JP2013503026A - Implement a sequence of standing up using a prosthetic leg or lower limb orthosis - Google Patents

Abstract

A knee brace or prosthesis can be used automatically when it is appropriate to initiate a sequence of standing up based on the position of the person's knee relative to the person's ankle while the person is sitting. When the knee is moved to a position that is in front of the ankle joint, at least one actuator of the brace or prosthesis is activated to help lift the person from sitting to standing. The at least one sensor includes an inertial sensor that measures a gravity vector relative to the femoral coordinate system. At least one sensor detects a rotation angle of the lower limb with respect to the thigh.
[Selection figure] None

Description

(Related application)
This application claims the benefit of US Provisional Application No. 61 / 238,305, filed Aug. 31, 2009.

  The present invention generally relates to a prosthesis, a lower limb orthosis and an exoskeleton device, components thereof, and a prosthesis, a lower limb orthosis and an exoskeleton device, and methods for controlling the components.

  Within the United States, more than 100,000 people lose their legs due to amputations each year. Hundreds of thousands of people around the world are suffering from this loss of debilitation. In addition, every year in the United States, there are 700,000 people who survive strokes that often cause various pathologies in the lower limbs that restrict locomotion. Until recently, prosthetic and lower limb orthosis systems were not stored on a step-by-step basis required by the body to perform lean walking movements, even on flat surfaces, not to mention uneven surfaces such as stairs and steps. It has adopted an entirely passive or low-power mechanism that does not have such ability to convey positive work.

  In order to assess the requirements of prosthetic legs, lower limb orthoses and exoskeleton devices, it is useful to understand the normal biomechanics associated with the subject's walking cycle. In particular, the function of a human ankle joint in sagittal plane rotation is described below for various motor function conditions.

  The mechanical properties of a conventional passive leg prosthesis (“AFP”), such as the Ossur Flex-Foot®, are essentially constant over the lifetime of the device. Patent Document 1 shows a significant advance over these conventional AFPs. In U.S. Pat. No. 6,057,027, the entire contents of which are incorporated herein by reference, the walking cycle is divided into five periods and the device mechanical properties are optimized independently for each of these five periods. It is recognized that performance can be improved.

  FIG. 1A shows an outline of various periods of the walking cycle of a subject on a flat ground. A gait cycle is typically defined as a period that begins with a heel contact on one foot and ends with the next heel contact on that same foot. The walking cycle is divided into two phases, that is, the stance phase (about 60% of the walking cycle) and the subsequent swing phase (about 40% of the walking cycle). The free leg period represents the part of the walking cycle when the foot is away from the ground. The stance phase begins with the heel touching when the heel touches the floor and ends with the toe off when the same foot rises from the ground. The stance phase is divided into toe-off when the same foot rises from the ground. The stance phase is divided into three sub-phases: Controlled Plantar Flexion (CP), Controlled Dorsal Flexion (CD), and Powered Plantar Flexion (PP).

  The CP begins with the heel contact illustrated at 102 and ends with the plantar contact illustrated at 106. CP describes the process by which the heel and forefoot first contact the ground. Researchers have demonstrated that CP ankle motion is consistent with a linear spring response where the joint torque is proportional to the joint displacement relative to the equilibrium position to the joint position. However, the motion of this spring is variable, and the joint stiffness is continuously modulated by the body at every step within the three sub-phases of the standing leg and the latter half of the swinging leg.

  After the CP period, the CD period continues until the ankle joint reaches the maximum dorsiflexion state as illustrated at 110 and starts the power plantar flexion PP. The relationship between the ankle torque and the position in the CD period is described as a non-linear spring whose rigidity increases as the ankle position increases. The ankle joint stores the elastic energy required to propel the body upward and forward during the PP period during the CD period.

  The PP period begins after the CD and ends at the toe take-off moment illustrated at 114. In PP, the ankle applies a torque according to the reflection response, which is the body pulls the body upward and forward. Next, the energy to be thrown out is released together with the spring energy stored in the CD period, and high buckling force is exerted in the latter half of the stance. This throwing action is necessary because the work done in the PP period is greater than the negative work absorbed in the CP and CD periods for medium to high walking speeds. The foot lifts off the ground during the swing leg period from the toe separation at 114 to the next heel contact at 118.

  Since the kinematic and dynamic patterns of the ankle joint on the stairs are different from walking on flat ground, another description of lower leg biomechanics is shown in FIGS. 1B and 1C. FIG. 1B shows the human ankle biomechanics when climbing stairs. The first phase of stair climbing is referred to as controlled dorsiflexion 1 (CD1), which begins at the dorsiflexed foot grounding shown at 130 and continues at 132 until the heel touches the step surface. to continue. In this period, the ankle joint can be modeled as a linear spring. Phase 2 is power plantar flexion 1 (PP1), which begins at the moment of plantar contact (when the ankle reaches maximum dorsiflexion at 132) and when dorsiflexion begins again at 134. End. The human ankle acts as a torque actuator and supplies more energy to support weight.

  The third period is controlled dorsiflexion 2 (CD2), where the ankle joint is dorsiflexed until 136 is released. In the CD2 phase, the ankle joint can be modeled as a linear spring. The fourth and final periods are power plantar flexion 2 (PP2), which begins at the detachment 136, where the foot kicks off the step, acts in parallel with the CD2 spring as a torque actuator, It continues until propelling upwards and forwards, and ends at 138 when the toe leaves the surface and begins the swing phase that ends at 140.

  FIG. 1C shows human lower leg biomechanics when the stairs are lowered. The stance phase of the staircase descent is divided into three sub-phases: control dorsiflexion 1 (CD1), control dorsiflexion 2 (CD2), and power plantar flexion (PP). CD1 begins with the foot grounding illustrated at 150 and ends with the foot grounding 152. In this period, the human ankle can be modeled as a variable attenuator. In CD2, the ankle joint continues to dorsiflex forward until it reaches the maximum dorsiflexion position shown at 154. Here, the ankle acts as a linear spring and stores energy throughout CD2. In the PP phase starting at 154, the ankle is bent at 156 until the foot lifts off the step. In this final PP phase, the ankle joints release stored CD2 energy and propel the body forward and upward. After toe lift at 156, the foot is controlled and positioned throughout the swing phase until the next foot contact at 158.

  In the stair lift shown in FIG. 1B, the human lower leg can be effectively modeled using a combination of actuators and variable stiffness mechanisms. However, the stair descent shown in FIG. 1 also needs to include a variable attenuator to model the lower leg complex, and the force absorbed by the human ankle joint is the force released when the stair is raised. Compared to stairs, it becomes considerably larger when the stairs are lowered. Therefore, it is appropriate to model the ankle joint as a combination of a variable attenuator and a spring in descending the stairs.

  Conventional passive prostheses, orthoses and exoskeleton devices do not adequately reproduce the biomechanics of the walking cycle. They are not biomimetic because they do not actively modulate impedance and do not apply a reflex torque response, even on flat ground, up or down stairs or slopes, or changing terrain conditions. Accordingly, there is a need for an improved prosthesis, lower limb orthosis and exoskeleton device, components thereof, and methods for controlling the prosthesis, lower limb orthosis and exoskeleton device, components thereof.

US Patent Application Publication No. 2006/0249315 US Pat. No. 7,313,463

  The invention described herein generally relates to prosthetic legs, lower limb orthoses and exoskeleton devices. Typical use cases for various embodiments of the invention include, for example, increased metabolism, permanent assistance of a subject with persistent limb lesions, or rehabilitation of a wearer with temporary limb lesions.

  One aspect of the present invention relates to an active orthosis or prosthetic device that includes a thigh member, a lower limb member, and a knee joint for connecting the thigh member to the lower limb member. The apparatus also includes a rotary motor having a motor shaft output, a motor drive transmission assembly coupled to the motor shaft output, and a drive transmission assembly coupled to the output of the motor drive transmission, the output of the drive transmission assembly Is coupled to the lower limb member and includes a drive transmission assembly that applies torque to the knee joint to rotate the lower limb member relative to the femoral member. The device is also at least one sensor having at least one output, wherein the position of the knee joint relative to the ankle joint can be determined from the at least one output while the wearer of the device is in a sitting position. A sensor and a controller that determines a time at which the knee joint is moved to a position in front of the ankle based on at least one output of the at least one sensor. In response to the determination, the controller controls the rotary motor to modulate the knee joint impedance, position, or torque to help lift the person from sitting to standing.

  Another aspect of the invention relates to a method for controlling a knee orthosis or prosthesis having at least one actuator. The method includes detecting the position of the person's knee relative to the person's ankle while the person is sitting and moving the knee to a position in front of the ankle based on the result of the determining step. Determining a time to be generated and generating an output representative of the time. Depending on the output, at least one actuator of the knee brace or prosthesis is activated to help lift the person from sitting to standing.

1 is a schematic diagram illustrating various periods of a wearer's walking cycle on flat ground. Fig. 6 is a schematic diagram illustrating various periods of a wearer's walking cycle when going up stairs. FIG. 6 is a schematic diagram illustrating various periods of a wearer's walking cycle when going down stairs. 1 is a schematic diagram illustrating a method for determining ankle, heel, and toe trajectories of a prosthetic limb, orthosis, or exoskeleton device, according to an illustrative embodiment of the invention. It is a graph of the experimental data which shows the acceleration of the ankle joint during a walk. 1 is a schematic diagram illustrating a method for measuring foot tilt (heel height), according to an illustrative embodiment of the invention. 6 is a schematic diagram illustrating a method for determining heel and toe coordinates for an ankle joint in a foot reference frame, according to an illustrative embodiment of the invention. 4 is a schematic diagram illustrating a method for estimating a wrinkle vector, according to an illustrative embodiment of the invention. It is a figure which shows the ankle joint axis locus | trajectory calculated | required by calculation with the inertial measurement apparatus in various walking movement situations. It is a figure which shows the two-dimensional geometric shape which describes the flight locus | trajectory of the ankle joint of a prosthetic limb apparatus. FIG. 6 illustrates a method of making a stair-slope discriminator using an ankle angle of attack as a trajectory feature that discriminates staircase and slope walking motion situations according to an exemplary embodiment of the present invention. FIG. 6 illustrates a method for positioning an ankle joint prior to foot contact, according to an illustrative embodiment of the invention. FIG. 7B illustrates the method of FIG. 7A that can be used to sense the presence of a staircase and the presence of a foot protrusion at a staircase landing, according to an illustrative embodiment of the invention. FIG. 6 illustrates a method for positioning an ankle joint in a slope walking motion situation, according to an illustrative embodiment of the invention. FIG. 7B illustrates how the method of FIG. 7B is adapted to use optimized impedance, according to an illustrative embodiment of the invention. It is a figure which shows the method for determining an inertia reference spring equilibrium state based on the angle of the topography in plantar grounding. It is a figure which illustrates the influence of the walking speed with respect to an ankle joint torque and the angle of an ankle joint, and shows how push pull actuator control is performed to the parallel elastic element selected appropriately. FIG. 6 illustrates a method for controlling a lower limb device, according to an illustrative embodiment of the invention. 1 is a schematic diagram illustrating a model-based controller for performing impedance and torque control in a prosthetic leg device, according to an illustrative embodiment of the invention. 1 is a schematic diagram illustrating a model-based controller for performing torque control in a prosthetic leg device, according to an illustrative embodiment of the invention. 10B is a schematic diagram illustrating the relationship of mechanical impedance that determines the impedance control performed in FIG. 10A. 11B is a schematic diagram illustrating the impedance and reflection relationship that determines the impedance and reflection control performed in FIG. 10B. FIG. 6 is a schematic diagram showing how zero moment pivot reference ground reaction force is used to determine the restoring torque required to stabilize the inverted pendulum mechanical behavior of a person wearing a prosthetic device. 1 is a schematic illustration of a leg foot member, ankle joint, and foot member of an artificial ankle joint showing ground reaction forces and zero moment pivots. FIG. 2 is a schematic diagram showing components of an artificial ankle joint showing the relationship between forces and moments between components required to determine ground reaction force and zero moment pivot. FIG. 2 is a schematic diagram showing components of an artificial ankle joint showing the relationship between forces and moments between components required to determine ground reaction force and zero moment pivot. FIG. 2 is a schematic diagram showing components of an artificial ankle joint showing the relationship between forces and moments between components required to determine ground reaction force and zero moment pivot. It is a figure which shows the biomimetic ((GAMMA)-(theta)) behavior of the artificial ankle joint in a flat ground as a function of the walking speed at the time of a power plantar bending. It is a figure which shows the biomimetic ((GAMMA)-(theta)) behavior of the artificial ankle joint in a flat ground as a function of the walking speed at the time of a power plantar bending. It is a figure which shows the influence of the transition of the leg with respect to the contact length. It is a figure which shows the influence of the transition of the leg with respect to the contact length. It is a figure which shows how the normalization contact length can be used as a means to perform a biomimetic behavior at the time of power planting in the table of contact length reduction depending on speed. FIG. 6 is a diagram showing how the estimated y component of a zero moment pivot vector changes during a typical walking motion. FIG. 6 illustrates a method for incorporating an attenuation factor into device performance, according to an illustrative embodiment of the invention. 2 is a schematic diagram illustrating a representation of a control system scheme in the case of heel grounding, according to an illustrative embodiment of the invention. FIG. 6 is a schematic diagram illustrating a representation of a control system scheme for toe grounding, according to an illustrative embodiment of the invention. FIG. 18A illustrates a method for performing position control applied to an artificial ankle joint (eg, the device 1700 of FIG. 17A), according to an illustrative embodiment of the invention. FIG. 6 illustrates a method for using step-by-step terrain adaptation, according to an illustrative embodiment of the invention. It is a figure which shows the example impedance which an artificial ankle joint applies with respect to three different walking movement situations. 1 is a schematic diagram illustrating a lower limb biomechanical system, according to an illustrative embodiment of the invention. FIG. 6 illustrates a method for performing posture reconstruction for a torso posture, a thigh posture, and a torso / body mass center posture, according to an illustrative embodiment of the invention. 1 shows a prosthetic device, according to an illustrative embodiment of the invention. FIG. FIG. 17B illustrates a portion of the lower limb device of FIG. 17A showing a passive parallel elastic element. FIG. 18B shows a passive parallel elastic element of the apparatus of FIG. 17B. 18 is a free body diagram for the passive parallel elastic element of FIG. 17C, according to an illustrative embodiment of the invention. FIG. FIG. 17B is a perspective view showing structural elements (pyramids) of the apparatus of FIG. 17A, according to an illustrative embodiment of the invention. FIG. 17B is a cross-sectional view illustrating an alternative method for measuring the axial force and moment applied to the lower limb member of FIG. 17A, according to an illustrative embodiment of the invention. FIG. 6 illustrates a method for calculating in-plane moment vectors and axial forces using circularly arranged displacement sensors on a printed circuit assembly, according to an illustrative embodiment of the invention. 18 is a schematic diagram illustrating a state and actuator controller for use with the apparatus of FIGS. 17-17G, according to an illustrative embodiment of the invention. 2 is a schematic wiring diagram illustrating an electrical equivalent circuit of a prosthetic leg device, according to an illustrative embodiment of the invention. FIG. 18B is a schematic wiring diagram of the electrical circuit of FIG. 17I including sensor measurement results used in controlling the device. FIG. 3 illustrates a passive series elastic member, according to an illustrative embodiment of the invention. FIG. 3 illustrates a passive series elastic member, according to an illustrative embodiment of the invention. FIG. 3 illustrates a passive series elastic member, according to an illustrative embodiment of the invention. FIG. 3 illustrates a passive series elastic member, according to an illustrative embodiment of the invention. FIG. 3 shows a prosthetic device incorporating a series leaf spring, according to an illustrative embodiment of the invention. FIG. 6 shows a prosthetic device using an alternative series spring, according to an illustrative embodiment of the invention. FIG. 6 shows a prosthetic device using an alternative series spring, according to an illustrative embodiment of the invention. 1 is a perspective view showing a linear actuator that can be used in various prosthetic legs, lower limb orthoses and exoskeleton devices, according to an illustrative embodiment of the invention. FIG. It is sectional drawing which shows the linear actuator of FIG. 20A. 1 is a perspective view showing a linear actuator that can be used in various prosthetic legs, lower limb orthoses and exoskeleton devices, according to an illustrative embodiment of the invention. FIG. 1 is a schematic top view of a lower limb brace or exoskeleton device (wearable robot knee brace), according to an illustrative embodiment of the invention. FIG. FIG. 22B is a side view of the device of FIG. 22A. FIG. 22D is a schematic diagram illustrating the inner portion of the knee joint drive assembly of the apparatus of FIGS. 22A and 22B. 6 is a schematic diagram illustrating the problem of human balance on an inclined surface. Fig. 6 is a schematic diagram showing an acceptable solution to the problem of balance based on variable knee flexion by the wearer. FIG. 6 is a schematic diagram showing how to use a unique sensing function to balance the representation of the human body and the wearer on the ground. FIG. 6 is a schematic diagram illustrating a method for balancing a wearer when the wearer stands up from a chair, according to an illustrative embodiment of the invention. FIG. 6 is a schematic diagram illustrating a method for balancing a wearer when the wearer stands up from a chair, according to an illustrative embodiment of the invention. FIG. 6 is a schematic diagram illustrating a method for balancing a wearer when the wearer stands up from a chair, according to an illustrative embodiment of the invention. It is a figure which shows the definition of the work of movement. It is a figure which shows the definition of a hip joint impact force. FIG. 26 shows the biomechanical characteristics of normal human walking during a walking movement. FIG. 27 shows the biomechanical mechanism by which quadriceps decline affects the walking motion on the flat ground. FIG. 28 shows how the knee device is used to restore normal walking motion. FIG. 29A shows a rising sequence of a healthy person. FIG. 29B shows a rising sequence of healthy persons. FIG. 29C shows a rising sequence of healthy persons. FIG. 29D shows a rising sequence of healthy persons. FIG. 30A illustrates a problem in implementing the same rising sequence of patients with disabilities. FIG. 30B illustrates a problem in implementing the same sequence of rising up patients with disabilities. FIG. 30C illustrates a problem in implementing the same rising sequence of patients with disabilities. FIG. 30D illustrates a problem in implementing the same sequence of rises for a disabled patient. FIG. 31A shows how the knee device is used to assist in the sequence of standing up a disabled patient. FIG. 31B shows how the knee device is used to assist in the sequence of standing up a disabled patient. FIG. 31C shows how the knee device is used to assist in the sequence of standing up a disabled patient. FIG. 31D shows how the knee device is used to assist in the sequence of standing up a disabled patient.

Detailed Description of Preferred Embodiments
Inventors use the lower limbs of humans to perform and adapt to many different activities in addition to normal walking, such as going up and down stairs and walking on slopes in daily flow I recognize that. The lower leg component requires the greatest force and must exhibit the most adaptable motion to the terrain because it is in direct contact with the underlying terrain. The inventors have further shown that the performance of AFP can be dramatically improved by dynamically optimizing the mechanical properties of the device in different ways and dynamically controlling the device in different ways for each of their activities. Recognized.

  For example, when a person is walking on a flat ground, the angle of the foot may be controlled so that the heel is lower than the toes when the foot is lowered and brought into contact with the ground. However, when a person is climbing the stairs, it is better to control the angle of the foot so that the toes are lower than the heel when the foot is lowered and brought into contact with the next step.

  In this application, various embodiments of AFP are described that operate appropriately in each of these different situations by detecting the terrain being traversed and automatically adapting to the detected terrain. In some embodiments, the ability to control the AFP built for each of these situations includes (1) the ability to determine the activity to be performed, and (2) the characteristics of the AFP based on the activity being performed. Functions to dynamically control, (3) functions to dynamically drive the AFP based on the activity being performed, (4) terrain texture, irregularities (eg, terrain stickiness, terrain slipperiness, terrain) The ability to respond to these with appropriate traction control, and (5) to respond to dynamic control and dynamic drive. Based on the five basic functions of the mechanical design of possible AFP.

  The inventors have found that an exemplary method for finding the activity being performed is a spot trajectory (typically ankle rotation) on the lower limb (or shin) between the ankle and knee joints. It was decided to track the virtual center). FIG. 6A shows shin loci corresponding to five different activities, with the additional slope trajectory distinguishing steep and gentle slopes. The system can use this information to find out what activity is being performed by mapping the tracked track to a set of activities.

  By examining the trajectory of the lower limb (shin part), it is possible to distinguish flat terrain, ascending stairs, descending stairs, ascending slope, or descending slope. For example, if the system recognizes a trajectory, it switches to the appropriate mode and dynamically controls (drives) the AFP as already established for that mode. If the trajectory does not fall within the classification range, the AFP controller optimizes the response to minimize the objective function in the sense of probability control, or fuzzy logic based on the probability that the terrain falls within the classification range. Or apply ad hoc control.

  One preferred method of tracking the shin locus is by attaching an inertial measurement unit (IMU) to the anterior surface above the lower limb member (shin) and processing the signal output by the IMU. . A preferred way to distinguish the various trajectories is to monitor the velocity of the ankle angle of attack. These topics are described in more detail below.

  In addition to dynamically optimizing the mechanical properties and dynamically controlling the device in different ways for each of the different activities, the inventors have determined that the performance of the device depends on various parameters. And recognize that further refinement of the control can be made.

  For example, when a person is walking slowly (eg, at a speed of less than 0.9 meters per second), performance can be improved by increasing the ankle impedance relative to the impedance used for normal walking. Alternatively, if a person is walking fast (eg, at a speed of 1.7 meters per second), performance can be improved by lowering the ankle impedance relative to the impedance used for normal walking.

  In addition, if the controller determines that the ankle does not respond as expected when crossing normal terrain, the controller may have features, textures, or irregularities in the terrain (eg, terrain stickiness, Terrain slipperiness, terrain roughness or smoothness, whether the terrain has obstacles such as rocks, etc.) (and modify the controller output).

  Each of the five functions identified above (i.e., the ability to find out the activity being performed, the feature of the terrain, the texture, or the presence of irregularities, the ability to dynamically control the characteristics of the AFP, AFP Will be described in more detail below.

Determination of Activity Performed Inertial Posture and Trajectory Estimation FIG. 2 illustrates a prosthetic, orthotic, or exoskeleton device (eg, device 1700 of FIG. 17A) based on the inertial posture of a lower limb member 220 coupled to an ankle joint 200. 6 is a schematic diagram illustrating a method for determining the trajectory of the ankle joint 200, the heel 212, and the toe 216 and the angle formed by the lower limb member 220 and the foot member 208; The posture is the position and orientation of the coordinate system. Device 1700 includes inertial measurement device 204 coupled to lower limb member 220. The inertial measurement device 204 includes a triaxial rate gyro for measuring angular velocity and a triaxial accelerometer for measuring acceleration. By placing the inertial measurement device on the lower limb member 220, the angular velocity and acceleration are measured together for all three axes of the lower limb member 220. The inertial measurement device 204 estimates the degree of freedom 6 of the posture of the lower limb member 220, the inertia (based on the world coordinate system) orientation, and the placement of the ankle joint 200 (ankle joint—the center of foot rotation).

  In some embodiments, the posture of the lower limb member 220 is used to calculate the instantaneous placement of the knee joint. Knowing the angle (θ) of the knee joint 200, the instantaneous posture of the bottom of the foot 208, including the placement of the heel 212 and the toe 216, can be calculated. This information can then be used to measure the terrain angle in a plane defined by the ankle / foot member rotation axis when the foot member 208 is flat. Mounting the inertial measurement device 204 on the lower limb member 220 has advantages over other potential placements. Unlike mounting on the foot member 208, mounting the lower limb member 220 prevents physical misuse and avoids flooding. In addition, tethering with a cable that would otherwise be required when attached to the foot member 208 is not required, ensuring mechanical and electrical integrity. Finally, the lower limb member 220 is centrally located in the kinematic chain of the hybrid system (see FIG. 15), thereby calculating the femoral and torso posture with minimal additional sensors. Becomes easier.

  The inertial measurement device 204 is an orientation of the artificial leg device in the world coordinate system with respect to the ground.

,position

, And speed

Used to calculate By a quaternion, or a 3 × 3 matrix of unit vectors that define the x, y, and z axis orientations of the ankle with respect to the world coordinate system

Can be expressed. The ankle joint 200 is defined such that its orientation is centered on the axis of rotation of the ankle joint associated with the lower limb member 220. From this center point, position, velocity, and acceleration can be calculated. For example, for a point of interest on a foot (eg, heel 212 or toe 216), a foot member-foot joint orientation transformation to derive a position

Is used and the relational expression

Use
However, in the formula

And
Where γ is the inertial leg member angle,

And
In the equation, θ is the angle of the ankle joint.

  In this embodiment, the inertial measurement device 204, including a 3-axis accelerometer and a 3-axis rate gyro, is placed on the anterior surface above the lower limb member 220 (eg, as shown in FIG. 17A). There is a need to eliminate the effects of scale, drift, and cross-coupling on global coordinate system orientation, velocity, and position estimates introduced by numerical integration of accelerometer and rate gyro signals.

Zero Speed Update Inertial navigation systems typically use zero speed update (ZVUP) by averaging regularly over a period of time, usually seconds to minutes. Such an arrangement of the inertial measuring device is hardly stationary in the prosthetic device. However, the sole of the foot is the only place where it is stationary and then the only controlled dorsiflexion state during the walking cycle. An exemplary zero rate update method that is not affected by this limitation for use with various embodiments of the present invention is further described below.

To solve this problem, ankle orientation, velocity and position integration is performed. After binarizing the acceleration ^IMU a of the inertial measurement device, the ankle joint acceleration

Is the equation of motion of the rigid body

Where, where

and

Are vectors of angular velocity and angular acceleration in the coordinate system of the inertial measurement device, respectively, and X represents an outer product.

  Relational expression

Uses standard strapdown inertial instrument integration methods and is known to those skilled in the art.

Is solved in the same manner as in the case of Equations 1 to 3. In equations 5-14 above, the matrix

Is the orientation matrix

Used interchangeably.

  Then, the speed and position of the ankle joint based on the world coordinate system

Derived at a time after the previous zero speed update (i th zero speed update), where

Is reset to zero for all i.

Through the experiment, by using a log of inertial measurement device data obtained from an exemplary prosthetic device (eg, prosthetic device 1700 of FIG. 17A), the inertial measurement device reference acceleration is approximately z-accelerated in a controlled dorsiflexion state. It equals 1 g (about 9.8m / ^{s 2),} the dispersion is a predetermined value of z acceleration (<0.005 g ²⁾ about 50.75 seconds is smaller than and 50.9 seconds and early - leg member Shows a period of time 220 is rotating with respect to a stationary ankle 200--determined to be sufficiently calm (see FIG. 2B). In other embodiments of this technique, a suitable quiet period can be detected in the part of the foot. By combining the known acceleration, angular velocity, and angular acceleration of the ankle joint with the known angle of the sensed foot joint (the angle formed by the foot member and the leg member), angular velocity, and angular acceleration. , Acceleration at any point on the foot can be calculated. A point at the bottom of the foot can often be used to perform a zero speed update in successive walking cycles. After this speed is known, the ankle speed can be calculated empirically. This speed (not zero) can be used as a reference in performing a zero speed update.

In a prosthetic device, the quiet period is almost always present in a controlled dorsiflexion state, so a zero speed update can be performed each time the wearer steps out. In each zero velocity update, preferably the three terms that each contribute to the velocity error to be evaluated are the z-axis tip δθ _{x of} the world coordinate system centered on the x axis (the feet at the zero velocity update in the previous step). Vector aligned with the rotation axis of the internode), z-axis tilt δθ _{y of the} world coordinate system around the _y- axis (vertical axis of the world coordinate system (opposite of the gravity vector) and x-axis of the world coordinate system) ), And inertial instrument scaling δg along the vertical axis. The values of these terms are used to correct the calculated posture, inertial orientation, and the previous calculated posture and inertial orientation of different components of the device (eg, leg member 1712 of FIG. 17A).
While performing orientation, velocity, and position integration, calculate a sensitivity matrix M (t) related to the velocity error that would be introduced by the error vector α = [δθ _x δθ _y δg] ^T. M (t) is a relational expression

Where M (t) is numerically integrated and the overall final velocity sensitivity M *

Is generated. In some embodiments, the error vectors are expanded to include accelerometer bias offsets when these errors are significant, thereby increasing the number of columns of M (t) and M *. In this case, M ^{* -1} takes the form of a Penrose parliamentary inverse matrix or K * with an optimal confidence gain. K ^* can be calculated using standard optimal linear filtering methods. Those skilled in the art can include or use other terms without loss of generality.

  For the zero speed update for step i, the estimated non-zero ankle joint speed

The value of α that would have generated

Where α is an innovations correction vector. Not all innovation corrections (α) are applied because some speeds are not zero due to noise in accelerometer and angular velocity measurement results. Instead, the correction is scaled by a filtering constant (fraction) k depending on the magnitude of the noise. At this point, the new orientation matrix

And the magnitude of gravity (g) is

g (ZVUP ⁺ ) = g (ZVUP ⁻ ) −kα (3) Equation 21
Where O _x (tip) and O _y (tilt) represent the tip and tilt incremental rotations about the x and y axes, respectively,

and

Represents the time before and after ZVUP.

  The zero velocity update can be extended to estimate accelerometer and rate gyro bias offsets using linear estimators. It is also possible to use a matching angular alignment error (eg, centered on a given axis) to estimate the rate gyro's bias about that axis. In one embodiment, this is done by creating linear stochastic models of accelerometer and rate gyro bias and using the zero velocity update prediction residual as an input to a linear filter applied to these models. Is done.

  The method described above is one method for continuously updating the orientation and the apparent magnitude of gravity. In this embodiment, an initialization procedure is used when the prosthetic device is first powered on. In this manner, the wearer moves forward one step and stops one step back to its original position when a request is made from the device (eg, by a vibration code transmitted by the device or alternative user interface). In this process, we walk with the affected leg (in a bilimb amputee, this calibration is performed in a serial fashion as selected by the amputee). In this calibration, two ZVUPs are invoked, one initializes the orientation and gravity magnitude, and the other checks the result. This ensures the integrity of the inertial measurement device signals, processing, and controller communication.

  The above process performs initialization of inertial orientation. However, perform a full calibration of the IMU to take into account the error offset vector (ε), which is the vector containing the bias offset, scale, and cross-sensitivity, and the gyro signal embodied in the accelerometer. That is a general concern. In manufacturing, a robot or 6-degree-of-freedom machine can carry the IMU and apply the reference trajectory continuously as a means of measuring the effects of these error sources. The sensitivity matrix (M (ε)) of the sensed reference trajectory to each of the error sources can be easily calculated by those skilled in the art. By measuring the perceived deviation from a set of multiple reference trajectories—typically, the deviation of the end of each trajectory segment—the vector (ε) is used using regression or other linear estimation methods. However, it is assumed that this set of reference trajectories is sufficiently large to stimulate the influence of the respective error sources. The inventors have found that a reference trajectory including a cycle such as a polygon and a circle in three orthogonal planes is sufficient to calibrate the complete vector of error sources. Such a reference trajectory, for example, walks in a sequence of closed patterns on the horizontal plane and rotates in turn around the vertical axis, thereby resolving important vector elements (accelerometer bias, scale, and cross sensitivity). It can also be guided by the wearer to recalibrate.

  In some embodiments of the present invention, these principles of the method include, for example, a wearer's femoral member in which a prosthesis, brace, or exoskeleton device treats these portions of the wearer's body and reinforces its performance. The same applies to the step of correcting or minimizing the effects of accelerometer and rate gyro drift errors associated with accelerometers and rate gyros located on the fuselage. In one embodiment, the method includes an accelerometer signal output by an accelerometer and a rate gyro coupled to a prosthetic or orthotic thigh member when the ankle joint is substantially stationary during the prosthetic or orthotic walking cycle; Determining an offset value for the rate gyro signal. The method can also include measuring an angle formed by the lower limb member with respect to the femoral member. In other embodiments, the method includes an accelerometer signal output by an accelerometer and a rate gyro coupled to the wearer's torso when the ankle joint is substantially stationary during the prosthetic or orthotic walking cycle, and Also included is determining an offset value for the rate gyro signal. The method can also include the step of measuring the angle the thigh member makes with the wearer's torso. Accordingly, these methods are extended to the wearer's femoral member and / or torso by performing these measurements and utilizing them in linkage constraint equations and related methods, as shown in FIG. be able to. During zero vector velocity updates, linkage constraints allow for back propagation of the joint velocity reference from the ground-based zero velocity of the lowest link in the kinematic chain (eg, the linkage defining the human-robot hybrid system). These velocity criteria can be used as input to attitude realignment and gravity correction as defined above.

Discrimination of exemplary ankle trajectory and terrain situation After the inertial instrument offset is calculated and corrected (zeroed), the foot inclination (β) (also known as the height of the heel) is shown in FIG. Determined as illustrated. From this illustration, considering that β = − (θ + γ), it is easy to check when the wearer is standing at the sole ground. By taking an average over a period of about 1/10 second, an accurate estimate of β can be determined. Since then, the coordinate system of the foot and ankle joint

The orientation component of the transformation that defines

Calculated based on As mentioned above, the translational motion component of this transformation remains zero.

  After the foot inclination is defined, it is then necessary to determine the coordinates of the heel 212 and the toe 216 in the foot coordinate system. One exemplary method for determining this is:

and

Is defined as the vector coordinates of the heel and toes in the new foot coordinate system. Since the rotation contribution of β has already been incorporated, the z components of these vectors are the same. It can be assumed that the x components of these vectors are both zero. So these vectors are

Where z ₀ defines the z coordinate of the bottom of the foot (shoe).

  FIG. 4 is a schematic diagram illustrating a method for determining the coordinates of the heel 212 and toe 216 with respect to the ankle joint 200 in the foot reference frame, according to an illustrative embodiment of the invention. In the first step of the foot calibration method defined in FIG. 4, the y-coordinate of the ankle joint 200 is aligned with the ground reference (eg, sidewalk seam, rug salient features, or linoleum floor). Here, this ground reference is arbitrarily defined as the origin of the world coordinate system. In mathematical notation, this alignment is

In the formula,

Is the starting position for the movements performed in steps 2 and 3. In the second step, the toe 216 is placed on the ground reference. In mathematical notation, this alignment is

Or

Takes the form

Similar relationships are also determined in the alignment in step 3. If the above equation is solved independently, two different estimates of z ₀ are obtained. By combining these two constraint equations into one, a least square estimate of Y _heel , Y _toe , and z ₀ can be determined.

  The heel 212 and toe 216 calibration method described above involves a series of steps used when first putting on a pair of feet / shoes. Such calibration can be performed, for example, at the office of a prosthetic orthosis.

  In another exemplary method, the heel and toe vectors are calculated during execution. As shown in FIG. 5, the ankle joint 200 draws an arc 500 in the first half of the stance between the foot contact and the plantar contact. The radius and orientation (midpoint angle) of the arc 500 completely determines the heel and toe vectors. Mathematically, this is a series of ankle positions recorded in the first half of the stance

Is described as Measurement of two ankle positions corresponding to the angle (θ _i ) position of two statistically distinct lower limb members 220 (γ _i ) and ankle joint 200 is required,

It becomes. Then, by taking the difference of the equations, the vector solution is

It becomes. This solution requires that (O (γ ₂ ) O (θ ₂ ) −O (γ ₁ ) O (θ ₁ )) is reversible. Furthermore, from the point of view of optimal linear filtering, this “gain matrix” must be large enough to obtain statistically significant results.

  In view of the fact that the prosthetic device experiences significant vibration during the first half of the stance, the above equation can be extended to N sets of ankle position / angle measurement results. The resulting N-1 equations can be solved using a least squares method with an optimal vector estimation. The above equation is similarly configured to solve for the toe vector when the first half of stance starts with toe grounding.

  FIG. 6A shows a flat ground (620), a 5 ° ascending slope (624), a 5 ° descending slope (628), a 10 ° ascending slope (632), a 10 ° descending slope (636), and an ascending staircase (640). And the ankle axis trajectory obtained by calculation by the inertial measurement device in various walking motion situations for the wearer in various terrain such as the descending stairs (644). The situation is the shape of the terrain and how the wearer interacts with the terrain.

  FIG. 6B shows a two-dimensional geometric shape describing the flight trajectory of the ankle joint of the prosthetic device. Here, when the flat ground walking is treated as a subset of the slope up / down walking motion situation (the flat ground is a slope of 0 degree), the situation discrimination shifts to the discrimination of the stairs up / down from the slope up / down. This discrimination typically requires that the bottom flexion (not dorsiflexion) of the ankle 600 should optimize the ground contact dynamics in a staircase situation, but typically in an inclined walking motion, the ankle joint. 600 is important because it is dorsiflexed (or held neutral) to optimize foot grounding dynamics. In the latter situation, it is only on extremely steep downhill slopes where the buckled ankle is properly oriented.

FIG. 6C shows a method of making a stair-slope discriminator using the ankle angle of attack (Ψ) as a trajectory feature that discriminates the staircase and slope walking motion status in the recorded data set. FIG. 6C is a graph of the estimated velocity vector angle of attack of the ankle 600 of the device over the entire walking cycle for each step taken by the wearer. In this data, a limb amputated patient wearing the prosthetic limb device 1700 of FIG. 17A advanced 31 steps as follows (meaning a walking cycle based on the right foot).
1. 1st to 6th steps: Go up the slope of 5 ° in 6 steps. Step 7: Get on the landing in one step 3. 8th to 9th steps: Go down the 10 ° slope in 3 steps. 4. Recording gap 10th to 11th steps: Go up the stairs in 2 steps. 12th step: Get on the landing in one step Steps 14-17: Go down the 5 ° slope in 4 steps. 18th to 19th steps: Advance two steps on the flat ground. 20th to 21st steps: climb a 10 ° slope in two steps Step 22: Go to the landing from a 10 ° slope in one step 23rd-24th steps: Go down the stairs in 2 steps 12. 25th to 31st steps: 7 steps on flat ground The walk at the time of this recording included both rising and falling walks on slopes and stairs. FIG. 6C shows that by monitoring the ankle velocity attack angle (Ψ), the staircase and the slope can be distinguished while the ankle is in flight before the ankle contact. The foot 604 always lands on the stairs when Ψ drops below a small positive value in this record (and other similar records). In all other cases, the foot will land on the slope regardless of the ramp angle (0 °, −5 °, + 5 °, −10 °, + 10 °). Therefore, Ψ is a preferred walking task status discriminant that can be used by the processor to determine what activity is taking place.

  Alternative methods for staircase-slope discrimination can also be used in other embodiments of the invention. Posture (orientation in inertial space) lower limb member 608 (shin) and ankle velocity attack angle (Ψ) can be used in one embodiment of the present invention to distinguish stairs or slopes / flats. The trajectory of the ankle 600 in the yz plane (see FIG. 6A) can be used in alternative embodiments of the present invention for staircase-slope discrimination.

Swing leg ankle positioning Stair slope discriminator

Real-time prediction. If the discriminator detects a step, including flat ground,

It becomes. Otherwise, the slope is

Is assumed. This inclination corresponds to the minimum possible value when the foot is not in contact with the ground.

This is the minimum of the two possible slopes-the angle that the heel is currently making with respect to the toe position from the last step and the angle that the toe makes with respect to the toe position from the last step. .

  value

Once found, a variety of different methods for positioning the ankle joint to adapt to this predicted terrain tilt can be applied. Two examples of such methods are described below. In one embodiment of the present invention, the method of the discriminator described above includes the impedance, position, or torque of the joint of a prosthetic leg, lower limb orthosis, or exoskeleton device (eg, device 1700 of FIG. 17A) worn by the wearer. Are used to control at least one of This method involves estimating the velocity vector attack angle (eg, the y-axis value of the data of FIG. 6C) of the ankle joint of the device throughout the second half of the swing leg. In one embodiment, the method matches the position of the foot member of the device to the toe landing position when the velocity vector angle of attack has a predetermined sign (eg, a negative value for the data of FIG. 6C). Step to adjust. In another embodiment of the present invention, the method includes setting the position of the foot member of the device to the heel landing position when the velocity vector angle of attack has a sign opposite to a predetermined sign (eg, a positive sign). With adjusting steps.

  In some embodiments, the method is applied to the lower limb member during a period from when the foot member's heel touches the underlying terrain to when the foot member is placed in the planted ground position relative to the underlying terrain. Adjusting the impedance (eg, ankle joint impedance) of the device to minimize the cost function based on the expected force being generated.

  FIG. 7A illustrates a method for positioning the ankle joint 700 prior to foot contact. In this method, the angle of the ankle joint is optimized to minimize the cost function based on the expected force (f (t)) applied to the ankle 700 from the grounding of the foot member 708 to the sole grounding. Is done. Both the heel first 716 and toe first 712 strategies are evaluated and the strategy that includes the optimal ankle 700 angle that minimizes the cost function is selected. FIG. 7A illustrates the method used.

  In other embodiments, the method of FIG. 7A is enhanced to sense the presence of stairs and constrain angle of attack optimization to toe grounding only in the case of stairs with short landing areas. When going up or down a steep, narrow set of stairs, the prosthetic device is programmed to track the volume that is swept by the foot when climbing-the volume that was not in contact between the foot and the stairs. The In the latter half of the swing leg, for example, when it is determined that there is no landing area for the heel, the optimization is restricted to a toe landing solution. In this embodiment, the z rotation is a rotation about the longitudinal axis of the lower limb member 704 of the device (eg, the z axis of FIG. 17A). When going down the stairs and thus rotating the foot member 708, the landing area is limited and the foot member 708 must be rotated to land on the stairs from the front. In this case, the tip of the toe 712 lands on the landing, so that the minimum force solution that can only be used for the method of FIG. 7A is obtained. Such z rotation sends a signal to the system that the landing area is limited, making the toe landing the safest alternative when compared to the heel landing.

The complex impedance calculation used in the above method is applied to the adaptive ankle positioning method as a means of minimizing the impact on the foot or the use of excessive braking force when the ankle 700 rotates to the foot landing state. Can be applied. FIG. 7D shows how the method of FIG. 7A is adapted to use optimized impedance. Optimal control to zero the ankle joint linearity and angular momentum without impacting the foot after finding the optimal angle of attack (ψ ^* )

Find out. The corresponding ankle angle response (θ * (t)) is then used as the equilibrium trajectory. In order to cope with the uncertainties of the momentum and the local terrain angle, it is possible to obtain the corresponding optimum impedance regarding this optimum locus.

  A simpler method can also be used, as shown in FIG. 7C. FIG. 7C shows a method for positioning the ankle joint in a slope walking motion situation. In this method, when the lower limb member 704 is vertical, the angle of the ankle joint 700 is the topography (tilt degree)

) The angle is such that it comes to the ground contact position above. Relational expression

It is also beneficial to generalize this method to adjust the angle of the ankle joint to be directly proportional to the degree of inclination predicted by. By using this relational expression, the angle of the ankle joint can be adjusted according to the wearer's preference.

  In either of the two methods described above, the ankle joint angle 700 before foot contact is continuously controlled to match the desired ankle joint 700 angle until the foot hits the ground ( Steered).

Control of stance phase impedance and torque The next step involves restoring the orientation of the lower limbs (shin) to align with the local longitudinal direction in the stance phase. FIG. 8 illustrates a method for determining an inertial reference spring equilibrium based on the angle of the terrain at the plantar contact of the prosthesis 800, eg, the prosthetic device 1700 of FIG. 17A. The prosthetic limb 800 has a foot member 808 having a toe 816 and a heel 820. The prosthesis also has an ankle joint 804 and a lower limb member (shin) 812. The terrain angle (φ) is an input to the control system. The control system shifts the curve (Γ-Θ) of FIG. 10A based on the change in the terrain angle (φ) (this changes the ankle impedance K _controlled plantant flexion) and causes the wearer's overall bend during control plantar flexion. Maintain or improve balance (as described or illustrated in FIG. 10F). The control system sets the impedance of the prosthetic ankle 804 so that the equilibrium angle of the ankle is equal to the terrain angle (φ), and the control system sets the lower limb member 812 (shin) to align with the local longitudinal direction 850. Restore the orientation of.

FIG. 9 shows the influence of walking speed on the relationship between ankle torque and ankle angle during controlled dorsiflexion. The control system sends the command to the ankle joint 804 to move the lower limb member (shin) 812 toward the equilibrium point, thereby changing the curve (Γ−Θ) of FIG. 10A based on the change in the terrain angle (φ). Shift (and thereby change the impedance K _{controlldodorflexion of the} ankle 804) to maintain or improve the overall balance of the wearer during controlled low flexion.

FIG. 10A illustrates a method for controlling a lower limb device, according to an illustrative embodiment of the invention. As shown in FIG. 10A, this is accomplished in the following manner in the control system.
1) Impedance in the latter half of the free leg (step 1000) (dynamic spring) to soften the impact during the time interval between foot ground contact and plantar ground contact as described herein with respect to FIG. 7A Constant and ankle angle balance angle) (the controller shifts the curve (Γ−Θ) based on the negative kinetic energy impact during power plantar flexion and the minimization of the hip impact force) Impedance _{Kpoweredplantarflexion} is changed)).
2) Applying lifting force to the rear thigh-achieved by causing an ankle (and knee) reflex response at or before the front thigh impact (step 1004).
3) Maintain the inertial reference equilibrium angle during the controlled dorsiflexion and balance (keep equilibrium) on the sloped terrain (as described and illustrated in FIG. 10F) (step 1008).

  FIG. 10B is a schematic diagram illustrating a controller for performing impedance and torque control in a prosthetic device (eg, device 1700 of FIGS. 17A-17E), according to an illustrative embodiment of the invention. FIG. 10E is a schematic diagram illustrating the relationship between impedance and reflection that determines the impedance and reflection control performed in FIG. 10B.

As shown, the impedance spring, damping, and inertial components are defined with respect to the trajectory θ ₀ (t). Both the impedance gain matrix and trajectory illustrated in FIG. 10B are loaded in real time, adaptively from the state controller processor, depending on the period, terrain condition, terrain texture, and walking speed in the walking cycle as described above. Is done.

  Studies have shown that the intact limb exhibits a reflex response resulting from nonlinear positive torque (force) and nonlinear positive joint velocity feedback. Both types of feedback are used in the reflection relationship as illustrated in FIG. 10E. As will be apparent to those skilled in the art, other non-linear implementations of these positive feedback relationships can be used, including piece-wise linear and piece-wise non-linear. In a preferred embodiment, positive torque feedback measures the knee torque of an artificial ankle and uses this as a non-linear feedback signal.

Is achieved by using as In other implementations, this reflected torque input can be estimated using model-based calculations of ankle motion.

  The inventors have observed that biomimetic impedance and reflexes in the stance are combined when considering the effects of walking speed and terrain tilt as shown in FIG. For this reason, in one preferred embodiment, the prosthetic limb parallel elasticity (eg, parallel or K3 spring) is chosen to represent stiffness for slow walking speeds as shown. In the biomimetic system, the rigid component of the prosthesis is attenuated when the walking speed is high, and the reflex response has a steep slope as shown in FIG. With this optimal biomimetic control and mechanical implementation, the response then requires the actuator to push the parallel spring in the control dorsiflexion and pull it in the power buckle. Here, this is called a bipolar or push-pull operation. In non-optimal control and mechanical implementation, the reflection is implemented only by the unipolar attraction of twice the magnitude. In a preferred embodiment, this reduces peak actuator force and motor current by a factor of two when an appropriate series spring bilateral response is selected, thus reducing actuator design. The service life is extended 8 times, and the ball nut speed is reduced to almost 1/2. This is very important in terms of increasing the durability of the actuator, reducing the weight of the actuator-reducing the number of ball bearings and ball nut diameter required to achieve the target design life-and reducing noise. Has great advantages.

10C is a schematic diagram illustrating a controller for performing impedance control in a prosthetic device (eg, device 1700 of FIGS. 17A-17E), according to an illustrative embodiment of the invention. FIG. 10D is a schematic diagram illustrating the relationship of the mechanical impedance that determines the impedance control performed in FIG. 10C. τ _M is a torque applied to the ankle joint of the artificial leg device by the linear actuator. Through a preferred “high gain” compensation G _c (z), where z is a discrete time signal transformation, the motor torque is 1) a series elastic actuator, 2) “K3” parallel elasticity, and 3) the desired result torque command Γ. Obviously, it acts to sum the torque applied by the acceleration torque on the ankle joint equal to _c . To represent model estimates for these mechanical parameters, and thus references to model-based controls,

and

Is used.

FIG. 10F shows how the zero moment pivot reference ground reaction force is used to determine the restoring torque required to stabilize the inverted pendulum mechanical behavior of a person wearing a prosthetic device. Applying a torque (Γ _CM ) to the center of mass of the system (eg, a person wearing a prosthetic limb and a prosthetic limb combination);

To maintain the wearer's balance. Wherein, f _l and f _t are each ground reaction force acting on the front foot and rear foot. v _CM is the velocity vector at the center of mass of the wearer. ZMP _l and ZMP _t represent zero moment pivots on the fore and hind legs.

and

Represents the coordinate reference vectors between the center of mass on the forefoot and hindfoot, respectively, and the zero moment pivot. The term zero moment pivot refers to the foot's inertia reference point, and the moment of the ground reaction force distribution is zero around this foot. Here, this point is also referred to as the center of pressure (CoP), and is used interchangeably throughout the remainder of this specification.

Ground reaction force and zero moment pivot The ground reaction force (GRF) is the force exerted on the foot (or leg member of the lower limb device) by the underlying surface. Ground reaction force is an important biomechanical input in the stance phase. By knowing the total ground reaction force acting on the zero moment pivot (referred to herein as ZMP and CoP), the prosthetic device control system (eg, controller 1712 in FIG. 17A) improves the balance (of the wearer). However, they will have direct steps to optimize power supply during the stance phase. Herr et al., U.S. Pat. No. 6,087,056, further describe methods for estimating ground reaction forces and zero moment pivot positions, as well as biomimetic motion and balance controllers and methods for use with prosthetics, orthoses, and robots (see The entire contents of which are incorporated herein).

  FIG. 11A shows the GRF component (in particular, the vector from the ankle 1104 to the ZMP).

And GRF vector

) Is a schematic diagram showing a lower limb member 1100, ankle joint 1104, and foot member 1108 of a prosthesis (eg, device 1700 of FIG. 17A) showing how it changes during the stance phase in a typical walking cycle. GRF estimation in research settings is often done by attaching a sensor to the sole. However, since reliable packaging means must preferably withstand contact stresses over millions of walking cycles, external sensing methods may not be practical for prosthetic devices and braces, typically in research settings. These sensors used cannot withstand that way. Furthermore, such means often require custom production of shoes that are often unacceptable to the wearer.

  In other embodiments of the present invention, GRF intrinsic sensing is novel by combining inertial state and leg member force / torque input 1112 (eg, using structural element 1732 of FIGS. 17A and 17E). Done in some way.

  11B, 11C, and 11D are schematic diagrams illustrating components of the apparatus 1700 of FIG. 17A. These figures show the components necessary to determine ground reaction force and zero moment pivot (linear series elastic actuator 1116 (eg, combination of linear actuator 1716 and series elastic member 1724 of FIG. 17A) and parallel spring 1120 (eg, The relationship between the force and moment between the passive elastic members 1724)) of FIG. 17A is also shown.

and

Is calculated based on the following steps.
1. The inertial measurement device and the angle input of the ankle joint 1104 are used to update the inertial state of the lower limb member 1100 and the foot member 1108. Assuming that the body is rigid, the acceleration based on the world coordinate system measured at the center of gravity (CM) of the lower limb member 1100 and the foot member 1108 and the angular velocity and acceleration of the lower limb member 1100 and the foot member 1108 are further calculated.
2. Solving for F || as a function of the force acting on the lower limb member 1100, these are resolved along the axis of the lower limb member 1100.
3. Solve for F_ | _ as a function of the moment applied by each of the force and moment components acting on the lower limb member 1100.
4). Using the values for F || and F_ | _ calculated in steps 2 and 3 above, then balancing the force applied to the foot member 1100

Solve about.
5.

Balance the moment around the ankle 1104 (i.e.

).
6).

Solve about.

Ankle Joint Behavior Due to Terrain Texture FIG. 12A shows the biomimetic Γ-θ behavior of a prosthetic limb device (eg, device 1700 of FIG. 17A) on flat ground as a function of walking speed. FIG. 12B shows that the applied ankle torque decreases rapidly with angle during power plantar flexion, especially when walking at high speeds, thereby deviating from an ideal biomimetic response, thereby performing the net The work (region under the Γ-θ curve) is significantly reduced.

  In conventional robotic systems, trajectories or other playback means are used to communicate repeatable and programmable responses. Such means are not preferred for prosthetic devices and braces because the wearer's intention may change in the middle of the regenerative segment. For example, the wearer may be walking at high speed and stop suddenly in front of, for example, a section of ice. When preprogrammed trajectories or other trajectories are reproduced, there is no easy way to interrupt them without abrupt changes in force and torque-and without introducing risk factors. In fact, that is why intrinsic means are used.

In order to extend the application of the ankle torque during power plantar flexion, a normalized contact length that depends on the walking speed is used as a means to attenuate the peak plantar flexion torque Γ ₀ . The ground contact length is estimated by measuring the inertial posture of the foot member during controlled dorsiflexion and power plantar flexion using an idealized model of the foot derived according to the description relating to FIGS. As shown in FIG. 12C, during the foot transition from plantar grounding to toe off, the idealized foot portion is below the terrain, allowing estimation of the ground contact length. FIG. 12D shows how the L _{ground-contact} changes from the plantar ground contact to the toe _lift-off .

  FIG. 12E shows how a normalized ground contact length can be used as a means of performing biomimetic behavior during power plantar flexion in a velocity dependent contact length reduction table. These tables can be calculated by dynamically measuring ground reaction forces and foot member postures of patients other than limb amputees in a controlled environment as a function of walking speed. The functional relationship between the attenuation function and the contact length can be calculated for each walking speed. These tables can be stored as reference relationships within the controller of the prosthetic device. The function can be tailored to the wearer's specific needs when the prosthetic device is worn by the wearer.

  As already explained, one motivation to use intrinsic feedback as opposed to explicit trajectory or replay means is to respond to changes in the wearer's intention (eg, decision to stop immediately) It is. The intrinsic sensing function as a means of attenuating ankle torque with ground contact length is not general enough to accommodate changes in the wearer's intentions with stops and changes in direction. Referring to FIG. 12G, in one embodiment of the present invention implemented on a prosthetic device, a time-dependent attenuation factor

Are used in series with a decrease in ground contact length. The time constant τ for this damping can be chosen to counteract the driving torque of the power plantar buckling to prevent the danger associated with changing the wearer's intention. τ is typically in the range of 50-100 milliseconds.

  Preferably, by using the prosthetic device, the wearer can walk faster with less effort on all terrain. It is not enough to respond to changes in terrain conditions (stairs, slope up / down). The change in terrain texture should preferably also accommodate this as it can also pose a risk of slipping (eg ice / snow) or sinking (slack, snow, sand, fine gravel). Using the intrinsic sensing function of the zero moment pivot trajectory can optimize walking performance and / or eliminate danger when walking on changing terrain textures.

  FIG. 12F shows a zero moment pivot vector during a typical walking motion.

It shows how the estimated y component of changes. As shown in the figure,

Must increase monotonically between the plantar ground (3) and the toe off (4). This is because during this period, it should be lifted off the surface of the terrain (as the walking cycle progresses). If at any time the velocity of the zero moment pivot moves along the negative y axis, the foot is slipping. Similar to the way anti-lock brakes are implemented in automobiles, prosthetic limb devices can reduce torque by a damping factor derived from the integration of the negative zero moment pivot speed. In one embodiment, only negative velocities below the noise threshold are integrated to reduce noise sensitivity.

  FIGS. 13A and 13B illustrate a state control situation of an exemplary embodiment of the invention applied to, for example, the apparatus 1700 of FIGS. Normal walking involves circulation between two phases, the swing phase and the stance phase. FIG. 13A shows a control system scheme with a walking motion in which the stance phase begins with a heel-contact 1320 to the ground.

Indicates the z component of the ankle joint velocity in the ground-based world coordinate system. FIG. 13B shows a walking motion in which the stance phase begins with a toe grounding 1324 to the ground.

Behavior of an exemplary control system for driving a prosthesis or orthosis throughout the walking cycle FIGS. 13A and 13B show that the control system 1300 is ankle joint when the ankle joint transitions between the swing phase 1304 and stance phase 1308 states. It shows that the behavior of is changed. The control system 1300 applies position control 1328 during the swing phase-avoids the risk of stumbling during the first half of the swing phase and avoids the risk of specific topographic conditions (slopes, stairs, steps) during the second half of the swing phase. Optimize toe contact angle of attack (adaptive ankle joint positioning). The control system 1300 provides impedance and torque in the stance phase when the ankle transitions between heel / toe grounding, foot landing, peak energy storage (dorsiflexion with exponential stiffness), power plantar flexion, and toe takeoff events. Apply control 1332—optimize the ankle inertia, spring, and damping characteristics.

FIG. 13C illustrates a method for performing position control applied to a lower limb device (eg, device 1700 of FIG. 17A), according to an illustrative embodiment of the invention. It is desirable not to advance the foot member 1348 until the wearer of the device and / or the controller of the device has confirmed that the toe 1340 has cleared the terrain in front of the wearer. One exemplary method for accomplishing this is to wait for the toe 1340 of the foot member 1348 to be sufficiently high from the last known position of the toe 1340 with respect to the underlying terrain. In this embodiment, the control system 1300 has a clearance distance measured along a vector in the normal direction with respect to the topographic surface between the toes 1340 of the foot member 1348 at time t and time tk ₋₁ from (ε ₀ ). Position control 1328 is applied by starting to rotate the ankle 1340 only after it is determined to be large. This minimizes the risk of encountering the danger that the toes 1340 will crawl. In one embodiment, the position of the toe 1344 at two different times (t and t _k-1 ) is determined using the measurement results of the inertial measurement device as previously described herein. One skilled in the art will appreciate that there are other ways of applying other schemes to determine when it is appropriate to advance the foot member 1348. In some embodiments, the controller may determine that it is appropriate to advance, for example, based on whether the swept volume of the foot provides the desired clearance with respect to the terrain surface during dorsiflexion.

  In summary, in this embodiment of the present invention, the prosthetic limb device aims to achieve true biomimetic behavior in all walking motor task situations, including walking on flat ground, stair climbing / falling, and slope climbing / falling. As step by step terrain adaptation. FIG. 14A shows an overview of the process for performing step-by-step adaptation. In the swing phase, the inertial measurement device is able to discriminate the terrain situation from the cues given by the trajectory trajectory features (ie opposite to the external neural / myoelectric input). To supply). Adaptive swing leg ankle positioning refers to the joint motion at the foot joint angle θ, which is a natural motion optimized for the most likely topographic situation determined by the topographic situation discrimination with respect to the trajectory of the swing leg period. Do nails or toe landings.

  FIG. 14B shows exemplary impedances applied by the artificial ankle for three different walking motion situations. FIG. 14B is a graph showing the relationship between the required ankle joint torque 1404 (unit: Nm / kg) and the ankle joint angle 1408 (unit: degree). The graph includes three curves 1412, 1416 and 1420. A curve 1412 shows the relationship between the ankle joint torque 1404 and the ankle joint angle 1408 for walking on a slope with an inclination of 5 degrees. A curve 1416 shows the relationship between the ankle joint torque 1404 and the ankle joint angle 1408 with respect to walking on a downward slope of 5 degrees. A curve 1420 shows the relationship between the ankle joint torque 1404 and the ankle joint angle 1408 for walking on a slope with an inclination of 0 degrees (flat ground). The slope of the curve is equal to stiffness (or generally impedance). The area enclosed by the closed Γ-θ curve corresponds to the amount of non-conservative work required for a particular terrain situation (eg, slope, stairs) and walking speed. As shown in the graph, the ankle joint prosthesis has a larger area in the curve 1412 than the area in the curve 1416. Needed to do further work to accomplish.

Generalization of Hybrid Leg Enhancement System FIG. 15 is a schematic diagram illustrating a leg biomechanical device 1500, according to an illustrative embodiment of the invention. In one embodiment, the device 1500 is a brace that enhances the wearer's ability to walk. In other embodiments, the device 1500 is an appliance that attaches to the wearer's body to support and / or correct malformations and / or abnormalities of the wearer's waist, thigh, lower limb, and foot musculoskeletal. It is. In other embodiments, the device 1500 is an exoskeleton device that attaches to the wearer's body to assist or enhance the wearer's lower limb biomechanical output (eg, to increase the strength or mobility of the wearer's lower limb). It is.

The device 1500 is a linkage represented by a plurality of links (or members) and joints connecting the links. Device 1500 includes a foot member 1508 (L ₀ ) coupled to a lower limb member 1516 (L ₁ ) by an ankle joint 1512. Apparatus 1500 also includes a femoral member 1524 (L ₂ ) coupled to lower limb member 1516 by knee joint 1520. The apparatus also includes a hip joint 1528 that couples the thigh member 1524 to the wearer's torso 1532 (L ₃ ). Center of mass 1504 is the center of mass of the combination of device 1500 and the wearer.

The foot member 1508 contacts the terrain 1536 below the lower limb member 1508 at the zero moment pivot 1540. The foot member 1508 includes a toe portion 1544 and a heel portion 1548. Each joint of device 1500 also includes an actuator having a generalized vector of torque (force) Γ _i , mutation ξ _i , and impedance K _i , where i = 0 corresponds to ankle joint 1512 and i = 1 corresponds to the knee joint and i = 2 corresponds to the hip joint. Each joint actuator may comprise a mechanical element (eg, a ball screw actuator or a rotating wave gear device), a human muscle, or both. The joint displacement typically takes the form of angular displacement (rotation), but may also include a combination of linear and angular displacement, such as found in a typical knee joint, for example. The link orientation i is represented by a 4 × 4 matrix that defines the unit vector of the coordinate system related to the position of the origin of the link and the unit vector of the world coordinate system W.

Thus, each link attitude j is transformed via linkage constraints-in particular the transformation defined by the generalized displacement ξ _j to the link attitude i-1 and the specific link parameters (link length, skew and convergence angle). It can be determined by multiplying with. For example, if the shin posture is known, the foot, thigh, and torso postures can be determined by direct sensing of their generalized displacements relative to their linkages or by using inertial sensors. It can be calculated assuming that it is known. The vector of sensor information inherent in each link is encapsulated in what is referred to herein as an intrinsic sensing unit (ISU). Examples of intrinsic sensors include direct or indirect measurement of generalized displacement, measurement of link angular velocity and acceleration (eg, using an inertial measurement unit, for example), measurement or estimation of force or torque components on the link, Multimodal computer images (eg, range maps) or measurements of the output of a particular neural pathway on or adjacent to the link.

  The terrain is modeled as a contour function z (x, y) with surface characteristics α (x, y). In this situation, the surface properties include the elasticity / plasticity of the surface, the damping properties, and the coefficient of friction that are sufficient to allow the foot to be bounced on the surface and capture the surface energy. It relates to the work required to land and push and release with the foot member.

  FIG. 16 is a schematic diagram illustrating a method for determining a posture of a wearer's thigh member, waist member, and torso, according to an illustrative embodiment of the invention. In lower limb systems that employ robotic knee prostheses or orthoses, human hip placement can also be calculated by incorporating an inertial measurement device in the thigh or by measuring the relative knee angle relative to the lower limb member. it can. If the inertial measurement device is further used in the torso, the torso posture can also be calculated instantaneously. Alternatively, the posture can be calculated by measuring the hip joint displacement with two degrees of freedom. Correction of torso posture prediction errors resulting from rate gyro and accelerometer drift on the torso inertial measurement device can be performed in a zero speed update of the lower limb member through a series of speed constraints through the hybrid system linkage.

  Figure 16 shows the body posture

, Thigh posture

And body / body mass center posture

Shows a method of posture restoration in which j, j-1 speed constraints are used to correct the prediction of. Step 1 (1604) captures the output of the zero speed update on the lower limb member 1620 (link 1) to determine the posture of the lower limb member, as described above with respect to FIGS. The solutions (steps 2 and 3) for the thigh member 1624 (link 2) and the torso member 1628 (link 3) follow the example of step 1 (1604), respectively, but in these cases the velocity constraint is non-zero. Yes, predicted by translation and rotational speed from the previous link.

Exemplary Mechanical Design FIG. 17A is a diagram illustrating a prosthetic device 1700, according to an illustrative embodiment of the invention. The device 1700 has an attachment interface 1704 that allows attachment to a wearer's complementary leg socket member. The device 1700 also includes a component 1732 (also referred to herein as a pyramid) coupled to a mounting interface 1704 and a first end 1752 of a lower limb member 1712 (also referred to herein as a shin). In some embodiments, the axial force and moment applied to the lower limb member of the device is determined based on the results of sensor measurements made using a structural member (pyramid) coupled to the lower limb member of the device. The A pyramid is a component with a measuring instrument that is a component of a prosthetic limb and couples to a wearer's limb socket. In one embodiment, pyramid measurements are used by the controller to determine the axial forces and moments applied to the lower limb member. In this embodiment, the structural element 1732 is coupled to the first end 1752 of the lower limb member 1712 with a set of pins 1711. Pin 1711 passes through a set of holes 1713 in lower limb member 1712 and a set of holes 1715 in the structural element (shown in FIG. 17E).

  The structural element 1732 has a top surface 1731 disposed toward the attachment interface 1704 and a bottom surface 1733 disposed toward the lower limb member 1712. Lower limb member 1712 is also coupled to foot member 1708 at ankle joint 1740 at second end 1744 of lower limb member 1712. Ankle joint 1740 (eg, a rotational bearing) allows foot member 1708 to rotate about the x-axis with respect to lower limb member 1712. The foot member includes a heel 1772 and a toe 1776.

  Apparatus 1700 also includes a linear actuator 1716 having a first end 1736 and a second end 1748. The linear actuator 1716 generates a linear motion 1703. The first end 1736 of the linear actuator 1716 is coupled to the first end 1752 of the lower limb member 1712 (eg, using a rotating bearing). The apparatus 1700 also includes a first passive elastic member 1728 that is in series with a linear actuator 1716. Passive elastic member 1728 is coupled to foot member 1708 and second end 1748 of linear actuator 1716. Passive elastic member 1728 is coupled to foot member 1708 at the proximal end 1730 of passive elastic member 1728 (eg, using a rotating bearing). The distal end 1726 of the passive resilient member 1728 is coupled between the second end 1748 of the linear actuator 1716 (eg, using a rotating bearing). Linear actuator 1716 applies torque around ankle joint 1740.

  The apparatus 1700 also includes an optional second passive elastic member 1724 having a first end 1756 and a second end 1760. The second passive elastic member 1724 generates a unidirectional spring force in parallel with the lower limb member 1712 (providing parallel elasticity). The first end 1756 of the second passive elastic member 1724 is coupled to the first end 1752 of the lower limb member 1712. A second end 1760 of the second passive elastic member 1724 is coupled to the foot member 1708. However, no springs are applied when buckling, thus only applying a one-way spring force to the device.

  In some embodiments, the second passive elastic member 1724 is a non-compliant stopper that stores little or no energy and limits further rotation of the ankle joint beyond a defined angle during power plantar flexion. To do.

  FIGS. 17B and 17C show a portion of the lower limb device of FIG. 17A showing a second passive elastic element 1724. The second passive elastic element 1724 stores energy when buckling, not when buckling. The elastic element 1724 has a double cantilever engagement (clamped at a location 1780 between the first end 1756 and the second end 1760). The elastic member 1724 has a tapered shape 1784 that allows the elastic member 1724 to efficiently store energy by maximizing bending strain along the entire length of the elastic element 1724 (along the y-axis). In some embodiments, the normalized spring constant is in the range of 0-12 Nm / rad / kg. At the upper limit of this range, energy storage is about 0.25 J / kg.

  With the cam / slope arrangement of the elastic member 1724, the spring constant can be easily adjusted according to the weight of the wearer. The cam element 1788 is disposed at the second end 1760 of the elastic member 1724. Slope element 1792 is disposed on foot member 1708. Cam element 1788 engages bevel element 1792 when dorsiflexed, but cam element 1788 does not engage bevel element 1792 or other portions of device 1700 when bent. Since the cam element 1788 does not engage the beveled element 1792 or other portions of the device 1700 during bottom flexion, the elastic member 1724 stores energy only during dorsiflexion. In one embodiment, the position of the bevel element 1792 is adjusted with a screw so that the wearer or a second party can modify the engagement of the bevel with the cam element 1788 to tailor the energy storage characteristics to the wearer's walking habits. Is possible. An operator can adjust the position of the slope element 1792 relative to the position of the cam element 1788 to modify the energy storage characteristics of the passive elastic member 1724.

  In an alternative embodiment, the actuator is integrated into the bevel to adjust the ankle joint angle (elastic member engagement angle) with which the second passive elastic member 1724 engages. Thereby, for example, when the wearer is climbing up the slope and the stairs, the ankle joint 1740 can be dorsiflexed during the swing phase without engaging with the elastic member 1724.

Passive elastic element 1724 also serves to increase the frequency response of device 1700 when elastic element 1724 is dorsiflexively engaged. The operation of device 1700 in dorsiflexion benefits from a fast response (bandwidth) series elastic actuator (ie, a combination of linear actuator 1716 and first passive elastic element 1728). The spring constant associated with the second passive elastic element 1724 increases the bandwidth of the device 1700 by a factor β, provided that
β = (K ₃ (1 + K _s / K ₃ ) ^1/2 / K _s ) ^1/2 formula 34
Where K ₃ is the spring constant of the second passive elastic member 1724 and K _s is the spring constant of the combination of the linear actuator 1716 and the first passive elastic element 1728. In one embodiment of the invention, the second passive elastic element provides a β of 1 to 3, thereby increasing the bandwidth of the device 1700 from about 5 Hz to about 15 Hz.

  The second passive elastic member 1724 uses dovetail features 1796 at both ends to allow for clamping at both ends without using attachment holes. In one embodiment, the second passive elastic member 1724 is fabricated from a composite fiber material. The attachment holes cause stress expansion and cause fiber dislocation in the passive elastic member 1724 that impairs the strength of the spring. The end clamp 1798 has a complementary shape that holds the passive elastic element 1724 in place. In one embodiment of the invention, epoxy is used in the clamp to permanently secure the second passive elastic member 1724 in the end clamp. Epoxy resin joints tend not to work well when dovetail feature 1796 is not present.

The passive elastic element 1724 employs a tapered design to maximize energy storage in the element 1724 so that the energy storage density is constant over the entire length for a given deflection. Referring to FIG. 17D, there is illustrated a free body diagram for a passive elastic element 1724 in which the roller force F _rollr and the lower limb member force F _shank combine to equal but not by the central pivot. It shows the mechanism that generates the direction force. In this embodiment, the roller force and the leg member force are applied equidistant from the central pivot. The multiple forces F at the ends combine to generate a 17F central pivot force. Using the standard thin beam relationship, the moment acting at a distance x from the central pivot varies linearly -L is the passive elastic element 1724 between the positions where the force is applied. Starts with the value of FL at the center, and becomes zero when x = L. The energy storage density along x is proportional to the product of the moment (M (x)) and the strain at the surface (ε ₀ (x)), provided that

It is.

For a given layup of the composite material, the surface strain is kept below the critical value ε ^* . For a given moment, the energy density in the beam is maximized when the surface strain is set to this critical value. In order to keep the energy density constant and at the maximum value, the optimum beam width w ^* (x) is

Defined by

  In one embodiment, the tapered shape 1784 varies linearly from the center of the beam. By using this design method, the energy storage of the spring was increased by a factor of two compared to the beam without the tapered shape 1784. Since the composite spring material is not homogeneous and the thin beam equation is not applicable, a calculation tool is used to estimate the energy storage density of the passive elastic member 1724. The shape that can store most of the energy is highly dependent on the fiber laminate, the laminate design, the thickness, and the exact method of attaching the passive elastic member 1724 to the device 1700. However, it was determined here that the linear taper shape stores energy within a range of about 10% of the optimum value. In one preferred embodiment, a linear taper is used because it is relatively easy to cut a linear taper pattern from a single layer composite plate using a water jet process. In an alternative, less preferred embodiment, a non-tapered spring can be used.

  FIG. 17E is a perspective view illustrating one embodiment of a structural element 1732 (also referred to herein as a pyramid). Structural element 1732 is coupled between attachment interface 1704 and first end 1752 of lower limb member 1712. The structural element 1732 is coupled to the first end 1752 of the lower limb member 1712 with a set of pins 1711 (shown in FIG. 17A). Pin 1711 passes through a set of holes 1713 in lower limb member 1712 and a set of holes 1715 in structural element 1732. The pin 1711 prevents rotational degrees of freedom for distortion in the structural element 1732 from being erroneously recorded as axial forces and moments in the structural element 1732. In this embodiment, the structural element 1732 can measure moments and axial loads on the ankle 1740, thereby providing for use with, for example, a state machine of a controller 1762 that controls the function of the device 1700. Positive detection of "foot landing", measurement of applied moments for use in positive feedback reflection controllers used during power plantar flexion, and whispering for use in safety systems integrated into controller 1762 Positive detection can be performed.

  In this embodiment, the structural element 1732 is designed as a bending element that amplifies the strain field induced by inner-outer moments and axial forces applied to the device 1700 during operation. Structural element 1732 has a high strain field (differential strain field) with opposite sign in regions 1738 and 1742 around the central adapter mounting hole 1734 when an inner-outer moment (moment about the x-axis) is applied. Is generated. These differential strain fields do not exist when only axial force is applied. Structural element 1732 includes one strain gauge (1782 and 1786) bonded to each of two moment sensing regions (1738 and 1742, respectively) on the bottom surface 1733 of structural element 1732. The gauge is attached to the opposite side of the Wheatstone bridge. Controller 1762 is coupled to the Wheatstone bridge to measure strain. The strain measurement results are used to measure the moment on the structural element 1732. In one embodiment, the sensitivity of the measurement is in the range of about 0.15 N-m, and in this situation, the sensitivity is the smallest change that can be resolved when digitally sampled at 500 Hz (signal to noise ratio≈1). Determine.

  In contrast to moment-induced strain, high strain is caused by axial forces along the inner and outer shafts in regions 1746 and 1754 around the central adapter mounting hole 1734. These distortions appear in 0.76 mm thick regions (regions 1746 and 1754) below the grooves machined along the inner and outer shafts (1758 and 1770, respectively). The upper part of this groove must be thick enough to transmit the moment load with minimal distortion in the thin lower part. The magnitude of the strain is significantly reduced in the thin area when a moment-only load is applied. Structural element 1732 includes one strain gauge (1790 and 1794) bonded to each of two axial load sensing areas (1746 and 1754, respectively) on the bottom surface 1733 of structural element 1732. The gauge is attached to the opposite side of the Wheatstone bridge. Controller 1762 is coupled to the Wheatstone bridge to measure strain. These strain measurement results are used to measure the axial force on the structural element 1732 and, therefore, to measure the axial force on the lower limb member 1712. Machined grooves 1758 and 1770 amplify axially induced strain without compromising the structural integrity of structural element 1732.

  In one embodiment, the structural element 1732 is preferably in the axial direction of more than 60 N / kg because it is in the critical chain of structural support between the wearer's remaining leg socket (not shown) and the device 1700. Designed to withstand loads. In this embodiment, the axial measurement sensitivity is in the range of about 50N, which is a threshold typically used in the device 1700 to sense that the device is firmly placed on the ground. Well below 100N. During calibration of the device 1700, a 2 × 2 sensitivity matrix is determined, whereby the true moment and axial force can be derived from several pairs of strain measurement results.

  FIG. 17F is a cross-sectional view illustrating an alternative method for measuring axial forces and moments applied to a lower limb member, according to an illustrative embodiment of the invention. In this embodiment, the structural element 1732 employs a bending design that amplifies the displacement of the bottom surface 1733 so that axial forces and in-plane moments (two degrees of freedom) can be derived redundantly. In this embodiment, the device 1700 includes a displacement sensing device 1735 that measures the deflection of the structural element 1732 to determine the moment (torque) and axial force applied to the leg member 1712.

  In this embodiment, the displacement sensing device 1735 comprises a printed circuit assembly (PCA) that employs one or more displacement sensors 1737 (eg, contact or non-contact displacement sensors). The sensor measures the distance between the sensor 1737 and the bottom surface 1733 of the structural element 1732 at each sensing coordinate.

  In one embodiment, changes in the mutual inductance of the coils printed on the PCA with respect to the bottom surface 1733 of the structural element 1732 are used to measure local surface deformation (displacement). In this embodiment, the counter-circulating “vortex” current in the structural element 1732 is used to reduce the inductance of the coil inversely proportional to the distance between the coil and the bottom surface 1733 of the structural element 1732. Other displacement sensing techniques can also be used, including contactless capacitance and contact sensors using force sensors, piezoelectrics, or strain gauges built into optical sensors or PCAs. By sampling the array of displacement sensors, the axial force and moment can be estimated using the sensitivity matrix calculated during the off-line calibration process.

  In this embodiment, the structural element 1732 is screwed to the lower limb member 1712, thus eliminating the pin 1711 used in the embodiment illustrated in FIG. 17E. By employing the screwing method, the weight and manufacturing complexity are reduced. Furthermore, this screwing method amplifies the displacement measured at the center of the structural element 1732 where the displacement sensing device 1735 is located. FIG. 17G shows how to calculate the in-plane moment vector and axial force using a circular array of displacement sensors on the printed circuit assembly. As shown, the bias and sine function-like demodulation of the displacement function is used to estimate moments and forces. Other displacement sensor arrangements can also be used as a means of intrinsic sensing of moments and forces.

  Moment and force sensing is useful as a means of signaling changes in walking conditions. In addition, the measurement of the moment of the lower limb member 1712 is used as a feedback means for causing a reflex action with power buckling. When combined with inertia and actuator feedback, the intrinsic moment and force measurements are used to calculate traction control and ground reaction forces and zero moment pivots used to balance. For these reasons, it is beneficial to package the intrinsic moment and force sensing functions along with inertial measurement devices and state control processing functions. FIG. 17F shows how these functions can be implemented on the PCA. Such a PCA has an FR-4 material and a stable core material used as a rigid interposer between a top displacement sensing FR-4 base layer and a bottom FR-4 base layer incorporating a signal processing layer. It can be implemented as a sandwich with (for example, Invar). By integrating the materials and functions into a single assembly, cabling and potentially unreliable means for interconnecting these functions are eliminated. This integration also allows for the use of stand-alone tools that can be used by prosthetic orthotics to set up passive prosthesis and research, gait parameters, including energy recovery and gait statistics. .

  Referring to FIG. 17A, the apparatus 1700 also includes a controller 1762 coupled to the linear actuator 1716 to control the linear actuator 1716. In this embodiment, the controller is placed in the housing 1764 of the device 1700 to protect it from the environment. A battery 1768 in the housing 1764 provides power to the device (eg, various sensors associated with the controller 1762 and the device 1700).

  Apparatus 1700 includes an inertial measurement device 1720 to predict the inertial posture trajectory of ankle joint 1740, heel 1772, and toe 1776 with respect to the previous toe off location. Inertial measurement device 1720 is electrically coupled to controller 1762 and sends an inertial measurement signal to controller 1762 to control linear actuator 1716 of device 1700. In one embodiment, inertial measurement device 1720 employs a triaxial accelerometer and a triaxial rate gyro. A triaxial accelerometer measures local acceleration along three orthogonal axes. A triaxial rate gyro measures angular rotation about three orthogonal axes. By using established numerical integration methods, the position, velocity, and orientation at any point on the foot structure can be calculated.

  In some embodiments, the inertial measurement device 1720 is used to detect the slope of the terrain and the presence of steps and stairs, so that the terrain below and the ankle spring in the stance phase before landing. The “attack angle” of the foot with respect to the equilibrium position can be optimized. In some embodiments, the inertial measurement device 1720 may determine the wearer's walking speed and terrain condition (eg, terrain features, texture, or irregularities (eg, how much the terrain is sticky, how slippery the terrain is). Used to determine if the terrain is rough or smooth, or if there are obstacles such as rocks on the terrain)). This allows the wearer to walk with confidence in all types of terrain. The inertial posture is a three-degree-of-freedom orientation of the lower limb member 1712 in a fixed ground reference (world) coordinate system—often the orientation component of the linear transformation (the x, y, and z axes in the world reference coordinate system Three unit vectors to be defined) or captured as a quaternion—the translation of the ankle 1740 in the world coordinate system and the velocity of the ankle 1740 in the world coordinate system. In this embodiment, inertial measurement device 1720 is physically coupled to lower limb member 1712. In some embodiments, inertial measurement device 1720 is coupled to foot member 1708 of device 1700.

FIG. 17H is a schematic diagram illustrating a state estimation and actuator controller (state and actuator control PCA-SAC) for use with the apparatus of FIGS. 17A-17G, according to an illustrative embodiment of the invention. In this embodiment, the controller 1762 uses a dual 40 MHz dsPIC (Microchip ^™ ) processor to control and adjust the linear actuator 1716 (eg, the rotary motor 504 of FIGS. 5A and 5B) and the inertial measurement device 1720. . In this embodiment, space vector modulation is used to implement a brushless motor control that generates an optimal pulse width modulated driving force that maximizes the RPM of the motor. Space vector modulation is a PWM control algorithm for multiphase AC generation in which a reference signal is periodically sampled. Signal or power supply PWM involves modulation of the duty cycle of the three-phase motor winding voltage (eg, rotary motor 504). After each time the reference signal is sampled, one or more of the non-zero active switching vectors and zero switching vectors adjacent to the reference vector are appropriate for the entire sampling period to synthesize the reference signal. Selected for part.

  The controller 1762 includes an inertial posture signal 1781 from the inertial measurement device 1720, a torque and axial force signal 1783 from the distortion measurement result of the structural element 1732, and an ankle joint angle signal from the Hall effect transducer disposed in the ankle joint 1740. 1785, motor position signal 1787 (quadrature encoder with index and absolute motor position) from an encoder (eg, encoder 2040 in FIG. 20A), strain signal 1789 from series elastic member 1728 strain sensor 1704 (see FIG. 18A). And various input signals including controller parameters 1791 (eg, device configuration data, wearer specific tuning, firmware updates). In addition, the controller 1762 outputs various signals including device performance data 1793 (eg, real-time data, error log data, real-time performance data), ankle joint status update 1795. In addition, the controller 1762 outputs a command to the linear actuator 1716 to provide an actuator feedback signal from the linear actuator 1716 (generally signal 1797), eg, a three-phase pulse width supplied to the power supply electronics of the linear actuator 1716. The modulation signal, battery power to linear actuator 1716, and current feedback measurements and temperature measurements from linear actuator 1716 are received.

In this embodiment, sensor feedback is used to identify a state change as device 1700 transitions between a stance phase and a swing phase state. By using redundant and diverse sensor measurement results, fault conditions are also identified, and the device 1700 is put into an appropriate safe state. By using an on-board real-time clock, the failure is time-tagged and stored in the on-board e ² PROM (error log). The content of the error log is retrieved wirelessly by a prosthetic brace and / or manufacturer service personnel. In this embodiment, the motor driver PCA (MD) receives a pulse width modulation (PWM) command from the SAC PCA and switches to the motor winding to pass current. The MD returns the sensed current and power information to the SAC PCA so that closed loop control can be applied.

In this embodiment, the IMU PCA is nominally mounted in the sagittal plane (a local plane parallel to the anterior part of the tibia), a 3-axis accelerometer, a 2-axis rate gyro (ω _z and ω _x ), and 1 An axial rate gyro (ω _y ) is used. In this embodiment, the definition of the coordinate system is to define the y-axis forward, the z-axis upward, and the x-axis as the outer product (y × x) of the y-axis and z-axis. The IMU receives status information from the SAC at a system sampling rate of 500 Hz. This transmits the ankle joint state vector—especially the position and velocity of the ankle pivot, the position of the heel, the position of the toe—all with respect to the sole ground contact position from the previous step.

  17I and 17J are schematic diagrams illustrating exemplary electrical equivalents of the device 1700 of FIG. 17A. Electrical circuit symbols are mechanical elements—resistors representing mechanical components with damping torque that is linearly related to speed, capacitors representing mechanical components with rotational inertia characteristics, and mechanical components with linear spring characteristics Is used to describe an inductor that represents Using this circuit notation, current corresponds to torque and voltage corresponds to angular velocity.

The circuit component is defined as follows. _Jshank is the unknown equivalent inertia of the lower limb member (shin) and the remaining leg under the knee (eg, inertia of the lower limb member 1712 of FIG. 17A), and J _Motor is the inertia of the equivalent motor and ball screw transmission assembly (eg, FIG. 17A linear actuator 1716 inertia),

Is the torsion spring constant for a series spring when compressed (eg, passive elastic element 1728 of FIG. 17A);

Is the torsion spring constant for the series spring when tension is applied, K ₃ is the torsion spring constant for the unidirectional parallel spring (eg, passive elastic member 1724 in FIG. 17A), and J _Ankle is , Rotational inertia of the foot structure under the ankle joint (eg, 17A foot member 1708). The current (torque) source in the model is defined as follows: Γ _Human is the unknown torque applied to the lower limb member (eg, lower limb member 1712) by the wearer's body, τ _motor is the torque applied by the actuator (eg, linear actuator 1716), and Γ _shank Is the torque measured using a structural element (eg, structural element 1732 of FIGS. 17A and 17E).

FIG. 17I illustrates the importance of series and parallel springs as energy storage elements. The use of stored energy reduces the power consumption required for a linear actuator if this energy is not used. In addition, as a further object of K ₃ springs, ankle - include functions as a shunt for the ankle inertia to increase the spring resonance.

  FIG. 17J illustrates how sensors are used in this embodiment to provide high fidelity position and force control and provide desirable sensor redundancy and versatility to achieve an intrinsically safe design. Indicates. As shown, ankle joint position

Is

Is derived from

It is.

  A redundant measure of θ is obtained by using a Hall effect angle converter, which can verify that the ankle joint is being properly manipulated by the control system. In one embodiment, the Hall effect transducer comprises a Hall effect device disposed on a SAC PCA in the housing 1764 of the apparatus 1700. The transducer also includes a magnet coupled to the foot member 1708. The magnitude of the magnetic field effect (the signal output by the transducer) changes in a known manner in response to angular joint rotation (ie, the movement of the magnet with respect to the Hall effect device). The Hall effect transducer is calibrated during manufacture of the apparatus 1700, for example, by measuring the transducer output against a known displacement of the Hall effect device relative to the magnet. In other ankle angle measurement embodiments, the mutual inductance measured with respect to the coil on the lower limb member has a known relationship as a function of the ankle angle, which is determined by the motor in the linear actuator, Or it can be calibrated to calculate angular displacement in a manner that is insensitive to magnetic fields generated by other stray fields. Also, as shown in FIG. 17J, the ankle moment as given by the wearer is also measured. This allows the linear actuator to adapt (eg, increase stiffness) to perform a reflective operation.

  18A, 18B, 18C, and 18D show the passive elastic member 1728 of FIG. 17A, according to an illustrative embodiment of the invention. Passive elastic member 1728 provides bi-directional stiffness and is connected in series with linear actuator 1716 and foot member 1708. Passive elastic member 1728 is coupled at one end to second end 1748 of linear actuator 1716 and at the other end to a foot member (not shown). Passive elastic member 1728 includes a strain sensor 1704 coupled to passive elastic member 1728 to measure strain in passive elastic member 1728. In this embodiment, strain sensor 1704 is a strain gauge whose response is calibrated to measure the force applied by linear actuator 1716 and then the moment about ankle 1740 applied by linear actuator 1716. The strain gauge signal is measured using the controller 1762 of FIG. 17A.

  In this embodiment, the passive elastic member 1724 is a molded carbon fiber layup that provides the desired bi-directional (function to bend in both directions) normalized drive mechanism stiffness. In one embodiment, the passive elastic member 1724 has a preferred compressive force of 14-25 Nm / rad / kg and a tension of 4-8 Nm / rad / kg. Biomechanical forces and torques strongly correspond to the wearer's weight. When scaling prostheses and braces, design parameter specifications are typically normalized. For example, such series and parallel elastic devices can be designed to yield discrete values that are intended to correspond to weight or cover multiple ranges of weight. The range of compressive force and tension reflects the variation in torque that results from the difference in the moment arm of the linear actuator relative to the ankle over the full range of rotation—from maximum buckling to maximum dorsiflexion. The series spring constant should be relatively non-compliant during swing phase dorsiflexion position control (while the spring is in compression), such as when the ankle joint is repositioned immediately after toe-off during walking. Optimized for. However, some degree of compliance is maintained to decouple the linear actuator from the impact load.

  Referring to FIGS. 18C and 18D, the high stiffness is achieved by inserting a dorsiflexion rotation bottom constraint 1708 toward the distal end 1726 of the passive elastic member (spring) 1728 and compressing the passive elastic member 1728. You can see that This constraint reduces the effective moment arm of the linear actuator 1716 after bending the series spring 1728 during compression (towards dorsiflexion). In tension, the moment arm is effectively increased by positioning the buckled upper constraint 1716 toward the proximal end 1730 of the spring constraint. The longer the moment arm, the more freely the spring beam bends, so that the spring constant in tension is reduced. In addition to the bilateral stiffness characteristics, in some embodiments, the spring constant of the passive elastic member 1728 is optimized to minimize the rotational speed of the ball screw. By design, this embodiment of the elastic member 1728 has asymmetric properties, i.e. higher compliance in tension than compliance in compressive force. The higher the compliance in tension, the greater the energy storage in the series spring 1728 for use in power buckling. Energy is released within the first approximately 100 ms involved in power planting, which reduces the required energy burden of the linear actuator 1716. In embodiments of the present invention that use a ball screw transmission assembly (eg, the ball screw transmission assembly 2024 of FIGS. 20A-20B) in conjunction with a rotary motor of a linear actuator, this is the required operation of the ball nut assembly portion of the ball screw transmission assembly. It has the further advantage that the speed and even the motor drive requirements of the rotary motor can be reduced. The spring throws out the foot member without the need for fast ball nut positioning in this case. The value optimized for in-line elasticity is in the range of 3-4 Nm / rad / kg.

  FIG. 19A is a diagram illustrating a prosthetic device 1900 incorporating a series leaf spring 1928, according to an illustrative embodiment of the invention. The device 1900 has an attachment interface 1904 that allows attachment to a wearer's complementary leg socket member. Device 1900 includes a lower limb member 1912 coupled to an attachment interface 1904. Lower limb member 1912 is also coupled to foot member 1908 at ankle joint 1940 of device 1900. Ankle joint 1940 allows foot member 1908 to rotate about the x-axis with respect to lower limb member 1912. The foot member includes a heel 1972 and a toe 1976.

  The apparatus 1900 also includes a linear actuator 1916 having a first end 1936 and a second end 1948. The first end 1936 of the linear actuator 1916 is coupled to the lower limb member 1912. Device 1900 also includes a passive elastic member 1928 in series with a linear actuator 1916. Passive elastic member 1928 is coupled between foot member 1908 and second end 1948 of linear actuator 1916. Passive elastic member 1928 is coupled to foot member 1908 at the proximal end 1930 of passive elastic member 1928. The distal end 1926 of the passive elastic member 1928 is coupled to the second end 1948 of the linear actuator 1916. Linear actuator 1916 applies a torque around ankle joint 1940.

  Apparatus 1900 also includes a controller 1960 that is coupled to linear actuator 1916 to control linear actuator 1916. In this embodiment, the controller 1960 is disposed within the housing 1964 of the device 1900 to protect it from the environment, but a portion of the housing has been removed in this view so that the contents of the housing can be seen. A battery 1968 coupled to the device 1900 provides power to the device 1900 (eg, the controller 1960 and various sensors associated with the device 1900).

  The passive elastic member 1928 of FIG. 19A is a leaf spring (for example, processed by a water jet cutting device). When a leaf spring is used, the cost of the passive elastic member 1900 is reduced, and the spring constant can be easily set to match the weight of the wearer. In one embodiment, the spring is split longitudinally (along the y-axis) and the linear actuator 1916 by lacking parallelism between the ball nut's rotational axis and the passive elastic member 1928 in series. Reduce the out-of-plane moment for the components of the ball nut (eg, FIGS. 20A and 20B). In this embodiment, the actuator torque feedback loop does not use a strain sensing function. Rather, in order to estimate the torque transmitted through the spring, the known spring constant of the leaf spring is measured against the measured spring deflection (when the measured angle θ of the ankle 1940 and the spring deflection is zero). The resulting angle of the ankle 1940 resulting from the kinematically defined angle β).

  FIGS. 19B and 19C are diagrams illustrating a series elastic spring made from two alternative portions of a prosthetic device 1900, according to an illustrative embodiment of the invention. The device 1900 has an attachment interface 1904 that allows attachment to a wearer's complementary leg socket member. Device 1900 includes a lower limb member 1912 coupled to an attachment interface 1904. Lower limb member 1912 is also coupled to foot member 1908 at ankle joint 1940 of device 1900. Ankle joint 1940 allows foot member 1908 to rotate about the x-axis with respect to lower limb member 1912. The foot member includes a heel 1972 and a toe 1976. Apparatus 1900 also includes a linear actuator 1916 having a first end (not shown) and a second end 1948. A first end of linear actuator 1916 is coupled to lower limb member 1912. The device 1900 also includes a coupling member 1988 (eg, a mounting bracket) that couples the foot member 1908 to the lower limb member 1912 at the ankle joint 1940 with a bearing for rotating the foot member 1908 about the x-axis of the ankle joint 1940. .

  Device 1900 also includes a passive elastic member 1928 in series with a linear actuator 1916. Referring to FIG. 19C, the passive elastic member 1928 has two member portions (eg, beam-like portions) 1994 and 1996. The elastic member 1928 also has a first end 1962 on the first member 1994 and a second end 1980 on the second member 1996. The elastic member 1928 also has an intermediate position 1996 where the two members 1994 and 1996 meet and the two members 1994 and 1996 pivot about each other about the x-axis. As the second member 1996 pivots toward the first member 1994, the elastic member stores energy in compression (indicated by arrow 1992) during dorsiflexion.

  The first end 1962 of the elastic element 1928 is coupled to the second end 1948 of the linear actuator 1916 with a bearing that allows rotation about the x-axis. The second end 1980 of the elastic element 1928 is coupled to a location on the coupling member 1988 with a bearing that allows it to rotate about the x-axis.

Exemplary Linear Actuators FIGS. 20A and 20B illustrate a linear actuator 2000 for use in various prosthetic, lower limb orthosis and exoskeleton devices, according to an illustrative embodiment of the invention. 20A is a perspective view of the linear actuator 2000. FIG. FIG. 20B is a cross-sectional view of the linear actuator 2000. The linear actuator 2000 can be used, for example, as the linear actuator 1716 of the apparatus 1700 of FIG. 17A or the apparatus 400 of FIG. The actuator 2000 includes a motor 2004 and a screw transmission assembly 2024 (in this embodiment, this is a ball screw transmission assembly, also referred to as a ball screw assembly) to transmit linear power along the A axis. The screw transmission assembly 2024 functions as a motor drive transmission that converts the rotational motion of the motor 2004 into linear motion. In one embodiment, the ball screw transmission assembly 2024 is a custom ball screw transmission assembly manufactured by Nook Industries (Cleveland, Ohio). As a specification, the custom ball screw transmission assembly has a thread spacing of 14 mm × 3 mm, a thrust of 4000 N at 150 mm / s, and an L1 rated life when subjected to 5 million cycles instantaneously. In some embodiments, the screw transmission assembly is a lead screw transmission assembly (also referred to as a lead screw assembly).

  The actuator 2000 includes a rotary motor 2004 having a motor shaft output 2008. The motor shaft output 2008 has a pulley 2032 coupled (eg, welded) to the motor shaft output 2008. In one embodiment, rotary motor 2004 is a high speed brushless motor (a model EC30 motor manufactured by Maxon Motor AG, Maxon Precision Motors, Inc., Hall River, Mass.). The motor 2004 includes an inductive incremental-absolute angular encoder 2040 that is incorporated into the motor 2004 to determine the angular alignment between the rotor and stator of the rotary motor 2004. The encoder 2040 controls the position of the screw 2060 of the linear actuator 2000 for “instant-on” motor commutation and redundant position feedback monitoring.

  By using the inductively-coupled encoding elements of encoder 2040, the system achieves absolute alignment between the rotor and stator (eg, 10 per revolution) simultaneously with high-precision incremental rotor position footback. (With bit resolution). By cross-checking these redundant feedback elements, it is possible to minimize the possibility that an encoder failure may cause ankle control instability. Incremental encoders have a runout of less than 300 μrad to eliminate speed fluctuations that are sensed when the ball screw transmission assembly 2024 (see below) is operating at constant speed. As a result, the torque fluctuation applied by the actuator 2000 is reduced.

The rotary motor 2004 also includes an integrated motor heat sink 2048. In one embodiment, the heat sink 2048 draws heat from the windings of the motor 2004, which allows the wearer to peak non-conservative work without exceeding the motor's coil temperature limit (typically 160 ° C). You can walk in. Heat generation of the motor is caused by resistance loss (i ² R loss) in the motor 2004 when the linear actuator 2000 transmits thrust. As the coil temperature increases, the coil resistance increases at a rate of 0.39% / ° C., which further increases the coil temperature. In addition, the motor K _t (a measure of torque as it is proportional to the motor current) typically decreases by about 20% as the coil temperature increases to a limit value. This requires increasing current consumption to do the same amount of work, further increasing the coil temperature. The heat sink in the linear actuator 2000 reduces the coil temperature rise by 40% or more. The “wear” phenomenon, which promotes motor winding insulation and motor bearing initial failure, is reduced by half each time the coil temperature drops by 10 ° C., so the life of the motor is lower than the operating temperature of the motor coil. If held, it will extend significantly. Also, using this intrinsic coil temperature sensing method, the power buckling power (current) is simply reduced when the maximum rating is approached, and finally the battery is reached when a predetermined limit of, for example, 150 ° C. is reached. The motor prosthesis can be protected from exceeding the absolute maximum rating of 160 ° C by shutting off the power of the robot, typically using a compact and lightweight motor drive, Send output in bursts. In some scenarios, output bursts can be applied repeatedly at high rates over time. Motor copper loss and eddy current loss cause excessive storage heating effects, which will increase the temperature of the motor windings. Since the copper winding resistance increases with temperature (0.39% / ° C.), the copper loss increases, thus amplifying the heating effect. Sometimes the critical winding temperature limit can be reached, in which case further temperature rise will cause permanent damage to the motor. Sensing when this temperature limit is reached is preferably done by the control system.

  Two conventional methods can be used to prevent or detect when the copper winding temperature limit is reached or reached. First, copper loss and eddy current loss are calculated during operation of the control system. These are used to operate a thermal model of the winding so that the winding temperature can be estimated. Sometimes the ambient temperature is measured to improve the measurement result of the winding temperature. The advantage of this method is that it can be implemented inexpensively. The disadvantage is that the coil temperature model is not available and is difficult to calibrate. In addition, it is often difficult to obtain a good measurement result of the ambient temperature around the motor, and the winding temperature measurement is erroneous.

  In the second method, sometimes combined with the first method, the motor case temperature is measured using a thermistor attached to the outside of the case or the inside of the motor. The advantage of this method is that direct measurements can be made. Disadvantages are that measurements are made at a single point, and attaching the sensor is expensive and often unreliable.

  A more preferred approach is to detect temperature and mitigate potential overheating conditions. In this approach, every time one step is taken at some point in the walking cycle when the ankle joint can be held in a fixed position for a short period of time to make a measurement (this eliminates the Back-EMF effect on the resistance calculation). Measure motor winding resistance. In one embodiment, the coil temperature is determined by passing a constant current (or applying a constant voltage) through the motor winding and measuring the corresponding voltage (or current) at that winding. In order to increase the accuracy, voltage is applied (or current is applied) in both the forward direction and the reverse direction, and the difference in current (or voltage) is measured.

  Since the motor drive electronics use the PWM current control method, there is all the infrastructure to make this measurement. By paying attention to the ratio of the difference between the winding resistance and the resistance when the ankle is in a stationary state (calibration constant), the winding resistance can be estimated in-situ without cost. In a typical servo system, this measurement cannot be made because the actuator must be constantly under closed loop control. However, in an artificial ankle joint, there are several times when the ankle position typically does not need to maintain precision control beyond the 5 milliseconds required to perform the measurement (the swing phase). After the winding temperature is calculated in this manner, the control system can detect when the winding is approaching the critical temperature. During these periods, the drive power available for walking is reduced or eliminated collectively until the temperature is lowered to a safe level.

  In some embodiments, the output of temperature sensor 2052 is provided to a controller (eg, controller 1762 of FIG. 17A), thereby controlling the torque output by linear actuator 2000 based on the temperature of motor 2004.

  The belt 2012 couples the pulley 2032 to the threaded shaft 2060 of the ball screw transmission assembly 2024, thereby converting the rotational motion of the motor shaft output 2008 to the linear motion of the ball nut assembly 2036 portion of the ball screw transmission assembly 2024. In some embodiments, two or more belts can be stretched in parallel, each driving only the ball screw transmission assembly 2024 of the linear actuator 2000, and thus a linear actuator 2000 for a single belt failure. Is operational. In such an event, the belt break sensor 2056 senses the condition and verifies the integrity of the belt during operation (eg, every wearer's walking cycle using a prosthesis).

  In one embodiment, an optical sensor (eg, a through beam sensor) is used as a belt break sensor, and the output signal of the optical sensor changes in a known manner when the belt breaks. In another embodiment of the present invention, a capacitive sensor is used as a belt break sensor, and the output of the capacitive sensor changes in a known manner when the belt breaks.

  In one embodiment, pulley 2032 and the belt are not used as a device for converting rotational motion to linear motion. Rather, a set of traction wheels is used as the transmission. In this embodiment, the threat of belt breakage is thereby eliminated.

  In one embodiment, if a belt failure occurs, the controller of the device in which the linear actuator 2000 is used (eg, the controller 1762 of the device 1700 of FIG. 17A) repairs the position of the foot member relative to the lower limb member. Until then, the device changes to a safe position where it can operate as a passive ankle joint. In one embodiment, the controller shorts the three leads of the rotary motor 2004 in response to the belt break sensor detecting a failure of one or more of the belts. When the three-phase electrical input is shorted, the motor 2004 provides a viscous resistance on the motor shaft output 2008. During walking, the viscous resistance keeps the rotor shaft output roughly fixed so that the device operates as a passive prosthesis. However, the device can be moved slowly so that it can move to a non-stationary equilibrium position when in the standing or sitting position. Each input lead is shorted to ground by its own individual switch.

  In one embodiment, the switch is operated by a rechargeable battery (a separate battery from the primary battery used to operate the device). By using a separate battery, the switch shorts the input lead (and puts the device in safe mode) even if a failure occurs (or even if the primary battery is exhausted).

  In one embodiment, threaded shaft 2060 includes a hollowed out portion that houses a noise attenuating material to reduce the noise generated from actuator 2000 and the device in which actuator 2000 is used. In one embodiment, threaded shaft 2060 is a stainless steel shaft with a bore diameter of 8.7 mm and a diameter of 14 mm, extending by 64 mm of the shaft length, manufactured by 3M (St. Paul, Minn.). Filled with ISODAMP® C-1002 sound attenuating material.

  The actuator 2000 also includes a radial thrust bearing 2028 that supports the tension of the belt 2024 by the rotation motor 2004 and the thrust of the screw 2036. Loads due to belt tension and thrust are present both statically and during the walking cycle.

  Ball nut assembly 2036 includes one or more recirculating ball tracks 2042 that hold a plurality of ball bearings, and this combination supports the linear motion of ball nut assembly 2036. In one embodiment, five ball tracks are used. Actuator 2000 includes a coupling element 2020 (eg, of linear actuator 1716 of FIG. 17A) that couples actuator 2000 to, for example, a passive elastic member of a leg member of a prosthetic device (eg, passive elastic member 1724 of FIG. 17A). A second end 1748).

  FIG. 21 is a perspective view illustrating a linear actuator 2100 for use in various prosthetic, lower limb orthosis and exoskeleton devices, according to an illustrative embodiment of the invention. The linear actuator 2100 can be used, for example, as the linear actuator 1016 of the apparatus 1000 of FIG. 17A or the apparatus 400 of FIG. Linear actuator 2100 is a variation of actuator 2000 of FIGS. 20A and 20B.

  The actuator 2100 includes a rotary motor 2004 having a motor shaft output 2008. The motor shaft output 2008 has a pulley 2032 welded to the motor shaft output 2008. The motor 2004 includes an inductive incremental absolute angle encoder 2040 that is incorporated into the motor 2004 to determine the angular alignment between the rotor and stator of the rotary motor 2004. The rotary motor 2004 also includes an integrated motor heat sink 2048.

  The two belts 2104a and 2104b are used in parallel rather than the single belt 2012 of FIGS. 20A and 20B. Each belt has the ability to drive the linear actuator transmission function itself with a margin of 1.5 times against belt failure, so that the linear actuator 2100 can withstand a single belt failure. In one embodiment, if a belt failure occurs, the controller of the device in which the linear actuator 500 is used (eg, the controller 1762 of the device 1700 of FIG. 17A) may be passive until the linear actuator 500 is repaired. The ankle is moved to a safe position so that it can operate as a typical artificial ankle. In one embodiment, the controller shorts the three leads of the rotary motor 504 in response to the belt break sensor detecting a failure of one or more of the plurality of belts. In such cases, one or more belt break sensors sense the condition and position the ankle in a safe position so that the system can operate as a passive prosthetic ankle until the linear actuator is repaired. Move to.

  Two belts 2104a and 2104b couple the pulley 532 to the threaded shaft of the ball screw transmission assembly (eg, the threaded shaft 2060 of FIG. 20B), thereby coupling the rotational motion of the motor shaft output 2008 to the ball screw transmission assembly. It converts into the linear motion of the ball nut assembly 2036 part. The actuator 2100 also includes a radial thrust bearing 2028 that supports the tension of the belts 2104a and 2104b by the rotation motor 2004 and the thrust of the screw. Loads due to belt tension and thrust are present both statically and during the walking cycle.

  The ball nut assembly 2036 includes one or more recirculating ball tracks that hold a plurality of ball bearings, and this combination supports the linear motion of the ball nut assembly 2036. Actuator 2100 includes a coupling element 2020 (eg, of linear actuator 1716 of FIG. 17A) that couples actuator 2100 to, for example, a passive elastic member of a leg member of a prosthetic device (eg, passive elastic member 1724 of FIG. 17A). A second end 1748).

  The actuator 2100 also includes a ball screw assembly seal 2108. Ball screw assembly seal 2108 protects the screw from, for example, contaminants (eg, sand, mud, corrosive materials, sticky materials). Such contaminants make it impossible to predict the design life of the actuator.

Exemplary lower limb orthosis (wearable robot knee orthosis)
22A, 22B, and 22C are schematic diagrams illustrating a lower limb brace or exoskeleton device 2200 (wearable robot knee brace), according to an illustrative embodiment of the invention. Device 2200 is a knee brace that enhances the lower limb function of the wearer. FIG. 22A is a top view of the device 2200. FIG. 22B is a side view of device 2200. FIG. 22C shows the inner portion of the knee joint drive assembly 2204 of the device 2200. Typical use cases of the device 2200 include, for example, increased metabolism, permanent assistance of a wearer with persistent limb lesions, or rehabilitation of a wearer with temporary limb lesions.

  Use cases of increased metabolism include, for example, when a wearer (eg, a soldier or other personnel) needs to traverse heavy and heavy weight terrain while carrying heavy loads. In this use case, the knee orthosis device 2200 enhances the wearer's own ability. Use cases for permanent assistance include wearers who suffer from persistent limb lesions (eg, knee tendon or meniscus degeneration) that are not rehabilitated. In this use case, the knee orthosis device 2200 provides permanent assistance to the wearer. Use cases involving rehabilitation of a wearer with a temporary extremity lesion include recovery of the wearer from trauma or other temporary symptoms. In this use case, the knee orthosis device 2200 functions as a programmable remote-controlled robotic tool deployed by the physiotherapist to accelerate recovery—through the progression of kinesthetic rehabilitation, and muscle memory and strength While recovering, gradually reduce the assistance. In other embodiments, the method defines a level of assistance performed by the wearer's device over a period of time, and the level of assistance performed by the wearer's device to assist in rehabilitation of a limb lesion. Including defining a physical therapy procedure to be reduced. In some embodiments, the level of assistance performed by the device is reduced based on the wearer's impedance and torque contribution to the device.

  Referring to FIGS. 22A and 22B, device 2200 includes a lower limb member 2216 (also referred to as a drive arm), a thigh member 2228, a lower limb cuff 2208, and an upper limb cuff 2212. Lower limb cuff 2208 is coupled to lower limb member 2228. Lower limb cuff 2208 attaches device 2200 to the wearer's lower limb. Upper limb cuff 2212 is coupled to thigh member 2228. Upper limb cuff 2212 attaches device 2200 to the wearer's thigh. Device 2200 includes a knee joint 2232 for connecting thigh member 2228 to lower limb member 2216. A knee joint 2232 (eg, a rotational bearing) allows the lower limb member 2216 to rotate about the x-axis with respect to the femoral member 2228.

  Referring to FIG. 22C, knee joint drive assembly 2204 includes a linear actuator that drives knee joint drum 2232 through belt drive transmission 2236. Linear actuators are a rotary motor 2240 (eg, a brushless motor) and a ball screw transmission assembly 2244 (eg, the motor 2004 and ball screw transmission assembly 2024 of FIGS. 20A and 20B). Device 2200 converts the rotational motion of motor shaft output 2256 of motor 2240 into linear motion of the ball nut assembly 2248 portion of ball screw transmission assembly 2244. The motor shaft output 2256 has a pulley 2260 coupled (eg, welded) to the motor shaft output 2256. The motor 2240 includes an inductive incremental absolute angle encoder 2264 that is incorporated into the motor 2240 to determine the angular alignment between the rotor and stator of the rotary motor 2240. The encoder also provides the position feedback signal necessary to control the position of the screw 2252 of the ball screw transmission assembly 2244 and to provide “instant-on” motor commutation and redundant position feedback monitoring.

  The belt 2268 couples the pulley 2260 to the threaded shaft 2252 of the ball screw transmission assembly 2244, thereby converting the rotational motion of the motor shaft output 2256 to the linear motion of the ball nut assembly 2248 portion of the ball screw transmission assembly 2244.

  In one embodiment, the threaded shaft 2252 includes a hollowed out portion that houses a noise reducing material to reduce noise generated by the knee joint drive assembly 2204. The knee joint drive assembly 2204 also includes a radial thrust bearing 2272 that supports the tension of the belt 2268 by the rotational motor 2240 and the thrust of the screw 2252. Loads due to belt tension and thrust are present both statically and during the walking cycle.

  The knee joint drive assembly 2204 also includes a spring 2280 for series elasticity, a spring cage 2284, a drive belt 2236, and a spring cage / belt connection 2288. In some embodiments, a drive band (eg, a spring steel flake) is used in place of the drive belt 2236. In some embodiments, a drive cable (eg, a stranded wire) is used in place of the drive belt 2236. The spring 2280 is a series passive elastic element having a function similar to that of the series elastic spring element 1728 of FIG. 17A. Spring cage 2284 includes a closed volume in which spring 2280 is disposed. Ball nut transmission assembly 2248 is coupled to screw 2252. Ball nut assembly 2248 is also coupled to drive belt 2236. Linear movement of the screw 2252 causes linear movement of the ball nut assembly 2248. Linear motion in the ball nut assembly 2248 causes linear motion of the drive belt 2236. The linear movement of the drive belt 2236 drives the knee joint 2232.

  The device 2200 includes a controller 2292 (a linear actuator 2204, a state and inertial measurement device 2294 (eg, a printed circuit assembly incorporating the inertial measurement device 1720 of FIG. 17A) control and processing functions), thereby controlling the operation of the device 2200. Drive and control. Referring to FIG. 22B, the device 2200 also includes a torque sensor 2220 coupled to the lower limb member 2216, thereby measuring the torque applied to the lower limb member 2216 by the knee joint drive assembly 2204. The sensor 2220 is used as a feedback element in the control loop of the controller 2292, thereby providing high fidelity closed-loop position, impedance, and torque (for reflection) control of the knee joint 2232. In one embodiment, an array of force sensing transducers is embedded in the cuff structure for the purpose of making a force measurement that is used to produce a fast biomimetic response.

  In some embodiments, a motor angle sensor (eg, encoder 2264) measures the motor position, and the controller controls the rotary motor and modulates the position, impedance, and torque of the knee joint 2232 based on the motor position. To do.

  In some embodiments, the device 2200 includes an angle sensor for determining the position of the belt drive transmission drum 2232 with respect to the output of the motor drive transmission, and the controller controls the rotation motor based on the position. Modulate impedance, position, or torque. In some embodiments, the device 2200 includes a displacement sensor that measures the displacement of the series spring in the motor driven transmission to determine the force applied to the series spring, and the controller controls the rotary motor to the spring. Modulates impedance, position, or torque based on applied force. In some embodiments, the inertial measurement device 2294 is coupled to the thigh member or the lower limb member to determine the inertial posture trajectory of the lower limb member, and the controller controls the rotation motor to determine the impedance, position, Or modulate the torque. In some embodiments, the device 2220 measures the torque applied to the lower limb member by the belt drive transmission, and the controller controls the rotation motor to determine the impedance, position, or based on the force applied to the lower limb member. Modulate torque. In some embodiments, the device 2200 includes an angle sensor that determines an angle between the femoral member and the lower limb member, provided that the controller controls a rotation motor based on the angle formed by the femoral member and the lower limb member. To modulate impedance, position, or torque.

  In some embodiments, the device 2200 includes a screw transmission assembly coupled to the motor shaft output instead of the motor drive transmission to convert the rotational motion of the motor shaft output into a linear motion output by the screw transmission assembly. In addition, the drive transmission assembly coupled to the output of the motor drive transmission converts the linear motion output by the screw transmission assembly into rotational motion, thereby applying torque to the knee joint to rotate the lower limb member relative to the femoral member. A redundant belt, band, or cable drive transmission coupled to the screw transmission assembly.

  Unlike the prosthetic device 2000 of FIG. 20A, the knee orthosis device 2200 operates in parallel with the human action. In metabolic augmentation and replacement applications, the knee orthosis control system provides all the necessary impedance and torque within the gait cycle. It is desirable for the wearer to be able to walk throughout the day without getting tired and without exerting force on the body augmentation side. For rehabilitation applications, the knee orthosis device 2200 supplies only a programmed proportion of impedance and torque. In such applications, the knee orthosis device 2200 is used as an extension of a physical therapist's remotely operated robot that monitors the wearer's rehabilitation.

  In one embodiment of a knee brace control system, a physiotherapist creates a procedure that is performed by a remotely operated robot with a knee brace over a period between visits to the therapist. By using the wireless interface, the patient's execution status can be fed back to the physical therapist, thereby realizing telepresence. This procedure specifies the rate at which assistance is reduced over time. As the knee brace device reduces assistance, the knee brace device calculates the impedance and torque contributions by the wearer via a biomechanical model--according to an improved response that maintains the desired net biomimetic response. Reduce assistance. The biomechanical model involves solving the knee's inverse dynamics-incorporating inertial rotation and acceleration of the lower limb, femoral and torso. This information on the degree of freedom 6 is derived from the inertial measurement device in terms of the thigh member and knee joint angular displacement. The zero speed update for the inertial measurement device is performed in the same manner as described herein.

  FIG. 26 shows the biomechanical characteristics of normal human walking during a walking movement that begins at the point of heel contact and ends at the point of heel contact. This is divided into a stance phase and a swing phase and requires movement of all elements at the hip, knee and ankle joints. Myopathy, such as IBM (inclusion body myositis), is mainly associated with quadriceps decline and adversely affects the patient's ability to run safely and efficiently. Equally important, patients with IBM cannot safely transition between resting (standing or sitting) and running.

  FIG. 27 shows the biomechanical mechanism by which quadriceps decline affects the walking motion on the flat ground. The entire biomechanical sequence lacks two methods relative to FIG. First, if the stiffness (or biomechanical impedance) of the knee joint in the first half of the stance is insufficient to absorb the impact of ground contact and apply restorative torque to maintain balance, Can be buckled in the first half of the stance (indicated by reference numeral 2710). The lack of the ability of the quadriceps to actively stiffen the knee to absorb foot contact has the ability to limit walking speed is similar to driving a car without braking. Biomechanical impedance refers to the stiffness that the joint acts on, and is defined by three components, the spring constant and parallel position, and the torque depending on the displacement of the joint, the spring component that applies a linear or nonlinear restoring torque, the speed of the joint Known to process damping components that apply linear or non-linear viscous restoring torques in response, and inertial components that apply linear restoring torque in response to joint accelerations, in the second, second half of stance and first half of the free leg Note that the quadriceps cannot apply brake torque to the knee to prevent fast knee flexion (indicated by reference numeral 2720).

  The knee device shown in FIGS. 22A-C can be used as a robot-assisted solution for handling the shortfalls identified above in connection with FIG. FIG. 28 is a biomimetic of the patient's knee joint according to the phase of the walking cycle by using essentially sensing the lower limb trajectory and to reproduce the desired walking trajectory state, terrain, posture and stability. We demonstrate restoring normal walking motion by applying dynamic impedance, incremental torque and position control.

  In the first half of the stance (indicated by reference numeral 2810), the knee device absorbs foot contact energy and provides increased stiffness that provides a stable platform where the hind leg can leave the ground. In the second half of the stance, such as plantar flexion (shown at reference numeral 2820), the knee device uses intrinsic measurements of thigh and shin orientation and the wearer's walking speed to produce a fully functional femoral quadrilateral. It is identical to that supplied by the head muscles and applies biomimetic and reflex torque that propels the wearer up and forward for metabolically efficient walking. Later, when the knee device senses that the lower limb is in the swing phase (shown at reference numeral 2830), the knee is bent, high impedance is applied, and the lower limb is absorbed when the foot touches the ground. (Apply the brake). This results in a safe and metabolically efficient gait.

Equilibrium Using Ground Reaction Force and Zero Moment Axis FIG. 23A illustrates the general problem of balancing in slopes with variable (positive or negative) slopes. The problem seems to involve the problem of multi-link “inverted pendulum” that is easy to implement nonlinear feedback control. In such a solution, the link angle and link mass properties (in this case, leg, conductor, head, and arms) are used to explicitly stabilize the multilink system. However, such explicit input is not incorporated into most implementations of prosthetic limbs, lower limb orthoses, or exoskeleton devices, and is therefore difficult if it is not possible to implement and package in a reliable manner to the wearer It is. Further, in some cases, the wearer has one intact leg tag and is part of the stabilization outside the prosthetic leg, lower limb orthosis, or exoskeleton device, and the prosthetic leg, lower limb orthosis, or exoskeleton The device enhances the function of the intact leg.

  In addition, FIG. 23B shows that there is an acceptable set of solutions to the balance problem. In particular, there are infinitely many solutions for knee bending that are completely acceptable and even desirable depending on the person's intention (eg, picking up heavy baggage or boxes, or balancing while playing games). Thus, the desired solution provides an intrinsic sensing function (for a prosthetic leg, lower limb orthosis, or exoskeleton device) that complements an intact balance-generating body part to achieve equilibrium in accordance with a person's intention. use.

  Solutions employed in some embodiments of the prosthetic leg, lower limb orthosis, or exoskeleton device use a simplified representation of the problem as modeled in FIG. 23C. In this representation, the stability of driving the ankle torque (eg, torque transmitted to the ankle joint by the linear actuator of the prosthetic device) through the inherent sensing of the lower limb member's inertial state, knee joint angle, and inertial reference ground reaction force Used as generalized feedback. The body is modeled as a series of masses on a thin, zero-buckled beam (only one is shown in the figure) with a mass and moment of inertia that change over time.

The balance is taken based on the following details. The desired equilibrium state is achieved when the following conditions are met:
1. ^W F _GRF align to the world z.
2. The line connecting the zero moment pivot and the ankle is aligned with the world z unit vector.
3. All time derivatives of inertial leg member angle γ and ankle joint angle θ are zero.

  Then

Leads to a feedback control law that drives to the equilibrium state of each of these conditions, where

Optimizes the secondary cost index J,

and

And the component of k is selected to emphasize the dynamic contribution of the link angle to the cost index. In this embodiment, the control law solution is realized by a linear secondary regulator (LQR) method. For non-professionals, this is determined by setting the (regulatory) controller that controls the machine or process using the mathematical algorithm described above and minimizing the cost function with the weighting factor supplied by the person. Is meant to be. “Cost” (function) is often defined as the sum of the deviations of important measurement results from the desired value. In fact, the algorithm therefore seeks a controller setting that minimizes undesirable deviations, such as deviations from the desired work performed by the wearer's prosthesis. In many cases, the magnitude of the control action itself is included in this sum to keep the energy consumed by the control action itself limited. In fact, the LQR algorithm optimizes the controller based on the weighting factor engineer designation. The LQR algorithm is just one automated means of finding an appropriate state feedback controller at its core.

  Although the use of a secondary cost index is not required, in one embodiment, the secondary cost index is used as an optimization criterion, which provides an objective frame for analysis and customization within the sales office for prosthetic leg wearers. A workpiece is formed, which allows an acceptable feeling to be maintained when the system attempts to maintain the wearer's equilibrium in different terrain. It is rare for a control engineer to prefer an alternative conventional method such as full state feedback (also called pole placement) rather than using an LQR algorithm to find a controller. This allows the engineer to see a fairly clear link between the adjusted parameters and the resulting changes in controller behavior.

Wearer Assist when Standing Up from a Chair FIGS. 24A, 24B, and 24C are for applying a balance control law to assist a wearer of a prosthetic device standing up from a chair, according to an illustrative embodiment of the invention. It is a figure which shows the method of. TUG (Timed Get Up and Go) is often used as an experimental means to assess dynamic and functional balance. The wearer is verbally commanded to get up from the chair, walk 3 meters, cross the line drawn on the floor, change direction, walk back, and sit down. In order to perform good “TUG” execution, prosthetic legs are often provided with “rise” and “sitting” buttons, which constitute the operating status of the prosthetic limb control system. In a prosthetic device incorporating the principles of the present invention, in one embodiment, there is no explicit need to set the operating status, for example, by pressing a button. Sitting, standing up, and sitting movements are identified by intrinsic sensors in the prosthetic device. The rising and sitting control actions are a simple part of maintaining the wearer's balance.

  FIGS. 24A, 24B, and 24C illustrate how the intrinsic balance control algorithm works to assist the wearer as he stands up from the chair. Referring to FIG. 24A, the transition from the start of seating to rising involves three states. Initially the foot is only off the ground or touching the ground lightly. A prosthetic device (eg, device 1700 of FIGS. 17A-17E) recognizes the wearer's weight, the inertial orientation of the lower limb and foot members, and ground reaction forces (eg, as determined with respect to FIG. 11A). Thus, the device “knows” or senses that the wearer is sitting. When the wearer starts to stand up, the increase in ground reaction force is recorded, and the state of the foot (ground contact with the sole) is determined through the measurement result of the inertial measurement device and the measurement result of the ankle joint angle sensor. Intrinsic balance control law execution begins. In this second state, the linear actuator 1716 is commanded to increase the torque applied to the lower limb member (eg, ankle joint 1740) using an imbalance sensed by an imbalance in ground reaction force. Propelled forward by the controller 1762) as a means of pulling the torso (center of mass) over the ankle joint.

  Referring to FIG. 24B, the inherent balance control continues the operation of balancing the wearer in front of the chair. FIG. 24C shows that the wearer in mid-stance balance can start walking if desired. As shown, the wearer's intent, and more specifically, the sitting / standing up situation, can be determined by a sensing function inherent in the prosthetic device. This avoids the implementation cost and complexity of explicit situation switching (pressing a button). Prosthetic devices complement and enhance body function in a natural way.

  Ankle torque induced by ground reaction force (GRF) is a preferred method of achieving exponential hardening in mid-stance. Unlike using torque on the lower limb (eg, torque measured using structural element 1732 of FIG. 17A), the ankle torque calculated with GRF is a measure of the torque applied by the ground on the ankle. It has become. GRF is often measured by a force plate in a gait research setting, thereby being used as a measure of how an intact ankle contacts the ground during gait. The GRF defines what the biomimetic ankle motion is in different terrain situations. The advantage of using GRF as a means to achieve exponential curing is that it is easy to measure performance with respect to biomimetic criteria. In addition, the use of this scale ensures invariance to the terrain orientation because it is derived from the intrinsic inertial sensing function (eg, using the inertial measurement device 1720 of FIG. 17A).

  Standing assistance can also be implemented using the knee brace shown in FIGS. 22A-C, or more generally any active knee brace or prosthesis. The apparently simple task of raising oneself from sitting to standing is in fact the involvement of knee and hip extensors to apply lift-off torque and restorative balance to step-like sequence movements. It is a complex task to incorporate. FIGS. 29A-D show the normal rising sequence of a healthy person, the rising sequence being four phases: sitting as shown in FIG. 29A; (b) stepping up in preparation for rising as shown in FIG. 29B Repositioning the knee; (c) the sitting-to-standing transition shown in FIG. 29C; and (d) the balance shown in FIG. 29D once the body is lifted.

  FIGS. 30A-D illustrate a problem in implementing the same standing for a person wearing an active leg prosthesis or prosthesis because some of the same four phases cannot be performed. More specifically, starting in the sitting position shown in FIG. 30A, many orthotic or prosthetic wearers maintain sufficient strength in their arms to position them on a chair, so that they Pre-position the knee and torso over the ankle joint, allowing phase b shown in FIG. 30B to be implemented. However, the lack of strength in the quadriceps and / or hip extensors prevents rotation of phase c that lifts the upper limbs and torso out of the chair as indicated by X, 3010 in FIG. 30C. In addition, for patients with inadequate knee and / or hip muscle strength when there is not enough extensor torque at the knee or hip, the torso shown by X, 3020 in FIG. It is virtually impossible to maintain balance once the posture is achieved.

  31A-D can be used to assist in the sequence of standing up for patients suffering from limb pathology, such as active braces (as shown in FIGS. 22A-C) or active knee prostheses. How the knee device is used. This requires the ability to recognize that the user is currently sitting and also to distinguish between the user's desire to remain seated and the desire of the user to start standing.

  One indication that the user is currently sitting is that the thigh is substantially horizontal, as shown in FIG. 31A. In the sitting position, the gravity vector is mainly in the xy plane of the thigh (perpendicular to the z-axis), so the thigh posture is an inertial sensor attached to the knee device that measures the gravity vector relative to the thigh coordinate system. Can be determined. Another guide that is sitting is based on the angle of rotation of the shin relative to the thigh (which can be obtained using an angle encoder) since the ankle joint is typically well in front of the knee in a relaxed sitting position. These states can be detected and relied upon by a system that remains in a sitting mode, for example, by maintaining a low joint impedance consistent with a relaxed sitting position.

  These same sensors then detect the desire to transition to the user's standing position by detecting when the ankle joint is placed closer to the knee joint, as shown in FIG. 31B. Can be used. When this situation is detected, the system preferably attempts to confirm the user's desire to stand by initiating a startup procedure on a trial basis. The startup procedure proceeds when the system receives positive feedback from the user. However, if the system has not received positive feedback from the user, the startup procedure is aborted.

  This trial start-up of the stand-up procedure can be implemented by having a knee device that applies a gradually increasing torque 3110 according to the estimated upward and forward speed of the hip shown in FIG. 31C. Positive feedback may be provided by the user by attempting to lift her body with her arm to vertically reposition the hip joint to help the knee device lift. In this situation, the hip joint repositions more vertically and the gravity vector begins to transition toward the z-axis. The system interprets these states as confirmation that the user actually wants to stand. In response, the system applies a greater torque to keep the user lifted.

  Alternatively, the positive feedback from the user may be in the form of an electromyographic recording signal measured from the wearer's thigh and / or hip musculature using surface or built-in electrodes. More specifically, once the device detects that the knee joint has moved forward relative to the ankle joint, a standing up procedure is initiated by the wearer bending her quadriceps and / or hip extensor. obtain. The device then measures these electromyographic recording signals, amplifies and filters each muscle output, and uses conventional techniques to characterize the signal such as amplitude, variance, and / or frequency. Extract. These extracted features can then be used to discriminate whether the user intends to remain seated or initiate a standing up procedure.

  Optionally, a foot pressure sensor can be used to detect ZMP, a torso attached to the IMU can be used to detect CMP, and the associated ground reaction force vector is It can be used as feedback to keep the wearer balanced across the lower limbs. Preferably, these pressure sensors are coupled to the knee device controller using a suitable wireless interface (e.g., Bluetooth (R)). As the patient is approaching standing, the knee device applies a restoring torque 3120 to assist in achieving balance while the wearer is standing, as shown in FIG. 31D. Apply. The application of knee torque manipulates the wearer's center of gravity force vector throughout, thereby zeroing the ZMP-CMP over time. An example of a suitable approach to implement this can be found in US Pat. No. 7,313,463, which is hereby incorporated by reference.

  Note that the user also occasionally shifts position in her chair while sitting in a way that triggers a trial start of the standing up procedure. However, the applied knee device torque is initially relatively small and potentially disappears over time (eg, 1-2 seconds) if no positive feedback is received from the user. An example of no positive feedback for a system without EMG recording may be when the user does not attempt to lift the torso with her arm and the hip joint does not begin repositioning. An example of a system with electromyography without positive feedback is when the user does not substantially activate her knees and / or hips to confirm her intentions against the intent of standing up It can be. If no positive feedback is received, the system aborts the start-up procedure and relaxes and returns to sitting.

Optimization Methods FIGS. 25A and 25B show: 1) Transition work W _t performed for weight transfer from the hind leg to the forefoot during both leg support periods of the walking cycle, 2) Minimization of hip joint impact force and power factor, or 3 ) Is a schematic diagram illustrating controlling the lower limb device based on a stochastic optimization of minimizing the combination of both cost (objective) functions FIG. 25A shows a simplified model that is used to calculate the work of the transition. FIG. 25B shows a simplified model used to calculate hip impact force and power factor.

  The term probabilistic is a hip-impact force and power factor constraint that is optimized by assuming probability (likelihood) functions for human intention, biomechanical feedback (including walking speed), terrain conditions, and terrain characteristics. Represents minimizing the expected value of the objective function. Optimization is achieved by modification of impedance, torque, and position control parameters in the control algorithm. By effectively minimizing the negative impact of ground contact and maximizing the positive impact of ground forces induced by reflections on the hybrid system energy, the kinetic energy is minimized and the hip impact force constraint is met. It is.

  The above optimization can be performed in real time by introducing “evolutionary” perturbations to key components contributing to biomimetic behavior and measuring the kinetic energy resulting from those evolutionary perturbations. The kinetic energy can be estimated using a biomechanical model to enhance the inertial measurement device feedback or, in special cases, a temporary inertial measurement device subsystem (torso and / or upper limb). Can be used to facilitate estimation of torso posture and body mass center velocity. Introducing intelligent evolution of parameters using the Fletcher-Powell method (or other suitable optimization methods known to those skilled in the art), An optimal value can be calculated. This optimal value may change over time due to the effect of augmentation on rehabilitation. By applying these evolutionary perturbations continuously and slowly over time, the optimum value can be determined continuously. Alternatively, this evolutionary optimization can be done in a fairly short time interval, i.e. between 5 and 10 minutes, as in the initial wearing of a prosthesis or brace or during medical diagnosis.

  The following describes different periods of the subject's walking cycle, and in one embodiment, the steps performed by the artificial ankle according to the principles of the present invention sense the movement of the artificial ankle and control the artificial ankle It is a step to do.

Controlled bottom flexion When subjected to impact, check that the ground reaction force and zero moment pivot correspond to the part of the foot that we expect to hit the ground first (from the terrain discrimination model). Verify that there is a corresponding change in ankle joint angle (or ankle torque) and that the appropriate end of the foot is stationary. After the impact, find a state where the local terrain slope corresponding to the inertial foot contact angle is much smaller than expected. Saturates the restoring force of the ankle spring and increases damping when it is detected. For terrain discrimination, based on biomechanical model feedback, confirm that the terrain assumptions (slope vs. staircase) are correct and that the wearer is not speaking. For example, a stair event may be detected as a large negative force in the y direction instead of a large z force centered on the front part of the foot. For terrain texture, either the heel or the front part of the foot is impacted first. The inelastic component of settlement associated with this impact is calculated. On hard ground, settlement is negligibly small and only elastic deformation (foot module, linear actuator) is observed. On muddy or soft ground, terrain plasticity is observed by looking at the trajectory of the foot causing the impact. Terrain plasticity is used as an attenuation for the net work performed in the walking cycle. Slip can also be detected by noting the forward speed of the foot that produces an impact after impact. The escalator or means of transporting the person notices that the shin angle does not rotate according to the forward speed of the foot and signals that the wearer is well balanced and is stepping on the moving surface step by step Can be detected. For ankle joint impedance control, an optimal impedance is applied using the estimated terrain reference speed attack angle (y) lower limb momentum, estimated terrain slope, and terrain characteristics. With respect to reflex control, when slip is detected, a balance restoration reflex is generated and the knee is moved over the ankle joint. For balance control, the optimal balance is usually obtained by taking the spring equilibrium state as a reference for inertia after the local terrain slope estimate is updated at the plantar contact. When the terrain is slippery, the algorithm to maintain balance introduces a positive torque “reflex” to “pull” the shin forward and help the wearer to position the knee over the ankle joint. Assist-This aligns the center of body mass with the estimated ground reaction force.

Controlled dorsiflexion After the plantar contact is detected, the controller is able to adjust the ankle joint under static conditions when the wearer is standing against gravity at this tilt. The spring equilibrium angle is used as the inertia reference so that the restoring torque is not applied by. At this point, the local terrain situation is now accurately known. The foot reference coordinates at the “foot contact” position are also defined for use in evaluating the impact of the terrain texture. For terrain texture, the algorithm measures an integrated measure of slip and deformation with respect to the “Foot Ground” criterion by measuring the way the foot moves under impact between the foot and ground contact. Use and update the terrain feature model-specifically measure the plasticity of the surface and its slipperiness. These measures can be used to attenuate ankle impedance and net work (reflex torque in late plantar flexion). Also, if a “slip” is detected between the foot contact and the plantar contact, the algorithm implemented in the controller also looks at the angular velocity of the shin (how the knee moves relative to the ankle joint). Distinguish between slippery surfaces and means of transporting escalators / people. In either case, a zero speed update is not scheduled because a reliable “ankle at zero speed” is not available at this step. If the terrain is slippery, the balance function needs to call a special scale. When the foot is on an moving escalator or means of transporting people, the nominal impedance can be used on the new inertial coordinate system. For impedance control, the control system maintains the inertial reference equilibrium angle and configures walking speed dependent stiffness (lower stiffness as walking speed is faster) so that high level net work can be used, and slippery, Or an optimum impedance can be applied that reduces the stiffness at the highly plastic surface. With respect to reflex control, when slip is detected, a balance restoration reflex is generated and the knee is moved over the ankle joint. For balance control, the optimal balance is usually obtained by taking the spring equilibrium state as a reference for inertia after the local terrain slope estimate is updated at the plantar contact. When the terrain is slippery, the algorithm to maintain balance introduces a positive torque “reflex” to “pull” the shin forward and help the wearer to position the knee over the ankle joint. Assist-This aligns the center of body mass with the estimated ground reaction force.

The power plantar model monitors slip and sinking into the surface and identifies ankle torque limits that can be used to make walking motion efficient in these conditions. For terrain texture, terrain feature estimation is refined in this state and used as input to impedance, reflection, and balance functions. For impedance control, the nominal impedance parameters are modified to accommodate changes in walking speed, terrain surface characteristics and deformation, and foot slip. A special “force field” —typically a non-linear actuator force that increases exponentially as the ball nut approaches a predetermined end stop limit—is applied by the motor controller, thereby allowing the K ₃ spring to It is ensured that the energy (in the parallel elastic members) does not exceed the lower limit of its failure limit. For reflection control, the reflection amplitude is adjusted to take into account the net work “setpoint” from the biomechanical model in combination with the extent to which the terrain is adapted to produce this net work. For balance control, the optimal balance is usually obtained by taking the spring equilibrium state as a reference for inertia after the local terrain slope estimate is updated at the plantar contact. When the terrain is slippery, the algorithm to maintain balance introduces a positive torque “reflex” to “pull” the shin forward and help the wearer to position the knee over the ankle joint. Assist-This aligns the center of body mass with the estimated ground reaction force.

In the first half of the swing leg, in the first half of the swing leg, the model immediately monitors the inertial trajectories of the ankles, heels, and toes, and the ankle is bent back without being obstructed by the terrain. Determine when you can return. The model calculates an optimal trajectory with suitable impedance gain and feedforward torque to move the ankle joint to the neutral position in the fastest, efficient and stable state (and avoid the risk of crawls). For terrain discrimination, the model starts tracking the swept (“contact” with the foot member) volume that the foot has passed, so that the toe landing solution is the only feasible solution (eg, shallow staircase) Or to notify the adaptive ankle joint positioning function in the latter half of the swing leg. For impedance control in the first half of the swing leg, the neutral value of impedance is applied by the controller. A force field function is applied so that the linear actuator does not affect the hard stop (end of movement)-a condition that could cause the actuator to stick there (at the end of movement). In the impedance control in the first half of the swing leg notified by the hybrid biomechanical model, the controller sets the impedance to form a trajectory that exponentially propels the equilibrium position (ankle angle setpoint) to the desired neutral position. Control. For example, a feedforward torque function is applied to reduce the interaction between the impedance characteristic and the ankle angle according to errors that could otherwise cause overshoot and ringing.

Late Legs For terrain discrimination, the model tracks the “clear” volume that the foot has passed, so that the toe landing solution is the only feasible solution for landing on a shallow staircase or ledge, for example. To notify the ankle positioning function in the second half of the swing leg. More generally, the ankle trajectory is monitored and a pattern recognition function is used to determine the likelihood that the foot will land on a staircase / projection as opposed to an inclined surface. One simple method that has been found to discriminate between the two states is to measure the angle that the ankle velocity makes with respect to the vertical direction, and various experiments have shown that this angle is 10 degrees. When it was less than, it was determined that the foot landed on a horizontal step. For impedance control notified by the terrain discrimination model, the ankle trajectory (equilibrium) is modified by the controller when necessary to avoid the danger of stumbling. For example, if the terrain discriminator assigns maximum likelihood to stair climb, an additional dorsiflexion can be commanded so that the toes do not get caught on the stair or ledge. As before, the hybrid biomechanical model schedules a continuously updatable equilibrium trajectory that can be traced safely and in a suitable manner with tight tolerances. In the second half of stance, the biomechanical model calculates the optimal equilibrium angle and ankle impedance that minimizes the objective function including some combination of kinetic energy and knee-hip impact forces. This optimization function can be performed via a table lookup in the state machine ROM. Alternatively, in a preferred embodiment, the state controller function performs the optimization in real time by calculating and optimizing the objective function using approximations of rigid body dynamics.

  Variations, modifications, and other implementations of what is described herein will occur to those skilled in the art without departing from the spirit and scope of the invention as set forth in the claims. . Accordingly, the invention is to be defined not by the preceding illustrative description but instead by the spirit and scope of the following claims.

Claims (20)

An active orthosis or prosthetic device, the device comprising:
A thigh member;
A lower limb member;
A knee joint for connecting the thigh member to the lower limb member;
A rotary motor with motor shaft output;
A motor drive transmission assembly coupled to the motor shaft output;
A drive transmission assembly coupled to the output of the motor drive transmission device, wherein the output of the drive transmission assembly is coupled to the lower limb member and applying torque to the knee joint to the thigh member; A drive transmission assembly for rotating the lower limb member;
At least one sensor having at least one output, wherein a position of the knee joint relative to an ankle can be determined from the at least one output while a wearer of the device is in a sitting position. When,
Including the controller and
The controller determines a time for the knee joint to move to a position in front of the ankle joint based on the at least one output of the at least one sensor, and in response to the determination, stands from the sitting position. An apparatus that controls the rotary motor to modulate the impedance, position, or torque of the knee joint to assist in lifting the person into position.   The apparatus of claim 1, wherein the at least one sensor includes an inertial sensor that measures a gravity vector relative to a femoral coordinate system.   The apparatus of claim 1, wherein the at least one sensor detects a rotation angle of the lower limb relative to the thigh.   The apparatus of claim 1, wherein the controller is programmed to maintain a low joint impedance in response to a determination that the wearer is in a sitting position.   The apparatus of claim 1, wherein the controller is programmed to confirm the wearer's desire to stand by triggering a start-up procedure on a trial basis.   6. The controller of claim 5, wherein the controller is programmed to abort the rising procedure if no positive feedback is received, so as to proceed to the rising procedure when positive feedback is received. The device described.   The controller controls the rotary motor to help lift the person from the sitting position to a standing position by gradually increasing torque according to the estimated upward and forward speeds of the hip joint. The apparatus according to 1.   The controller controls the rotary motor so that when the patient approaches the standing position, a restoring torque assists in achieving balance while the wearer is in the standing position. 8. The device of claim 7, wherein the device is applied as follows.   The apparatus of claim 1, further comprising at least one pressure sensor configured to measure a force applied to the wearer's foot.   The controller helps to lift the person from the sitting position to a standing position by increasing knee torque according to an electromyographic recording signal measured from at least one of knee and hip muscles; The apparatus of claim 1, wherein the apparatus controls the rotary motor. A method for controlling a knee orthosis or prosthesis having at least one actuator, wherein the knee orthosis or prosthesis is worn by a person, the method comprising:
Detecting the position of the person's knee relative to the person's ankle joint while the person is sitting;
Determining the time that the knee is moved to a position in front of the ankle based on the result of the determining and generating an output representative of the time;
Activating at least one actuator of the knee brace or prosthesis to assist in lifting the person from the sitting position to a standing position in response to the output.   The method of claim 11, further comprising measuring a gravity vector with respect to the femoral coordinate system.   The method of claim 11, further comprising detecting a rotation angle of the person's lower limb relative to the person's thigh.   The method of claim 11, further comprising maintaining a low joint impedance in response to a determination that the person is in a sitting position.   12. The method of claim 11, further comprising the step of confirming the wearer's desire to stand by initiating a stand-up procedure on a trial basis. Determining whether positive feedback is received;
If it is determined in the determining step that positive feedback is received, proceeding to a start-up procedure;
16. The method of claim 15, further comprising the step of aborting the start-up procedure if it is determined in the determining step that no positive feedback has been received.   The method of claim 11, further comprising gradually increasing torque according to the estimated upward and forward speed of the hip joint.   18. The method of claim 17, further comprising applying an applied restoring torque to assist in achieving balance while the person is in the standing position.   The method of claim 11, further comprising measuring a force applied to the person's foot.   12. The method of claim 11, further comprising increasing knee torque according to an electromyographic recording signal measured from at least one of knee muscle and hip muscle.