Table 2 shows the angle reflection equations used in the preferred embodiment.

({} _PTO is a constant value taken when falling earlier.)

result. The human model does not have a central pattern generator (CPG) that moves muscles in a feed-forward manner, but uses a sensor located on each foot's ball and heel to sense the ground, standing on each leg. And different reflexes for glazing. As a result, the dynamic interaction of the model with its mechanical environment becomes an important part of generating muscle activity. 3 diagrammatically illustrates pattern generation according to this aspect of the invention. In FIG. 3, instead of the central pattern, reflexes generate muscle stimulation (S _m ) 305, 310. The legs 320 on the left (L) and the legs 330 on the right (R) have separate stances 340, 345 and angles 350, 355 reflecting, which are ball and heel sensors 370, 375. ) Is selected based on contact sensing 360, 365 from. The reflex output depends on mechanical inputs (M _i ) 380, 385, intertwined instruments and motor control.

Walking gait. To see how important this interdependence of mechanisms and motor control is to human movement, the model puts its left leg in standing position and its right leg in elevation, at a normal walking speed (v0 = 1.3 ms- ^{). 1} ) set off. Since the reflexes of modeled muscles include a time delay of up to 20ms, all muscles are not initially activated. 4A and 4B respectively show the dynamic interaction between the model and the ground according to one aspect of the invention and the walking of a human model that is self-organized from the corresponding ground reaction forces. 4A and 4B, snapshots of a human model taken every 250ms (FIG. 4A) and snapshots of the corresponding model GRF (FIG. 4B) are shown, left leg 405, 410 and right leg 415. 420, there is a separate plot (30 Hz low pass filtered). Starting at a horizontal speed of 1.3 ms ⁻¹ , the model slows down in the first two steps and rapidly recovers walking at the same speed. The leg muscles are shown only for the right leg 415 and show greater than 10% muscle activity. The initial conditions of φ _{a, k, h} (definition of ankle, knee and hip angles) for each leg are φ _{a, k, h} = 85 degrees, 175 degrees, 175 degrees (left leg) and φ _{a, k, h} = 90 degrees, 175 degrees, and 140 degrees (right leg).

Because of this disturbing initial condition, the model sits slightly and slows down in the first step (FIG. 4A). However, if its parameters are properly selected, the model quickly recovers in subsequent steps, and the walk itself is organized from the dynamic interaction on the ground with the model. The vertical ground reaction force (GRF) of the leg in the standing state exhibits M-shaped pattern characteristics for gait (FIG. 4B), and similar systemic dynamics of the model and human in steady state gait.

Steady state pattern of angle, torque and muscle activity. This similarity is maintained even during close examination; The model shows qualitative identity with known angle, torque and muscle activity patterns from human gait data. 5A-5C compare the steady state gait of 1.3 ms ⁻¹ of the model and human subject in the hip (FIG. 5A), knee (FIG. 5B) and ankle (FIG. 5C) according to one aspect of the present invention. 5A-5C, based on one stride between heel-strikes of the same leg, the steady state pattern of muscle activity, torque, and angle of the hip, knee, and ankle of the model is shown in human walking data. Compared to Perry's (1992). Vertical dashed line 510, approximately 60% of the stride length, represents toe-off. The compared muscles were found in the adductor longus (HFL) 520, the upper gluteus maximum (GLU) 530, the semimembranosis (HAM) 540 and the lateral tussalis (vastus lateralis). 550).

The strongest match between model prediction and gait data can be found in the ankle (FIG. 5C). The reflex model not only generates ankle kinematics φ _a and torque τ _a observed from walking human ankles, but also predicts SOL, TA, and GAS resembling experimental SOL, TA, and GAS estimated from surface electromyography. In SOL and GAS these activities are generated by their local F + reflexes entirely in the standing state. For TA, its L + reflex responds with high activity to the plantar flexion of the foot in the initial standing state, but is suppressed by F- from SOL during the remainder of this step. Only when SOL activity decreases from transition from stance to stir (60% of stride), the resumption of the TA's L + pulls against foot flexion.

This comparison shows a weaker agreement on the knees and hips. For example, the general trajectory φ _k of the human knee is captured by the model, but the knee of the model bends about 10 degrees or 30% more than the human knee in the initial standing state (FIG. 5B). In connection with this large knee flexion, the model lacks the VAS activity observed in the late stent state leading to the initial standing state. Only after the heel strike, the F + of the VAS can meet the muscle groups in response to the load on the legs and operate them. Delays in extensor activity not only cause relatively weak knees in the early standing state, but also cause the heavy HAT to tilt forward after impact. Balance control of the HAT progressively meets the weight supported by the stance legs, resulting in no balanced reflexes until heel-stroke, thus generating unusually large GLU and HAM activity to return the HAT to its baseline slope. Should be done (FIG. 5C). Thus, the hip trajectory φ _h and torque pattern τ _h of the model are at least similar to those of humans, where hip extensor GLU and HAM are activated prior to impact and can prevent such excessive tilt of the torso.

Self adaptation to ground changes. Despite limited reflex control, the human model tolerates abrupt changes in ground level and adapts itself to long-lasting ground level changes. 6A-6D show an example where the model meets a series of stairs each up to 4 cm. 6A-6D illustrate steps for climbing stairs, including a snapshot of the model according to one aspect of the invention (FIG. 6A), pure water (FIG. 6B), extensor activity pattern (FIG. 6C), and corresponding ground reaction force (FIG. 6D). Shows adaptation. 6A-6D, approaching from a steady state walk of 1.3 ms ⁻¹ , eight strides of the human model are shown to cover five steps of 4 cm uphill each. The model recovers steady state walking at the eighth stride. One stride is defined as the heel strike from the right leg to the heel strike. Shown in FIG. 6A is a snapshot of the model at heel-strike and toe-off of the right leg. For this leg, pure work during the standing state that occurs in the hips, knees and ankles with a positive action extending in FIG. 6B is shown, and in FIG. 6C the active patterns of five extensors of each stride are shown. And FIG. 6D shows the corresponding ground reaction force 650 based on the weight bw in which the ground reaction force of the left leg 660 is included for comparison.

Approaching from steady state walking (first stride), the model touches the stairs at the end of the second stride with the feet of the stretched right leg (FIG. 6A). This initial impact tilts the upper body forward, causing the model to slow down, which results in the large hip torque generated by the GLU and HAM (third stride, Figures 6b and 6c). Since hip extension torque also extends the knee, the VAS does not feel as much force as it does in the normal state, and its force feedback control degrades its muscle irritation (FIG. 6C), but pure work in the knee during the standing state It remains approximately the same as it was in the steady state. In contrast, the slowing of the model reduces the force felt by the ankle extensor GAS and SOL during the stance phase, and their force feedback reflexes result in somewhat less muscle stimulation, lowering the net work of the ankle (FIGS. 6B and FIG. 6). 6c). At the fourth and fifth strides, the model is stabilized with a stair climbing walk of about 1 ms ⁻¹ , with forward and rear thrust occurring primarily on the hips and knees. After reaching normal at the sixth stride, the model recovers the original steady-state stride rate, which is 1.3 ms ^-1 at the eighth stride.

7A-7D illustrate a step down adaptation in accordance with one aspect of the present invention, with a snapshot of the model (FIG. 7A), a pure day (FIG. 7B), an extensor activity pattern (FIG. 7C), and a corresponding ground reaction force (FIG. 7d). 7A-7D, approaching from a steady state walk of 1.3 ms ⁻¹ , eight strides of the human model are shown to cover five steps of 4 cm uphill each. The model recovers steady-state walking at the eighth stride. One stride is defined as the heel strike from the right leg to the heel strike. Shown in FIG. 7A is a snapshot of the model at heel-strike and toe-off of the right leg. For this leg, pure work during the standing state that occurs in the hip, knee and ankle with an extended positive action in FIG. 7B is shown, and the active pattern of five extensors of each stride is shown in FIG. 7C. And a corresponding ground reaction force 750 based on the weight bw in which the ground reaction force of the left leg 760 is included for comparison. The model recovers a steady state walk of 1.3 ms ^{−1 on} the fourteenth crawl after covering five steps with 4 cm downhill each.

7A-7D continue to illustrate the walking sequence where the model meets the down stairs. At the end of the ninth stride, the model reaches the first step with the right leg (FIG. 7A). The downward motion accelerates the model, causing the first shock of the right leg to be greater overall at the tenth crawl and the response of most extensors to be greater (FIGS. 7A-7D). Only the GAS generates less force because the knee stays more bent at this stride. As a result, the positive net work at the ankle substantially increases (FIG. 7B). This increase and greater HFL stimulation (not shown), caused by the anterior tilt of the upper body at take-off (FIG. 7A), propels the right leg forward in an angled state, increasing the step length. (FIG. 7A). After the transitional tenth stride, the model maintains a larger step in downward motion (eleventh and twelfth stride), where the downward acceleration of the model is due to the increased activity of GLU, HAM and VAS immediately following the impact. Disabled (FIG. 7C and FIG. 7D), this reduces the positive net work in the heap and increases the negative net work in the knee (FIG. 7B) and stabilizes the model with a downward walk of about 1.5 ms ⁻¹ . Returning to flat ground, the absence of downward acceleration slows the model, which automatically reduces the step length (FIG. 7A) and returns to steady state walking of 1.3 ms ⁻¹ within thirteenth and fourteenth steps.

In both climbing and descending stairs, a single control is not reliable. An important factor for acceptance and adaptation of the model is its dynamic muscle-reflex response. The stance of the standing foot depends on how much load the leg extensors SOL, GAS and VAS feel that allow the leg to bend sufficiently to allow forward progression when ascending but actually brake when descending. On the other hand, the angle of play is that the forward propagation of the bridge depends on model dynamics. Sudden deceleration after impact of the opposite leg, forward tilting of the upper body, and ankle extension near the end of the stance all contribute to the propagation of the leg. These combined features ensure that the angled leg extends sufficiently during walking up the stairs and substantially walking down the stairs. To this end, the force feedback of the GLU and HAM suppresses excessive rotation of the legs and instead quickly causes the legs to fold and straighten.

Muscle tendon unit. All tendon units (MTUs) of 14 bipedal animals have the same model structure. 8 is a schematic diagram of a muscle tendon model according to one aspect of the present invention. In FIG. 8, the active contractile element (CE) 810 forms a muscle tendon unit (MTU) in normal operation with the tandem elastic (SE) 820. Once the CE 810 extends beyond the optimal length l _CE 830 (l _CE > l _opt 840), the parallel elastics (PE) 850 are engaged. Conversely, if the SE 820 becomes loose (l _MTU 870-l _CE 830 <l _slack 880), the cushioning elastic portion (BE) 860 prevents the active CE 810 from collapsing.

As shown in FIG. 8, the active heel-type contractile element CE generates a force in line with the series elastic section SE. The MTU is fitted to the skeleton such that individual CEs act primarily on the rising limbs of the force-length relationship, but the MTU model includes parallel elastics (PE), which are coupled when the CE goes beyond the optimal length l _opt . In addition, the cushioning elastic portion BE ensures that the CE does not collapse when the leg geometry shortens the MTU to loosen. BE is a numerical tool that allows the MTU to describe loose muscles, eg loose GAS, when the knees are bent excessively. However, BE does not generate forces outside the MTU.

Table 3 shows the individual MTU parameters. All parameters are predicted from Yamamaguchi et al. [Yamaguchi, GT, Sawa, AG-U., Moran, DW, Fessler, MJ, winters, JM, 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 equivalent force (F _max ) is evaluated from individual or grouped muscle physiological cross-sectional areas with a force of 25 N / cm ⁻² . The maximum contraction velocity (V _max ) is set at 6L _opt S ^{- 1} for slow muscles and 12L _opt S ^-1 for muscles of medium pace. The optimal CE length l _opt and SE slack length l _slack reflect the muscle fiber and tendon length.

How CE and SE are modeled is described by Geeyer, H., Seyfarth, A., Blickhan, R., 2003. Positive force feedback in bouncing gaits? . R. Soc. Lond. B 270, 2173-2183. Force of _CE , F _CE = AF _max f _ℓ (ℓ _CE ) f _v (v _CE ) is the muscle activity (A), CE force-length relationship F _ℓ (ℓ _CE ) and CE force-velocity relationship F _v (v _CE ) Is the product of Based on this product method, the MTU kinetics can be calculated by integrating the CE velocity (v _CE ), which is obtained by reversing f _v (v _CE ). If F _SE = F _CE + F _PE -F _BE, then f _v (v _CE ) = (F _SE -F _PE + F _BE ) / (AF _max f ₁ (L _CE )). This equation has a numerical critical point during muscle morbidity when F _SE -F _PE approaches zero. To facilitate the simulation, this critical point is avoided by introducing F _v (v _CE ) into the parallel elastic part F _PE ˜ (L _CE −L _opt ) ² f _v (v _CE ). PE binds outside the normal operating range in the model and plays a small role for muscle dynamics during normal movement, like BE. However, with this method, f _v (v _CE ) = (F _SE + F _BE ) / (AF _max F ₁ (L _CE ) + F _PE ) is obtained, which is numerically obtained using an approximate time step. Can be integrated. This method is convenient for facilitating the simulation of the model, but is also important when muscle dynamics are emulated on a PC board with a fixed and limited time resolution.

MTUs have common and separate parameters. Common parameters include the time constant (t _ecc = 0.01) of the excitation contraction coupling; The width of the CE force-velocity relationship (w = 0.56 L _opt ) and the residual response factor (c = 0.05); Eccentric force improvement (N = 1.5) and shape factor (K = 5) of the CE force-velocity relationship; Includes SE reference elongation (ε _ref = 0.04) [For more information see: Geyer, H., Seyfarth, A., Blickhan, R., 2003. Positive force feedback in bouncing gaits?) Proc. R. Soc. Lond. B 270, 2173-2183. A common parameter is that PE reference elongation is ε _PE = w, where F _PE = F _max (ℓ _CE / ℓ _opt -1) ² / ε _PE ² f _v (v _CE ), and the BE idle length is ℓ _min = ℓ _opt -w, the reference compression is ε _BE = w / 2, where F _BE = F _max [(L _min −L _CE ) / L _opt ] ² / ε _PE ² . Individual MTU attachment parameters can be easily obtained from the paper and distinguish each muscle or muscle group. These values are listed in Table 4 (MTU Attachment Parameters).

Musculoskeletal connections and mass distribution. The MTU is connected to the skeleton by cutting one or two joints. The transformation from muscle force F _m to joint torque τ _m is modeled using variable lever arm r _m (φ) = r ₀ cos (φ-φ _max ) for the ankle and knee, where φ is the joint angle and φ _max Is the angle at which r _m reaches its maximum and τ _m = r _m (φ) F _m . For the heap, it is simply assumed that r _m (φ) = r ₀ . On the other hand, the change of the MTU length △ ℓ _m is modeled by _{△ ℓ m = ρr [sin (} φ-φ max -sin (φ ref -φ max)] to the ankle and the knee, for the heap △ ℓ _m = modeled as ρr (φ-φ _ref ) The reference angle φ _ref is the joint angle for ℓ _m = ℓ _opt + ℓ _slack The factor ρ is the muscle pennation angle and the fiber length of the MTU is It ensures that it stays within physiological limits throughout the operating range.The specific parameters of each muscle and joint are listed in Table 4. These values are empirical evidence ["muscle fiber and tendon length in human lateral muscles during slow pedaling". Muscle fiber and tendon length changes in the human vastus lateralis during slow pedaling ", Muraoka, T., Kawakami, Y., Tachi, M., Fukunaga, T., 2001. J. Appl. Physiol. 91, 2035 -2040; "Force-length characteristics of in vivo human skeletal muscle", Maganaris, C Acta Physiol.Scand. 172, 279-285; "Force-length characteristics of the in vivo human gastrocnemius muscle", Maganaris, C, 2003. Clin. Anat. 16 , 215-223; "In vivo lenth-force relationships on muscle fiver and muscle tendon complex in the tibialis anterior muscle", Oda, T., Kanehisa , H., Chino, K., Kurihara, T., Nagayoshi, T., Kato, E., Fukunaga, T., Kawakami, Y., 2005. Int. J. Sport and Health Sciences 3, 245-252, or obtained through a schematic anatomical evaluation.

The seven segments of the human model are simple rigid bodies whose parameters are listed in Table 5. The values are for example synthesized by Guenther and Ruder [Guether, M., Ruder, H., 2003. Synthesis of two-dimensional human walking: a test of the λ -model). Biol. Cybern. 89, 89-106. The segments are connected by a revolute joint. In relation to the human body, these joints have a free operating range (70 ° <φ _a <130 °, φ _k <175 ° and φ _h <230 °) with which mechanical soft limits are outside the range, It is modeled in the same way as the ground impact point. The segments of the model are the local center of mass (d _{G , S} ), and the joint position (d _{J , S} ) (measured from the distal end), and the different mass (m _S ) and length (l _S ) of inertia (Θ _S ). , And characteristic distance.

Ground contact and joint limits. Each foot segment of the bipedal model has a point of contact at the toe and heel. When impacting the ground, the contact point (CP) is pushed back by the vertical reaction force (Fy = -F _ref f _l f _v ) similar to the muscle force, which is the force-length relationship [ f _l (Δy _CP ) = Δy _CP / Δ _ref ] and the force-velocity relationship [f _v (dy _CP / dt) = 1-dy _CP / dt / v _max ]. This product approach to modeling vertical reaction forces is an existing approach that describes vertical forces as the sum of spring and nonlinear spring-damper terms [Scott, S., Winter, D., 1993. Biomechanical Model of Human Foot: Gait 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; Guether, M., Ruder, H., 2003. Synthesis of two-dimensional human walking: a test of the λ-model. Biol. Cybern. 89, 89-106. However, by separating the spring and damper items, the parameters of the contact model can be interpreted as two basic material properties, ground stiffness (k = F _ref / Δy _ref ) and maximum relaxation velocity (v _max ), It shows how quickly the shape can be restored afterwards. For example, v _max = ∞ describes a fully elastic ground shock that the ground always pushes against the CP, and v _max = 0, like sand, pushes the ground back out of the CP for downward velocity but backwards for upward velocity. Explain fully inelastic shocks that cannot be pushed out. Note that the impact model is used to account for the mechanical soft limit of the joints of the model (see previous item) with a soft limit stiffness of 0.3 N m deg ^-1 and a maximum relaxation rate of ¹ deg s ^-1 .

9 illustrates a contact model in accordance with an aspect of the present invention. In FIG. 9, contact occurs when contact point 920 falls below y ₀ (910). Similar to muscle force, the vertical reaction force (F _y ) of the ground is ground compression where F _y = F _ref when Δy _ref is dy / dt = 0, and dy _ref / dt is the maximum relaxation velocity of the ground (small figure). , Is modeled as the product of the force-length (f _l ) and force-velocity (f _v ) relationships. Initially, the horizontal reaction force F _x of the ground is modeled as sliding friction proportional to Fy with the sliding coefficient μ _sl . However, the contact point 920 decelerates below the minimum speed v _lim 930 and the horizontal model is converted to static friction 930. During static friction 930, F _x is also modeled as the product of the force-length and force-velocity relationships, which are different from those in the past to allow interaction with the ground in both directions around the static friction reference point (x ₀ ). Is a bit different. If F _x exceeds the limit static friction force (μ _st F _y ), the model switches back to sliding friction. Parameters: F _ref = 815N, Δy _ref = 0.01m, dy _ref / dt = 0.03ms ^-1 , Δx _ref = 0.1m, dx _ref / dt = 0.03ms ^-1 , v _lim = 0.01ms ^-1 , μ _sl = 0.8, μ _st = 0.9.

In addition to the vertical reaction forces, horizontal reaction forces are applied to the CP during ground contact. Initially, this force is modeled as a kinetic frictional force against the movement of the CP on the ground with a force (Fx = μ _sl F _y ). When the CP is decelerated below v _lim , the horizontal reaction force is modeled as a static friction force calculated in a similar manner as the vertical impact force is calculated (FIG. 9). If the static friction exceeds the limit force (F _lim = μ _st F _y ), the static friction changes back into kinetic friction. Thus, depending on the transition conditions, both types of horizontal reaction force are interchanged until the CP leaves the ground.

This result suggests that the machine and motor control cannot be seen separately for human movement. The neuromuscular model of human migration according to one aspect of the present invention is self-organized with gait steps after initial push, withstands abrupt changes in ground level, and adapts to stepped walking without intervention . The center of tolerance and compliance of this model is its dependence on muscle reflexes, which integrate sensory information about movement into the activation of leg muscles. Since there is no CPG, the model does not, in principle, require a central input to generate a walking movement, and the reflex input constantly adjusting between the nervous system and its mechanical environment may take precedence over the central input in the control of normal human motion. Suggest.

The model results also suggest that these continuous reflection inputs encode the principles of the gait mechanism. Current experimental and modeling studies of the function of spinal reflexes during exercise focus on their contribution to the generation of muscle forces at timing bearing and extensor loads in the altitude and standing state [Pang, MY, Yang, JF, 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. L, 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; 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]. Reflective contributions to load bearing began to translate into positive force feedback on the fundamental dynamics of the motor system [Prochazka, A., Gillard, D., Bennett, D., 1997. Positive force feedback control of muscles 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. It seems that there has never been a systematic expansion of the idea of encoding the principles of gait mechanics in motor control systems. Some muscle reflexes performed in the human model were a simple way to enter cyclic movements (body balance, incidence-leg initiation), but predominantly standing conditions corresponded to standing leg movements [Blickhan, R., 1989. Driving and hopping. 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 behavior explains the basic dynamics of walking and running. Proc. R. Soc. Lond. B 273, 2861-2867, Stabilization of segmented chains [Seyfarth, A., Guenther, M., Blickhan, R., 2001. Stable operation of an elastic three-segmented leg). Biol. Cybern. 84, 365-382; Guenther, M., Keppler, V., Seyfarth, A., Blickhan, R., 2004. Human leg design: optimal axial alignment under constraints. J. Math. Biol. 48, 623-646, and a 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, reflect the previously described control and walking mechanics and encoded principles of control. Based on these functional reflections, the model not only converges to the known joint angles and torque trajectories of human walking, but also predicts some individual muscle activation patterns observed in walking experiments. This matching between predicted muscle activation and observed muscle activation plays a greater role in motor control than the walking machine principle previously predicted, with muscle reflexes linking these principles to neural networks that respond to movement. I suggest you can.

In a preferred embodiment, the neurochanchanical model of the present invention has been implemented as a muscle reflex controller for powered ankle-foot prosthetics. This embodiment is an adaptive muscle-reflex controller based on simulation studies and utilizing an ankle plantar flexor that includes a Hill-type muscle with positive force feedback reflex. The parameters of the model were appropriate to match the torque-angle profile of the human ankle obtained from ground walking measurements of weight and height-matched intact subject walking at 1 m / sec. Using this single set of parameters, a transtibial amputee walk was performed at ground, ramp uphill, and ramp downhill conditions. During this test, adaptation of artificial ankle operation was observed in response to changes in ground slope in a way comparable to complete subjects, with no apparent terrain detection difficulties. In particular, the energy provided by the prosthesis was directly related to the ground tilt angle. This study highlights the importance of neuromuscular controllers to improve the adaptability of motorized prosthetic devices over variable surfaces.

To produce a controller with adaptive capability, a neuromuscular model with a positive force feedback reflex concept as the basis of the control of the present invention has been used as part of a control system for powered ankle-foot prostheses. The controller provided herein employs a model of the ankle-foot composite to determine the physical torque used at the ankle joint. In this model, the ankle joint is provided with two virtual actuators. In the case of plantar flexion torque, the actuator is a heel-type muscle with a positive force feedback reflex concept. This concept models the reflex muscle response due to some combination of afferent signals from the Golgi tendon organ and muscle spindle. In the case of dorsiflexion torque, the impedance is provided by an imaginary rotating spring-damper.

The parameters of this neuromuscular model are adjusted by the optimization procedure to provide the best match between the measured ankle torque of the complete object gait and the model's output torque when entering the measured object's complete movement at a target speed of 1.0 m / sec. lost. Neuromuscular model-based prosthetic controllers were used to provide torque commands to motorized ankle-foot prostheses worn by the amputator. This control strategy was evaluated using two criteria. First, the controller was tested for the ability to produce a prosthetic ankle torque and ankle angle profile that quantitatively matches the comparable torque and ankle angle of a complete subject at target ground walking speed. The second performance criterion was the controller's ability to exhibit a biologically consistent trend to increase gait cycle net work for increasing gait slope without changing controller parameters. It is difficult to detect changes in ground slope using conventional sensors, so a controller with inherent ability to adapt to these changes is particularly important.

10A-10C show a physical system (FIG. 10A), a view of a drive train (FIG. 10B), and a mechanical model (FIG. 10C) for an exemplary embodiment of ankle-foot prosthesis used in the preferred embodiment. The ankle-foot prosthesis used for this study is under development by iWalk, LLC. This prosthesis inherits a series of prototypes developed by the Biomechanical Group of the MIT Media Laboratory, described in U.S. Provisional Patent Application No. 12 / 157,727, filed June 12, 2008, which is incorporated herein in its entirety. Included by reference. Prostheses are fully self-contained devices with the size and weight (1.8 kg) of the complete biological underfoot-foot complex. 10A shows a housing 1005, ankle joint 1010, foot 1015, unidirectional parallel leaf spring 1020, and tandem leaf spring 1025 for motors, transmission, and electronics. 10B includes a timing belt 1030, pin joint main housing 1035, motor 1040, ball screw 1045, ankle joint 1010, ball nut 1050 pin joint (serial spring) 1055, and Foot movement indicator 1060 is shown. The mechanical model of FIG. 10C includes a parent link 1065, a motor 1040, a transmission 1070, a series spring 1025, a unidirectional parallel spring 1020, a foot 1015, a series spring moment arm ( r _s ) 1075, spring rest length 1080, and SEA 1085 are shown. Rotational elements of the physical system are shown as linear equivalents in the model conceptual diagram for clarity.

Ankle joint is a rolling bearing design that connects the lower foot structure to the upper leg shin structure, topping with a prosthetic pyramid fixture to attach to the socket of the cutter. The foot includes a passive low profile Flex-Foot ™ (Osur ™) to minimize ground contact impact to the cutter. Unidirectional leaf springs, parallel springs, act across the ankle joint, which engages when the ankle and foot are perpendicular to each other. The spring acts parallel to the powered drive train providing the manual function of the Achilles tendon. The powered drive train is a motorized link across the ankle joint as shown in FIG. 10B. In series from the upper leg shank, the powered drive train is a brushless motor (Powermax EC-30, 200 Watt, 48V, maxon) operating at 24V, a belt drive transmission with 40/15 reduction ratio, and a 3mm pitch linear ball. It is made of screws. At this operating voltage, the theoretical maximum torque that can be generated by the motor through the drive train is approximately 340 Nm.

At the foot, the tandem spring, the Kevlar-composite leaf spring, connects the foot to the ball nut with a direction dependent moment arm (r _s ). Thus, the effective rotational stiffness of the tandem spring evaluated by locking the drive train and torqueing around the ankle joint is 533 Nm for positive torque and 1200 Nm for negative torque and positive torque (or plantar flexion). Torque) presses the tandem spring as shown in FIG. 10C. The drive train and series spring together comprise a series elastic actuator (SEA) [GA Pratt and MM Williamson, "Series elastic actuators," Proceedings on IEEE / RSJ International Conference on Intelligent Robots and Systems, Pittsburgh, pp. . 399-406, 1995]. An arrangement of these components is shown schematically in FIG. 10C.

sensor. The hall effect angle sensor is the primary control input at the ankle joint and has radians in the range -0.19 to 0.19, with zero corresponding to the foot perpendicular to the shin. The joint angle is measured with a linear Hall effect sensor (Allegro A1395) mounted in the main housing. The sensor is close to a magnet that is firmly connected to the foot structure such that the magnetic axis is perpendicular to the arc of movement of the magnet. As a result of this arrangement, the magnetic field strength changes at the sensor position as the magnet rotates past the sensor. The strain gauge is located inside the prosthetic pyramid attachment, allowing measurement of torque at the ankle joint. The strain gage located in the series spring allows the detection of the output torque of the motorized drive train, thereby allowing closed loop force control of the SEA. The motor itself includes a Hall effect communication sensor and is fitted with an optical shaft encoder that enables the use of advanced brushless motor control technology.

Microcontroller. Overall control and communication for the ankle-foot prosthesis is provided by a single chip, 16 bit, DSP oriented microcontroller, Microchip Inc. dsPIC33FJ128MC706. The microcontroller has 128 kilobytes of flash program memory, and 16384 bytes of RAM to perform 40 million instructions per second. The microcontroller provides the appropriate operation to support real time control.

Motor controller. A second 16-bit dsPIC33FJ128MC706 is used as the dedicated motor controller. The high computational load and speed requirements of modern brushless motors control the methodology in addition to the separation of work from the real-time demands of key microcontrollers that motivate these structures. The high speed digital link between the main microcontroller and the motor microcontroller provides the virtually instantaneous command of the motor.

Wireless interface. The microcontroller's high-speed serial port is dedicated to external communication for deployment and data collection. Such ports may be used directly via cables or may have a wide variety of wireless communication devices attached. In the present study, 500 Hz sensor and internal state information is measured over a serial port of 460 kilobolo and transmitted over an IEEE 802.11g wireless local area network device (Lantronix Wiport).

battery. All power for the prosthesis is supplied by a 0.22 kg lithium polymer battery with an energy density of 165 Watts / H / kg. The battery can provide daytime power demands, including 5000 steps of motorized walking.

Optimum mechanical part selection. Meeting the requirements for mass, magnitude, torque, speed, energy efficiency, shock tolerance, and nearly silent operation is not a trivial task. Of course, optimization and modeling of drive trains for the movement of walking and the production of biological torques is important. Some of the effects of motor selection, overall transmission rate, series elastic springs, and parallel springs are described in IEEE Robotics & Automation Magazine, September 2008, pp. 52-59; AY S. On the Design of a Powered Ankle-Foot Prosthesis: The Importance of Written by SK Au and H. Herr. Series and Parallel Elasticity ".

Control architecture. The purpose of the control architecture is to command ankle torque appropriate for the gait cycle of the amputee, determined from possible sensor measurements of the prosthetic ankle state. The controller uses the appropriate torque using a neuromuscular model of the human ankle-foot complex. In this model, the hinge joint representing the human ankle joint is bidirectional with two competing virtual actuators, the unidirectional plantar flexor and gait phase, a heel shaped muscle model. It is actuated by a proportional differential position controller or a dorsiflexor acting as a unidirectional virtual rotary spring damper. The finite state machine maintains an estimate of the cutter's gait cycle. Depending on the estimated gait cycle, one or the other, or both, of the virtual actuators generate torque in the virtual ankle joint. The total virtual torque is then used as ankle torque to command the prosthetic hardware. Physical torques at the ankle joint are produced by both motorized drive series and parallel springs. The ankle angle sensor is used to determine the torque generated by the parallel springs and the remaining predetermined torque is commanded through the motor controller.

Top level state machine control. The highest level of control of the prosthesis is realized by a finite state machine that allows concurrency in the walking cycle. During walking, two states are known: swing phase and stance phase. Prosthetic sensor inputs (ankle torque, ankle angle, and motor speed estimated from pyramidal strain gauges) are continuously observed to determine state transitions. The conditions for this state transition were determined experimentally. 11 illustrates the operation of the transition condition and state machine. The dorsal and plantar flexor virtual actuators develop torque from the state machine according to the walking state estimate.

In FIG. 11, the angle state 1110 is visually represented as SW 1120 and the elevation 1130 is controlled plantar flexion (CP) 1140, controlled flexion flexion (CD) 1150, and powered. Plantar flexion (PP) (1160). State transitions 1170 and 1180 are determined using the prosthetic ankle torque T _p and the prosthetic ankle angle θ measured from the pyramidal strain gauge.

When the foot leaves the ground, the transition to the angle of incidence is measured with ankle angle less than -0.19 radians to prevent descent to total ankle torque less than 5 N · m or ankle sensor concentration measured using a pyramidal strain gauge. It is detected by a drop in (θ). Positive torque is defined as the actuator torque of the tendency to plantar flex the ankle, and the positive angle corresponds to flexion flexion. In order to prevent premature transition, the ankle torque developed during the standing phase should allow this transition to exceed 20 N · m. In addition, a buffer time of 200ms provides a minimum time during the standing period. Transition to the heel-strike stance is detected by a reduction to torque lower than -7 Nm measured using a pyramidal strain gauge.

A block diagram of an exemplary embodiment of a control system for a prosthetic prosthesis according to this aspect of the invention is shown in FIG. 12. Neuromuscular model 12010, parallel spring model 1220, lead compensator 1230, friction compensator 1240, motor controller 1250, and prosthesis 1260 (shown as a mechanical model according to FIG. 10C) are shown in FIG. 12. Is shown.

는 신경근 모델로부터의 토크 명령(τ_d)을 생성하는데 사용된다. 이러한 소정의 발목 토크는 보철 작동기로의 현재 명령을 획득하기 위해 토크 제어 시스템을 통해 공급된다. 이러한 토크 제어 시스템의 3개의 주요 성분으로는 피드포워드 이득(feedforward gain; K_ff), 리드 보상기, 및 마찰 보상 항(term)이 있다. 소정의 작동기 토크(τ_{d, SEA})를 획득하기 위해 보철 발목 토크로의 병렬 용수철 기여(τ_p)는 소정의 발목 토크로부터 제외된다. 이후, 폐루프(closed-loop) 토크 제어기는 측정된 작동기 토크(τ_SEA)를 이용하여 소정의 작동기 토크를 실행시킨다. 마지막으로, 마찰 보정 항은, 폐루프 토크 제어기의 출력에 부가된 추가적인 토크 값(τ_f)을 생성한다.Prosthetic Measurement Ankle Condition

Is used to generate the torque command τ _d from the neuromuscular model. This predetermined ankle torque is supplied through the torque control system to obtain a current command to the prosthetic actuator. Three main components of this torque control system are feedforward gain (K _ff ), lead compensator, and friction compensation term. The parallel spring contribution τ _p to the prosthetic ankle torque is subtracted from the predetermined ankle torque to obtain the desired actuator torque τ _{d, SEA} . The closed-loop torque controller then executes the predetermined actuator torque using the measured actuator torque τ _SEA . Finally, the friction correction term produces an additional torque value τ _f added to the output of the closed loop torque controller.

13A-13C show one complete gait cycle during three gait conditions: flat ground (FIG. 13A), ramp ascent (FIG. 13B), and ramp descent (FIG. 13C). This is a graphical representation of prosthetic torque over the same heel touch to the heel touch). Shown in each figure are prosthetic torques estimated using torque averages 1305, 1310, 1315 (thin line) ± standard deviation (hidden line), and angle-based estimate of parallel spring torque contribution and measured SEA torque contribution. (1320, 1325, 1330) (thick lines). Vertical (dashed) lines 1335, 1340, and 1345 represent the terminations of the standing positions.

Dorsal muscle model. 14A-14C show spring damper attachment to two link ankle joint models and an heel shaped muscle model (FIG. 14A), a detailed heel shaped muscle model (FIG. 14B), and an ankle coordinate frame and various moments for the muscle model. An exemplary embodiment of a musculoskeletal model implemented in a prosthetic microcontroller including the shape of a muscle model skeletal attachment including arm transitions (FIG. 14C) is shown. Shown in FIGS. 14A and 14C are mechanical views of the dorsal flexor (spring-damper) 1405, plantar flexor (MTC) 1410, foot 1415, shank 1420, and heel 1425. to be.

의 설정값을 갖는 가상 회전식 스프링 댐퍼 및 관계식(1)에 의해 구현된다.The dorsal muscle of Fig. 14A is a dorsal muscle actuator. This represents Tibialis Anterior and other biological flexor muscles. These dorsal muscles,

It is implemented by a virtual rotary spring damper having a set value of and relation (1).

는 발목 각도이고,

는 발목 각속도이다. 입각기에서, K_p의 값은, 통상의 지면 레벨 보행에 대한 생물학적 발목의 입각기 행위를 최적으로 적응시키기 위해 다른 근육 모델 파라미터와 함께 최적화된다. 댐핑 항(K_v)은 앞발이 발평면부(foot-flat)의 지면에서 튀는 것을 막기 위해 입각기 동안 5Nm·s/rad로 실험적으로 조정되었다. 또한, 입각기 동안, 배굴근은, 생물학적 근육의 단일 방향 특성을 모방하기 위해 배굴근 토크를 제공하기 위해서만 작용한다. 또한, 배굴근에 의해 생성된 토크가 입각 동안 발이 정강이에 수직이 되게 됨에 기인하여 0으로 떨어질 때, 배굴근은 입각기의 나머지에 장애를 입히게 된다. 따라서, 인간 배굴근 근육이 중요한 역할을 하는 것으로 알려져 있을 때, 배굴근은 단지 입각기의 초기의 토크 생산에만 기여한다[1992년 발행된 뉴저지 슬랙 아이엔씨.(New Jersey: SLACK Inc.)의 4장, 55-57쪽; 페리 제이.(J. Perry)의 "걸음걸이 분석: 통상적이고 병리학적인 기능(Gait Analysis: Normal and Pathological Function)" 참조]. 유각기에서, 배굴근은 설정값

로 발을 구동시키는 위치 제어기로서 작용한다. 여기서, K_p=220N·m/rad 및 댐핑 상수 K_v=7N·m·s/rad의 이득은 유각기에서의 초기의 발의 빠른 지상고(ground clearance)를 제공한다.Where K _p is a spring constant, K _v is a damping constant,

Is the ankle angle,

Is the ankle angular velocity. In standing, the value of K _p is optimized along with other muscle model parameters to optimally adapt the standing behavior of biological ankles to conventional ground level walking. The damping term (K _v ) was experimentally adjusted to 5 Nm · s / rad during standing to prevent the forefoot from bouncing off the ground of the foot-flat. Also, during the standing phase, the dorsal muscles act only to provide the dorsal muscle torque to mimic the unidirectional properties of biological muscles. In addition, when the torque generated by the dorsal muscle drops to zero due to the foot being perpendicular to the shin during standing, the dorsal muscles will cause the rest of the standing muscles to fail. Thus, when human dorsal muscles are known to play an important role, dorsal muscles only contribute to the initial torque production of the standing phase [4 of New Jersey: SLACK Inc., published in 1992]. Chapters, pp. 55-57; See J. Perry's "Gait Analysis: Normal and Pathological Function". At the drone, the flexor muscle is set

Acts as a position controller to drive the foot. Here, the gain of K _p = 220 N · m / rad and the damping constant K _v = 7 N · m · s / rad provides a fast ground clearance of the initial foot in the staging phase.

Plantar flexor model. The virtual plantar flexors of FIGS. 14A-14C include muscle tendon complex (MTC) showing a combination of human plantar flexors. MTC, The Netherlands Proc. IEEE Int. Conf. On Rehabilitation Robotics, pp. 298-303; AY S. Based on the "Biomechanical design of a powered ankle-foot prosthesis" written by K., J. Weber, and H. Herr. This is described in more detail below. It consists of a contraction element (CE) modeling muscle fibers and a series element (SE) modeling the gun. The contraction element consists of three unidirectional components: a heel shaped muscle with a positive force feedback reflective scheme, high limit parallel elasticity, low limit or parallel elasticity of the buffer. A series element, a nonlinear unidirectional spring representing an Achilles tendon, is in line with the contraction element. The attachment shape of the root canal composite to the ankle joint model is nonlinear, complicating the calculation of torque resulting from the actuator force.

Plantar flexors serial elastic elements. The tandem elastic element (SE) acts as a line in line with the muscle contraction element [Proc. R Society. Lond. Pp. 270, pp. 2173-2183; "Positive force feedback in bouncing gaits?" Written by H. Geyer, A. Seyfarth, and R. Blickhan. )" Reference]. Ε obtained is the tendon tension, which is defined as follows.

Where l _SE is the length of the serial element and l _slack is the remaining length of it. The serial element was published in 2003 in Proc. R Society. Lond. Pp. 270, pp. 2173-2183; It is embodied by the nonlinear spring described in "Positive force feedback in bouncing gaits?" Written by Gay H., Seyphas A., and Brick Egg.

Where F _max is the maximum isometric force the muscle can exert. Proc. R Society. Lond. Pp. 270, pp. 2173-2183; In "Positive force feedback in bouncing gaits?" Written by Gay H., Seyphas A., and Brick R., this form of quadratic equation is typically modeled distinctive. Used as an approximation of an exponential linear stiffness curve. This approximation is made to reduce the number of model parameters.

Plantar flexor contraction element. The contraction element CE (see FIG. 14B) of the plantar flexor virtual actuator is a heel shaped muscle model with a positive force feedback reflection scheme. This is, H. "A muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities", published by Her H. Two parallel elastic parts and active muscle fibers generating force. Heel-shaped muscle fibers exert a unidirectional force. This force is a function of muscle fiber length (I _CE ), velocity (V _CE ), and muscle activity (A). The final force (F _MF ) was published in Proc. R Society. Lond. Pp. 270, pp. 2173-2183; It is defined by the following equation as described in "Positive force feedback in bouncing gaits?" By Gay H., Seyphas A., and Brick Egg.

The force-length relationship [f _L (I _CE )] of the heel shaped muscle is a bell-shaped curve defined by the following equation.

Where I _opt is the contractile element length I _{CE at} which the muscle can provide the maximum isotonic force F _max . Parameter w is the width of the bell curve and parameter c represents the magnitude of the curve near the extreme of the bell.

The force-velocity relation of CE [f _v (V _CE )] is the heel equation.

Where V _max <0 is the maximum contraction rate of the muscle, V _CE is the rate of fiber contraction, K is the curvature constant, and N is the slight muscle force (standardized by F _max ).

The following is H. "A muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities", published by Her H. The force-length relationship for the high limit parallel elasticity (HPE) established in response to CE is defined as follows.

Lower limit buffer parallel elasticity (LPE) also gains. "A muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities", published by Her H. Submitted for). This is given by the following nonlinear spring shape.

Thus, the total plantar flexion force is described as follows.

Where F _CE is the force developed by the contraction element. Since CE and SE are continuous, the following equation is maintained.

Reflective Schematic. Shrinkage element activity (A) is generated using the positive force feedback reflection scheme shown in FIG. 15 [gay H. "A muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities", published by Her H. Submitted for) and Proc. R Society. Lond. Pp. 270, pp. 2173-2183; See “Positive force feedback in bouncing gaits?” Written by Gay H., Seyphas A., and Brick R.]. FIG. 15 illustrates an exemplary embodiment of a reflective scheme for virtual plantar flexor muscles, including relationships between ankle angle, muscle force, and plantar flexor component of ankle torque.

As shown in FIG. 15, this feedback loop includes a stance transition to impede plantar flexion force development during the stray phase. During standing, plantar flexor force F _MTC is multiplied by the reflection gain Gain _RF delayed by Delay _RF and added to the offset stimulus PRESTIM to obtain a neural stimulus signal. The stimulus is defined in the range of 0 to 1 and is low pass with a time constant (T) to stimulate muscle excitation-contraction coupling. The final signal is used as the activation of equation (4) with an initial value of PreA. In addition, following H. "A muscle-reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities", published by Her H. The suppression gain (Gain _supp ) described in (Submitted for) is implemented to help prevent two actuators from interfering with each other during _standing . At this time, the torque produced by the flexor times is reduced until the value or by Gain _supp · F _{M TC} falls to zero.

Planar flexor shape and implementation. Within the muscle model framework, the ankle angle θ _foot is defined as shown in FIG. 14C. Using these angles entered into the model, the length of the tendon complex was written by H. Geyer, H. Herr, “A muscle reflex model representing the principles of leg dynamics predicts human gait dynamics and muscle activity. reflex model that encodes principles of legged mechanics predicts human walking dynamics and muscle activities "(submitted for publication).

Where ρ represents the pennation angle of the muscle fiber and φ _ref is the ankle angle at 1 _CE = l _opt when there is no load.

The fiber length l _CE can be calculated from l _CE = l _MTC -l _SE , where l _SE is the current value F _CE = F _SE = F _MTC based on muscle dynamics. Can be obtained from the inverse. The shrinkage rate (V _CE ) of the fiber can be obtained by differential method. This produces a first order differential equation determined by the dynamics of the neuromuscular model. This equation can be found a value for F _MTC given the initial condition and the time history of θ _{f o ot} the ground. However, because the integral method is more computationally effective than the differential method, the muscle reflex model that expresses the principles of leg dynamics published by H. Geyer, H. Herr, predicts human gait dynamics and muscle activity (A muscle-reflex model The integral form of this process was used to obtain F _MTC as described in that encodes principles of legged mechanics predicts human walking dynamics and muscle activities.

Knowing the attachment radius, r _foot , and angle φ _max at the maximum tendon moment arm, the relationship between the final plantar flexor and F _MTC for the ankle torque T _plantar is derived by the equation below.

T _plantar = F _MTC cos (θ _foot ~ Φ _max ) r _foot = F _MTCR (θ _foot ) (13)

Where R (θ _foot ) is the variable moment arm resulting from muscle attachment to the ankle joint model. This relationship is shown graphically in FIG. That is, the plantar flexor model can be treated as a dynamic system that essentially links a single input value θ _foot to a single output value T _plantar .

Parameter determination of neuromuscular model. Plantar flexor model represents all plantar flexors. Similarly, the dorsal muscles represent all living dorsal muscles. In this work, joint and torque measurements were taken only at the ankle joint. As a result, the state of articular muscles such as gastrocnemius muscle could not be estimated accurately. Thus, plantar flexors are based on the predominant mono-articular plantar flexors, or soleus, which occupies most of the human. Thus, most plantar flexor parameters are based on the plantar muscles published by H. Geyer, H. Herr, "A muscle-reflex model expressing the principles of leg dynamics predicts human gait dynamics and muscle activity (A muscle-reflex model). that encodes principles of legged mechanics predicts human walking dynamics and muscle activities ". However, some parameters of plantar flexors and parameters for dorsal flexors were thought to be significantly influenced by the previous model or not well known in biology. These six parameters were determined by a combination of genetic algorithms and gradient descent methods to ensure that the neuromuscular model fits well with the walking data of intact subjects.

Unoptimized parameter values are shown in Table 6.

Data collection for non-cutting targets. Kinematics and kinematic gait data were collected in a traveling laboratory at Harvard Medical School's Spaulding Rehabilitation Hospital through research into human use as an experimental subject approved by the Spaulding committee. This is known in "Angular momentum in human walking" published by H. Herr, M. Popovic in "The Journal of Experimental Biology, Vol. 211, pp 487-481, 2008". A healthy adult male (81.9 kg) was required to walk at slow motion along a 10m footpath in a motion capture laboratory with prior consent.

Motion capture was performed using the VICON 512 motion capture system with eight infrared cameras. The infrared camera tracked the location during the test as the reflective marker was placed at 33 locations on the subject's body. The camera was operated at 120 Hz and was able to track a given marker within approximately 1 mm. The markers were placed on the following bone targets to track the lower body: bilateral anterior superior iliac spines, posterior superior iliac spines, lateral femoral condyles, and lateral ankle bones. , Forefoot and heel. A wand was placed over the tibia and the femur and a marker was attached to the circle located over the middle bone of the tibia and middle femur. The marker was also placed at the following points in the upper body: the sternum, clavicle, C7 and T10 lumbar vertebrae, head, and both sides of the shoulder, elbow and lumbar joints.

Ground repulsion was measured using two staggered force plates, Model No. 2222 or OR6-5-1, commercially available from Advanced Mechanical Technology, Inc., Watertown, Massachusetts, USA. The accuracy of these measuring plates for measuring ground repulsion and center of pressure is approximately 0.1 N and 2 mm, respectively. Force plate data was collected at 1080 Hz and synchronized with VICON motion capture data. Joint torque was calculated from ground repulsion and joint kinematics using a modified version of the standard dynamics model. Commercially available Vicon Bodybuilder by Oxford Matrix, UK, was used to perform dynamic dynamics calculations.

Six tests were selected for slow walking speed on the ground (average 1.0 m / s) and one test was used to represent the target ankle and torque trajectory for this walking condition. The end of the stance was defined as the point over time when the torque dropped to zero after the peak had reached its peak in the walking cycle. This case occurred at a walking cycle of 67% for the selected test.

16A and 16B show torque and angular trajectories of the subject's living ankles with intact limb weight and height, and torque and angular trajectories measured through testing on the prosthesis of the subject undergoing amputation. 16A and 16B show ankle torque (FIG. 16A) and ankle angle (FIG. 16B) for the ground walking cycle from the start of the heel to the toe drop (100% cycle) to the toe drop (100% cycle). 16A and 16B plot the mean (1610, 1620 (thin line)) ± 1 standard deviation (solid line) for the measured torque and angular profile for the prosthesis and the measured torque of the prosthesis resulting from neuromuscular model control. And averages 1610, 1620 for ankle biomechanics 1630, 1640 (bold line) for the subject's gait cycle that conformed to weight and height with intact limbs at an angle profile and the same walking speed (1 m / sec); Thin line) ± 1 standard deviation (dotted line) is shown graphically in FIGS. 16A and 16B. The vertical line represents the average time of the initial stages of the stages 1650, 1660 (thin dashed line) for the prosthetic gait cycle, and the initial stage of the stages of the living ankle stages 1670, 1680 (bold dotted line).

Adjustment of model parameters to experimental data through optimization techniques. The following parameters were selected to adjust F _max , Gain _FB , Gain _supp , Φ _ref, and Φ _max . The goal of parameter adjustment was to find a parameter setting that would allow the neuromuscular model to better match the bioankle torque trajectories under certain walking conditions given the corresponding bioankle angle trajectories as inputs to the model. The cost function for optimization was defined as the squared error between the biotorque profile and the model torque profile during the standing phase, given the bioankle angle trajectory.

(14)

(14)

T _m is the torque output of the model and T _bw is the living ankle torque.

Optimization techniques using genetic algorithms were selected to perform an initial search for optimal parameter values, and direct search was included to pinpoint an optimal set of parameters. Genetic algorithmic tools from Matlab program were used to perform both optimization methods. Human ground walking data at walking speed selected at 1.0 m / s was used to provide target behavior for optimization. The allowable range of each of the optimization parameters is shown in Table 7 (range of optimization parameters).

The initial population was selected by the optimization program. The parameter values derived from parameter optimization are shown in Table 8 (adjusted values of neuromuscular model parameters).

Result of parameter optimization. In validating the optimization, the optimization was applied to the final parameter using the bio-ankle angle profile as input to the neuromuscular model. The biotorque profile and the final torque profile are shown in comparison to FIG. 17.

As shown in FIG. 17, bioankle moments (gray line, 1710), modeled dorsal flexor component (dotted dashed line, 1720), modeled plantar flexor component (thin line, 1730), whole nerve root model (flexible and flexor muscle) ) Moment (dashed line, 1740) is compared with the ankle moment profile from the intact living ankle to the ankle of the neuromuscular model with the bioankle angle profile as input and optimized parameter values. The ankle moment of the neuromuscular model almost exactly matches the bioankle moment for most walking cycles.

Low level torque control. The physical torque actually occurring at the ankle joint during the standing phase comes from the joint action of the parallel spring and the motor drive train. The stiffness of the rotary parallel spring with a spring stiffness of 500 Nm / rad appears to be approximately linear in the operating range. Using this spring constant, the parallel spring contribution is predicted and subtracted from the desired ankle torque. The remaining torque is produced by the motor drive train.

The performance of the motor drive train is improved by using lead compensation, friction compensation and feed-forward techniques, as shown in FIG. Experimental studies of open loop drive train dynamics have been performed and used to implement these improvements [M. Eilenberg, "A Neuromuscular-Model Based Control Strategy for Powered Ankle-Foot Prostheses," Master's Thesis, Massachusetts Institute of Technology, Cambridge, MA, 2009]. Output torque versus command torque as ground walking, rising slope and falling slope are shown in FIGS. 13A-13C. The prosthetic output torque is estimated using the stress gauge of the serial spring to SEA torque contribution and the parallel spring torque estimate based on the ankle angle to the parallel spring torque contribution.

Clinical evaluation. Prostheses were performed on the right foot of a healthy, active 75 kg lower leg patient. Subject was given time to walk naturally through the prosthesis. A radio link to the prosthesis was used to record walking data from these tests. During the ground walking test, the subject was asked to walk across a 10 m long path. The intended target walking speed was set at 1.0 m / s to match the walking speed of the intact object. Subject began walking approximately 5 meters from the beginning of the path and stopped walking approximately 3 meters past the end of the path. Marks on the ground were used to indicate the beginning and end of the 10 m of the path. A stopwatch was used to confirm the average walking speed for each test by informing each subject as the center of gravity passed each mark. A total of ten tests were set. Tests with walking speeds within 5% of the target speed were used for processing, resulting in 45 walking cycles. The subject was then asked to walk on a 2 m ramp, at an angle of 11 degrees, at his or her own chosen speed. Subject began walking on the ground approximately 2m from the beginning of the ramp during 10 climb ramp tests and stopped walking approximately 1m past the ramp on the platform. This same process was reversed during the 12 descent ramp tests.

Data analysis. In the ground test, the first three walk cycles and the last three walk cycles were assumed to be temporary and were ignored. Each remaining walk cycle was resampled to span 1000 data points. Mean and standard deviation trajectories were calculated from the final data. For ascending and descending ramps, the last step of the ramp was used as a representative walking cycle. Each walk cycle selected was resampled and averaged in the same manner as described in connection with the ground experiment.

The network was calculated for each individual gait cycle by numerically integrating ankle torque against ankle angle from the start of the heel to the fall of the toe. At this point, the vane was ignored in network calculations. The average network for each walking condition was calculated from the individual walking cycle network values.

result. Torque tracking. The precondition for this experiment was the ability of the ankle-foot prosthesis to actually produce the torque and speed commanded by the neuromuscular controller. This capability is shown in FIGS. 13A-13C which show the command torque for the measured output torque of ground walking, rising ramp and falling ramp.

Adaptation to ground slope. Ground slope adaptation assessment of neuromuscular model control prostheses was confirmed by the clinical trial data of FIGS. 9A-9C. The numerical integration data of these tests provided network values (work loop area) as follows.

5.4 ± 0.5 joules above ground

Rising slope 12.5 ± 0.6 joules

Down slope 0.1 ± 1.7 joules

Comparison for living ankles. The purpose of this neuromuscular model is to present the intrinsic dynamics of the human ankle-foot complex in a useful way. Thus, the final prosthetic controller can be evaluated based on its ability to mimic human behavior. 16A and 16B described above show ground walking torque and angular profiles from prostheses along with ground walking torque and angular profiles of subjects consistent with weight and elongation with intact limbs.

18A-18C are plots of prosthetic torque-angle trajectories measured for three different walking conditions: ground (FIG. 18A), rising slope 18b and falling slope 18c. 18A-18C show the mean 1810, 1820, 1830 ± 1 standard deviation. Arrows indicate forward propagation over time. The average prosthetic network increases with increasing ground slope. This result is described in A. S. Mclntosh, 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].

The measured ankle torque and ankle angle profile of the prosthesis is quantitatively consistent with each comparable intact ankle torque and ankle angle profile for ground walking. There were few differences that could significantly affect a number of factors, including atrophy and hypertrophy of the leg muscles of the clinical subjects due to amputation, differences in limb lengths, and lack of functional biaxial joint gastrocnemius muscle. In addition, the limited range of prosthetic angle sensors prevented the prosthesis from reaching the full range of intact ankle movements.

Ground slope adaptation. The neuromuscular control provided herein represents an inherent adaptation to ground tilt without explicit sensing of the terrain. Increased ankle net work during ramp up and decreased ankle net work during ramp down is [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, consistent with the behavior of a complete human ankle under the same conditions. This variation in the standing positive net work over walking conditions represents the urgent slope-adaptation behavior of the neuromuscular model. The ability of the neuromuscular model to generate these biomimetic changes in behavior suggests that the model implements important features of the human plantar flexor. It is also envisaged that the model has the potential for speed adaptation. In an attempt to move faster, the user may press harder on the prosthesis. This additional force causes reflexes that are modeled to command higher virtual muscle forces, resulting in greater energy output, resulting in higher walking speeds.

Although preferred embodiments are disclosed, many other embodiments may be practiced by those skilled in the art within the scope of the present invention. Each of the various embodiments described above can be combined with other described embodiments to provide a number of features. In addition, the foregoing describes many distinct embodiments of the apparatus and method of the present invention, and those described herein are merely illustrative of the application of the principles of the present invention. Therefore, other arrangements, methods, modifications and substitutions by those skilled in the art are also contemplated within the scope of the invention, which is not limited except as by the claims below.

Claims (9)

Model-based neuromachine controller for robotic limbs with one or more joints,
A finite state machine configured to receive feedback data relating to the state of the robot limb and to determine the state of the robot limb using the received feedback data;
A muscle model processor configured to receive state information from the finite state machine and muscle shape and reflex architecture information from one or more databases, and determine one or more desired joint torque or stiffness commands to be sent to the robot limb using the neuromuscular model;
A joint command processor configured to command the biomimetic torque and stiffness determined by the muscle model processor in the robotic limb joint;
Model-based Neural Machine Controller. The method of claim 1,
Feedback data is provided by one or more sensors mounted at each joint of the robot limb.
Model-based Neural Machine Controller. The method of claim 1,
Robotic limbs are legs and finite state machines are synchronized to the gait cycle
Model-based Neural Machine Controller. The method of claim 3,
Legs include motorized ankle-foot prosthetics
Model-based Neural Machine Controller. The method of claim 3,
The leg contains the knee joint
Model-based Neural Machine Controller. The method of claim 4, wherein
The legs contain more knee joints
Model-based Neural Machine Controller. The method of claim 6,
The legs contain more hip joints
Model-based Neural Machine Controller. The method of claim 2,
One or more sensors may be angular joint displacement and speed sensors, torque sensors, or inertial measurement units.
Model-based Neural Machine Controller. To control robotic limbs with one or more joints,
Receiving feedback data relating to the state of the robot limb in a finite state machine,
Determining the state of the robot limb using the finite state machine and the feedback data received;
Using the neuromuscular model, muscle shape and reflex architecture information, and state information from a finite state machine, determining one or more desired joint torque or stiffness commands to be sent to the robot limb;
Instructing the biomimetic torque and stiffness determined by the muscle model processor in the robotic limb joint.
Robot limb control method.

Images