U.S. patent application number 13/115175 was filed with the patent office on 2011-09-15 for powered leg prosthesis and control methodologies for obtaining near normal gait.
This patent application is currently assigned to Vanderbilt University. Invention is credited to Michael Goldfarb, Jason Mitchell, Frank Charles Sup, IV, Huseyin Atakan Varol, Thomas J. Withrow.
Application Number | 20110224803 13/115175 |
Document ID | / |
Family ID | 41201795 |
Filed Date | 2011-09-15 |
United States Patent
Application |
20110224803 |
Kind Code |
A1 |
Goldfarb; Michael ; et
al. |
September 15, 2011 |
POWERED LEG PROSTHESIS AND CONTROL METHODOLOGIES FOR OBTAINING NEAR
NORMAL GAIT
Abstract
A powered leg prosthesis includes powered knee joint comprising
a knee joint and a knee motor unit for delivering power to the knee
joint. The prosthesis also includes a prosthetic lower leg having a
socket interface coupled to the knee joint and a powered ankle
joint coupled to the lower leg opposite the knee joint comprising
an ankle, joint and an ankle motor unit to deliver power to the
ankle joint. The prosthesis further includes a prosthetic foot
coupled to the ankle joint, at least one sensor for measuring a
real-time input, and at least one controller for controlling
movement of the prosthesis based on the real-time input.
Inventors: |
Goldfarb; Michael;
(Franklin, TN) ; Varol; Huseyin Atakan;
(Nashville, TN) ; Sup, IV; Frank Charles;
(Nashville, TN) ; Mitchell; Jason; (Franklin,
TN) ; Withrow; Thomas J.; (Brentwood, TN) |
Assignee: |
Vanderbilt University
Nashville
TN
|
Family ID: |
41201795 |
Appl. No.: |
13/115175 |
Filed: |
May 25, 2011 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
12427384 |
Apr 21, 2009 |
|
|
|
13115175 |
|
|
|
|
61046684 |
Apr 21, 2008 |
|
|
|
Current U.S.
Class: |
623/40 |
Current CPC
Class: |
A61F 2002/764 20130101;
A61F 2002/6836 20130101; A61F 2/64 20130101; A61F 2/72 20130101;
A61F 2002/665 20130101; A61F 2002/5073 20130101; A61F 2002/701
20130101; A61F 2/6607 20130101; A61F 2002/6657 20130101; A61F
2002/30359 20130101; A61F 2002/7625 20130101; A61F 2002/6642
20130101; A61F 2002/704 20130101; A61F 2002/7645 20130101; A61F
2220/0033 20130101; A61F 2002/7635 20130101; A61F 2/66 20130101;
A61F 2002/768 20130101; A61F 2002/705 20130101; A61F 2/60
20130101 |
Class at
Publication: |
623/40 |
International
Class: |
A61F 2/64 20060101
A61F002/64 |
Goverment Interests
FEDERAL RIGHTS STATEMENT
[0002] The U.S. Government may have certain rights to the invention
based on National Institutes of Health Grant RO1EB005684-01.
Claims
1-25. (canceled)
26. A powered leg prosthesis, comprising: a powered knee joint
comprising a knee joint and a knee motor unit for delivering power
to the knee joint; a powered ankle joint coupled to the knee joint
comprising an ankle joint and an ankle motor unit to deliver power
to the ankle joint; a prosthetic foot coupled to the ankle joint; a
plurality of sensors for measuring a real-time input; and at least
one controller for controlling movement of the prosthesis based on
the real-time input, wherein at least one of the knee motor unit or
the ankle motor unit is configured to deliver power to a
corresponding one of the knee joint and the ankle joint using at
least one of a cable drive or a friction drive.
27. The leg prosthesis of claim 26, wherein the one of the knee
motor unit and the ankle motor unit comprises a motor having a
drive shaft for generating torque, an output stage comprising the
corresponding one of the knee joint and the ankle joint and an
output gear for providing the power to the corresponding one of the
knee joint and the ankle joint based on the torque, and a drive
stage comprising a pinion and at least one drive gear for
transmitting the torque between the drive shaft and the output
stage.
28. The leg prosthesis of claim 27, wherein the friction drive
comprises at least one roller preloaded against at least one drive
gear for transmitting the torque between the drive shaft and the
output stage.
29. The leg prosthesis of claim 27, wherein the pinion is preloaded
against the output gear to define the friction drive and to
transmit the torque between the drive stage and the output
stage.
30. The leg prosthesis of claim 27, wherein a drive cable is
wrapped around the pinion and the at least one drive gear to define
the cable drive and to transmit the torque between the pinion and
the at least one drive gear.
31. The leg prosthesis of claim 27, wherein a drive cable is
wrapped around the output gear and the pinion to define the cable
drive and to transmit the torque between the drive stage and the
output stage.
32. The leg prosthesis of claim 27, wherein a drive cable wrapped
around the output gear and the one of the knee joint and the ankle
joint to define the cable drive and to transmit the torque between
the output gear and the one of the knee joint and the ankle
joint.
33. A powered leg prosthesis, comprising: a powered knee joint
comprising a knee joint and a knee motor unit for delivering power
to the knee joint; a prosthetic lower leg having a socket interface
above the knee joint; a powered ankle joint coupled to the lower
leg opposite the knee joint comprising an ankle joint and an ankle
motor unit to deliver power to the ankle joint; a prosthetic foot
including a ball and a heel; a plurality of sensors for providing a
sagittal plane moment, and ground interaction forces at the ball
and at the heel, and at least one controller coupled to the
plurality of sensors for extracting real-time input from the user
based on data from the plurality of sensors for controlling
movement of the prosthesis, wherein at least one of the knee motor
unit or the ankle motor unit delivers power to a corresponding one
of the knee joint and the ankle joint using at least one of a cable
drive or a friction drive.
34. The leg prosthesis of claim 33, wherein the one of the knee
motor unit and the ankle motor unit comprises a motor having a
drive shaft for generating torque, an output stage comprising the
corresponding one of the knee joint and the ankle joint and an
output gear for providing the power to the corresponding one of the
knee joint and the ankle joint based on the torque, and a drive
stage comprising a pinion and at least one drive gear for
transmitting the torque between the drive shaft and the output
stage.
35. The leg prosthesis of claim 34, wherein the friction drive
comprises at least one roller preloaded against at least one drive
gear for transmitting the torque between the drive shaft and the
output stage.
36. The leg prosthesis of claim 34, wherein the pinion preloaded
against the output gear to define the friction drive and to
transmit the torque between the drive stage and the output
stage.
37. The leg prosthesis of claim 34, wherein a drive cable is
wrapped around the pinion and the at least one drive gear to define
the cable drive and to transmit the torque between the pinion and
the at least one drive gear.
38. The leg prosthesis of claim 34, wherein a drive cable is
wrapped around the output gear and the pinion to define the cable
drive and to transmit the torque between the drive stage and the
output stage.
39. The leg prosthesis of claim 34, wherein a drive cable wrapped
around the output gear and the one of the knee joint and the ankle
joint to define the cable drive and to transmit the torque between
the output gear and the one of the knee joint and the ankle
joint.
40. A powered leg prosthesis, comprising: a powered knee joint
comprising a knee joint and a knee motor unit for delivering power
to the knee joint; a powered ankle joint coupled to the knee joint
comprising an ankle joint and an ankle motor unit to deliver power
to the ankle joint; a prosthetic foot coupled to the ankle joint; a
plurality of sensors for measuring a real-time input; and at least
one controller for controlling movement of the prosthesis based on
the real-time input, wherein at least one of the knee motor unit or
the ankle motor unit is configured to deliver power to a
corresponding one of the knee joint and the ankle joint using at
least one cable drive and at least one friction drive.
41. The leg prosthesis of claim 40, wherein the one of the knee
motor unit and the ankle motor unit comprises a motor having a
drive shaft for generating torque, an output stage defining the
cable drive and comprising the corresponding one of the knee joint
and the ankle joint and an output gear for providing the power to
the corresponding one of the knee joint and the ankle joint based
on the torque, and a drive stage defining the friction drive and
comprising a pinion and at least one drive gear for transmitting
the torque between the drive shaft and the output stage.
42. The leg prosthesis of claim 41, wherein the friction drive
comprises at least one roller preloaded against at least one drive
gear for transmitting the torque between the drive shaft and the
output stage.
43. The leg prosthesis of claim 41, wherein the pinion is preloaded
against the output gear to define the friction drive and to
transmit the torque between the drive stage and the output
stage.
44. The leg prosthesis of claim 41, wherein a drive cable wrapped
around the output gear and the one of the knee joint and the ankle
joint to define the cable drive and to transmit the torque between
the output gear and the one of the knee joint and the ankle joint.
Description
CROSS REFERENCE TO RELATED APPLICATIONS
[0001] This application claims the benefit of Provisional
Application Ser. No. 61/046,684 entitled "POWERED LEG PROSTHESIS
AND CONTROL METHODOLOGIES FOR OBTAINING NEAR NORMAL GAIT", filed
Apr. 21, 2008, which is herein incorporated by reference in its
entirety.
FIELD OF THE INVENTION
[0003] The invention relates to a powered leg prosthesis and
control methodologies for controlling the prosthesis.
BACKGROUND
[0004] Leg prostheses can provide an artificial ankle, and
artificial knee, or both an artificial ankle and an artificial
knee. A transfemoral prosthesis is a prosthesis designed for above
the knee amputees. Transfemoral prostheses are generally more
complicated than transtibial prostheses, as they must include a
knee joint.
[0005] Nearly all current commercial transfemoral comprising
prostheses are energetically passive devices. That is, the joints
of the prostheses either store or dissipate energy, but do not
provide net power over a gait cycle. The inability to deliver joint
power impairs the ability of these prostheses to restore many
locomotive functions, including walking up stairs and up slopes.
Moreover, there is a need for a leg prosthesis that provides a more
natural gait behavior.
SUMMARY
[0006] This Summary is provided to comply with 37 C.F.R.
.sctn.1.73, presenting a summary of the invention to briefly
indicate the nature and substance of the invention. It is submitted
with the understanding that it will not be used to interpret or
limit the scope or meaning of the claims. Embodiments of the
invention provide power leg prostheses and associated methods for
control.
[0007] In a first embodiment of the invention, a powered leg
prosthesis is provided. The prosthesis includes a powered knee
joint comprising a knee joint and a knee motor unit for delivering
power to the knee joint, a prosthetic lower leg having a socket
interface above the knee joint, a powered ankle joint coupled to
the lower leg opposite the knee joint comprising an ankle joint and
an ankle motor unit to deliver power to the ankle joint, and a
prosthetic foot coupled to the ankle joint. The prosthesis also
includes at least one sensor for measuring a real-time input, and
at least one controller for controlling movement of the prosthesis
based on the real-time input.
[0008] In a second embodiment of the invention, a method is
provided for controlling a powered leg prosthesis comprising at
least one of an ankle and a knee joint, the leg prosthesis coupled
to a user of the prosthesis. The method includes representing
behavior of the at least one joint as a plurality of different
activity modes, and within each activity mode a plurality of
different internal phases, and responsive to sensing of at least
one input initiated by the user, switching between the internal
phases and activity modes. In the method, net energy can be
delivered to the at least one joint upon the switching between the
internal phases, and no net energy is delivered to the at least one
joint if the internal phases remain unchanged.
[0009] In a third embodiment of the invention, a powered leg
prosthesis is provided. The prosthesis includes a powered knee
joint comprising a knee joint and a knee motor unit for delivering
power to the knee joint, a prosthetic lower leg having a socket
interface above the knee joint, a powered ankle joint coupled to
the lower leg opposite the knee joint comprising an ankle joint and
an ankle motor unit to deliver power to the ankle joint, and a
prosthetic foot including a ball and a heel. The prosthesis also
includes at least one sensor for providing a sagittal plane moment,
and ground interaction forces at the ball and at the heel, and at
least one controller coupled to the sensor for extracting real-time
input from the user based on data from the sensor for controlling
movement of the prosthesis.
BRIEF DESCRIPTION OF THE DRAWINGS
[0010] FIG. 1A is a view of a powered knee and ankle prosthesis,
according to an embodiment of the invention.
[0011] FIG. 1B is an exploded view of the powered knee and ankle
prosthesis shown in FIG. 1A, according to an embodiment of the
invention.
[0012] FIG. 2 is an exploded view of knee motor unit, according to
an embodiment of the invention.
[0013] FIG. 3 is an exploded view of ankle motor unit, according to
an embodiment of the invention
[0014] FIG. 4 is an exploded view of knee joint, according to an
embodiment of the invention.
[0015] FIG. 5 is an exploded view of ankle joint, according to an
embodiment of the invention.
[0016] FIGS. 6A and B are views of a foot having toe and heel force
sensing elements, according to an embodiment of the invention.
[0017] FIG. 7 shows the joint angle and torque convention used
herein. Positive torque is defined in the direction of increasing
angle.
[0018] FIG. 8 shows the subdivision of normal walking into four
internal phases showing the knee and ankle angles during the
phases, according to an embodiment of the invention.
[0019] FIG. 9 shows a finite-state model of normal walking,
according to an embodiment of the invention. Each box represents a
different internal phase and the transition conditions between the
internal phases are specified.
[0020] FIG. 10 shows piecewise fitting of knee and ankle torques
during normal speed level walk scaled for a 75 kg adult to a
non-linear spring-damper impedance model.
[0021] FIG. 11 is a diagram for an active/passive decomposition
based control of the powered knee and ankle prosthesis, according
to an embodiment of the invention.
[0022] FIG. 12 is a diagram for a general form of active-passive
decomposition control including intent recognition that provides
supervisory modulation, according to an embodiment of the
invention.
[0023] FIG. 13A is a side view of powered knee and ankle
prosthesis, according to another embodiment of the invention.
[0024] FIG. 13B is a front view of powered knee and ankle
prosthesis of FIG. 13A.
[0025] FIGS. 14A and 14B show perspective and bottom views of an
exemplary sagittal moment load cell suitable for use in the various
embodiments of the invention.
[0026] FIG. 15 is a block diagram of an exemplary embedded
microcontroller in accordance with an embodiment of the
invention.
[0027] FIG. 16 is a control state chart for the three activity
modes corresponding to walking, standing, and sitting, and for the
internal phases and their corresponding transitions within each
activity mode.
[0028] FIG. 17 shows knee angle modulated knee stiffness during
pre-stand (solid line) and pre-sit (dashed line) phases.
[0029] FIG. 18 is a plot of axial actuation unit force versus ankle
angle.
[0030] FIG. 19 shows a normal speed walking phase portrait of the
knee joint and four stride segments.
[0031] FIG. 20 shows the selection of indexing data samples during
a first segment of a walking stride.
[0032] FIG. 21 is the output of the decomposition for Segment 1
showing the spring and dashpot constants and the active and passive
knee torques.
[0033] FIG. 22 is a state chart for governing the discrete dynamics
of an active-passive decomposition controller in accordance with an
embodiment of the invention.
[0034] FIG. 23 is a state chart for governing the discrete dynamics
of the cadence estimator in accordance with an embodiment of the
invention.
[0035] FIG. 24 is a schematic diagram of accelerometer measurements
for slope estimation in accordance with an embodiment of the
invention.
[0036] FIG. 25 is a state chart for slope estimation in a
controller in accordance with an embodiment of the invention.
[0037] FIGS. 26A and 26B show front and back views of a
friction/cable drive motor in accordance with an embodiment of the
invention.
DETAILED DESCRIPTION
[0038] The invention is described with reference to the attached
figures, wherein like reference numerals are used throughout the
figures to designate similar or equivalent elements. The figures
are not drawn to scale and they are provided merely to illustrate
the instant invention. Several aspects of the invention are
described below with reference to example applications for
illustration. It should be understood that numerous specific
details, relationships, and methods are set forth to provide a full
understanding of the invention. One having ordinary skill in the
relevant art, however, will readily recognize that the invention
can be practiced without one or more of the specific details or
with other methods. In other instances, well-known structures or
operations are not shown in detail to avoid obscuring the
invention. The invention is not limited by the illustrated ordering
of acts or events, as some acts may occur in different orders
and/or concurrently with other acts or events. Furthermore, not all
illustrated acts or events are required to implement a methodology
in accordance with the invention.
[0039] The present inventors have observed that biomechanically
normal walking requires positive power output at the knee joint and
significant net positive power output at the ankle joint.
Embodiments of the invention provide a prosthesis that delivers
power at both the knee and ankle joints. Unlike prior disclosed leg
prosthetics that generate a desired joint trajectory for the
prosthetic leg based on measurement of the sound leg trajectory and
thus requires instrumentation of the sound leg, embodiments of the
invention do not generally require instrumentation of the sound
leg. Prostheses including transfemoral prostheses according to
embodiments of the invention generally provide power generation
capabilities comparable to an actual limb and a gait-based control
framework for generating the required joint torques for locomotion
while ensuring stable and coordinated interaction with the user and
the environment. Embodiments of the invention thus enable the
restoration of substantially biomechanically normal locomotion.
[0040] One design for a prosthesis according to an embodiment of
the invention is shown in FIG. 1A through FIG. 6B. The prosthesis
100 comprises a prosthetic lower leg 101. Lower leg 101 can be
coupled to a powered knee joint comprising a knee motor unit 105
coupled to a knee joint 110, and a powered ankle joint comprising
an ankle motor 115 coupled to an ankle joint 120. A sagittal plane
moment sensor 125 can be located between the prosthesis and the
user to measure the moment, and in one embodiment is located
immediately below the socket interface. In the embodiment shown,
sensor 125 measures the sagittal plane moment, while separate
sensors described below measure the ball of foot force and heel
force with respect to the ground or other object the foot is
pressed against. A load sensor 135 can be positioned at the ball of
the foot, and a load sensor 140 can be positioned at the heel of
the foot. However, in another embodiment (not shown) sensor 125 can
measure the sagittal plane moment, the frontal plane moment and the
axial force, such as provided by the three-axis socket load cell.
This alternate embodiment can eliminate the need for sensor 135 and
sensor 140.
[0041] Load sensors 141 and 142 are in series with each motor unit
105 and 115, respectively for motor unit force control. Position
sensors 151 and 152 are provided at each joint 110 and 120 as shown
in FIGS. 4 and 5 respectively. Position sensors 151 measure joint
angle (.theta. as used below) and can be embodied as
potentiometers. The computer/process controller, and power source
(e.g. a battery such as a Li ion battery, and electrical
connections in the case of an electrical power source are not shown
to avoid obscuring aspects of the invention. Non-electrical power
sources may also be used, such as pneumatic power, or non-battery
electrical sources, such as hydrogen-based fuel cells.
[0042] Prosthesis 100 is shown in an exploded view in FIG. 1B.
Joints 110 and 120 are more clearly shown as compared to FIG.
1A.
[0043] FIG. 2 is an exploded view of knee motor unit 105, according
to an embodiment of the invention. Load sensor 141 is shown as a
load cell (e.g. strain gauge). Load sensor 141 measures force and
moments. The motor unit 105 comprises a motor-driven ball screw
assembly which drives the knee joint through a slider-crank linkage
comprising screw 145. Other motor drive assemblies may also
generally be used.
[0044] FIG. 3 is an exploded view of ankle motor unit 115,
according to an embodiment of the invention. Load sensor 142 is
generally analogous to load sensor 141. The motor unit 115
comprises a motor-driven ball screw assembly which drives the ankle
joint through a slider-crank linkage comprising screw 145. The
ankle motor 115 includes a spring 147 positioned to provide power
in parallel (thus being additive) with power provided by the motor
unit 115. Spring 147 biases the motor unit's force output toward
ankle plantarflexion, and supplements the power output provided by
motor unit 115 during ankle push off.
[0045] FIG. 4 is an exploded view of knee joint 110, according to
an embodiment of the invention. As described above, knee joint 110
includes position sensor 151 that can be embodied as a
potentiometer for angle measurements of the knee joint 110.
[0046] FIG. 5 is an exploded view of ankle joint 120, according to
an embodiment of the invention. As described above, ankle joint 120
includes position sensor 152 that can be embodied as a
potentiometer for angle measurements of the ankle joint 120.
[0047] FIG. 6A is a view of a foot 170 having ball of foot sensors
135, according to an embodiment of the invention. Sensors 135 are
provided to measure the ground reaction forces near the ball of the
foot, such as when the foot strikes the ground. FIG. 6B is a view
of a foot 170 having ball of foot sensors 135 and heel sensors 140,
according to an embodiment of the invention. Sensors 140 are
provided to measure the ground reaction forces on the heel of the
foot when the foot 170 strikes the ground. Sensors 135 and 140 can
be embodied as strain based sensors.
[0048] As described above, prostheses according to embodiments of
the invention generally provide a gait-based control framework for
generating the required joint torques for locomotion while ensuring
stable and coordinated interaction with the user and the
environment. This enables embodiments of the invention to restore
substantially biomechanically normal locomotion.
[0049] Regarding control of the prosthesis, conventional prosthetic
control schemes utilize position-based control which comprises
generation of a desired joint angle/position trajectories, which by
its nature, must utilize the prosthesis itself as a position source
(e.g. "echo-control" based approaches). Such an approach poses
several problems for the control of a powered prosthesis, such as
prostheses according to embodiments of the invention. First, the
desired position trajectories are typically computed based on
measurement of the sound side (normal) leg trajectory, which 1) is
not well suited to bilateral amputees, 2) requires instrumentation
of the sound side leg, and 3) generally produces an even number of
steps, which can present a problem when the user desires an odd
number of steps. A subtle yet significant issue with conventional
position-based control is that suitable motion tracking requires a
high output impedance (i.e., joints must be stiff), which forces
the amputee to react to the limb rather than interact with or more
generally control the prosthetic limb. Specifically, in order for
the known prosthesis to dictate the joint trajectory, it must
generally assume a high output impedance, thus substantially
precluding dynamic interaction with the user and the
environment.
[0050] Unlike existing passive prostheses, the introduction of
power into a prosthesis according to embodiments of the invention
provides the ability for the device to also act, rather than simply
react. As such, the development of a suitable controller and
control methodology that provides for stable and reliable
interaction between the user and prosthesis is provided herein.
Control according to embodiments of the invention has been found to
enable the user to interact with the prosthesis by leveraging its
dynamics in a manner similar to normal gait, and also generates
more stable and more predictable behavior.
[0051] Thus, rather than gather user intent from the joint angle
measurements from the contralateral unaffected leg, embodiments of
the invention infer commands from the user via the (ipsilateral)
forces and moments of interaction between the user and prosthesis.
Specifically, the user interacts with the prosthesis by imparting
forces and moments from the residual limb to the prosthesis, all of
which can be measured via suitable sensor(s), such as sensors 125,
140 and 141 described above which measures moments/forces. These
forces and moments serve not only as a means of physical
interaction, but also serve as an implicit communication channel
between the user and device, with the user's intent encoded in the
measurements. Inferring the user's intent from the measured forces
and moments of interaction according to embodiments of the
invention provides several advantages relative to the known echo
approach.
[0052] In one embodiment of the invention the torque required at
each joint during a single stride (i.e. a single period of gait)
can be piecewise represented by a series of passive impedance
functions. A regression analysis of gait data indicates that joint
torques can be characterized by functions of joint angle (.theta.)
and angular velocity by an impedance model, such as the following
exemplary passive impedance function shown in equation 1 below:
.tau.=k.sub.1(.theta.-.theta..sub.c)+b*{dot over (.theta.)}
where k.sub.1, b, and the equilibrium joint angle .theta..sub.c are
all constants that are generally generated empirically, and are
constants for a given joint during a given internal phase (e.g.
knee, internal phase 3). k.sub.1 characterizes the linear
stiffness. b is the linear damping coefficient, .theta. is the
measured joint angle which can characterize the state of the
prosthesis, .theta..sub.c is the equilibrium angle, {dot over
(.theta.)} is the angular velocity of the joint, and .tau. is the
joint torque. Given these constants, together with instantaneous
sensor measurements for .theta. and {dot over (.theta.)}, the
torque (.tau.) at the joints (knee and ankle) can be
determined.
[0053] Positive directions of the angle (.theta.) and torque
(.tau.) as used herein are defined as shown in FIG. 7. If the
coefficients b and k.sub.1 are constrained to be positive, then the
joints will each exponentially converge to a stable equilibrium at
.theta.=.theta..sub.c and {dot over (.theta.)}=0 within each
internal phase. That is, within any given internal phase, the
actuators are energetically passive (i.e. the joint will come to
rest at a local equilibrium). While the unactuated prosthesis can
be energetically passive, the behavior of one joint (knee or ankle)
or the combined behavior of the knee and ankle joints, can be
likewise passive, and thus will generally respond in a predictable
manner.
[0054] Responsive to direct input from the user (e.g. a heel
strike) to trigger a change in internal phase, power (torque) can
be delivered from the power source (e.g. battery) to the prosthesis
in the proper magnitude to provide the desired movement. Since the
switching can be triggered by direct input from the user related to
the current internal phase, the user maintains direct influence
over the power applied to the prosthesis. If the user does not
trigger the next internal phase (i.e. remains stationary) no net
energy is delivered. That is, the prosthesis will generally cease
to receive power from the power source for moving the joint, and
will instead, due to the damped response, soon come to rest at the
local equilibrium identified with the present internal phase.
[0055] As described above, the decomposition of joint behavior into
passive segments requires the division of the gait cycle into a
plurality of internal phases or "finite states" characterized by an
impedance function and a set of constants for the impedance
function, as dictated by their functions and the character of the
piecewise segments of the impedance functions described above. The
switching rules between internal phases should generally be well
defined and measurable, and the number of phases should be
sufficient to provide a substantially accurate representation of
normal joint function. In one embodiment of the invention, the
swing and stance phase of gait can constitute a minimal set of
internal phases.
[0056] Based on least-squares regression fitting of Equation 1 to
empirical gait data, the present Inventors determined that such
fits were improved significantly by further dividing the two modes
of swing and stance each into two additional internal phases to
realize four phases, as shown in FIG. 8. A fifth internal phase can
also be added, as illustrated in FIG. 16. The angle (.theta.) of
the prosthetic knee (above) and ankle joint (below) can be provided
during each internal phase as a function of the % of the stride.
Angle values shown can be used as threshold values to trigger phase
changes as described below relative to FIG. 9. As clear to one
having ordinary skill in the art, the number of phases can be other
than two or four.
[0057] FIG. 9 shows exemplary switching rules between internal
phases for walking. FIG. 16 shows another set exemplary switching
rules, for walking, standing, and sitting activity modes. As
described above, if the user does not initiate actions that trigger
the next phase (e.g. based on the switching rules), the prosthesis
will cease to receive power and will come to rest at the local
equilibrium identified with the present phase. For example,
switching can be based on the ankle angle>a threshold value
(mode 1 to mode 2), or ankle torque<threshold) (mode 2 to mode
3), the angle or torque measurements provided by on board sensors
as described above.
[0058] Phase 1 shown in FIG. 8 begins with a heel strike by the
user (which can be sensed by the heel force sensor), upon which the
knee immediately begins to flex so as to provide impact absorption
and begin loading, while the ankle simultaneously plantarflexes to
reach a flat foot state. Both knee and ankle joints have relatively
high stiffness (and can be accounted for by k1 in equation 1)
during this phase to prevent buckling and allow for appropriate
stance knee flexion, because phase 1 comprises most of the weight
bearing functionality. Phase 2 is the push-off phase and begins as
the ankle dorsiflexes beyond a given angle (i.e. user's center of
mass lies forward of stance foot). The knee stiffness decreases in
this mode to allow knee flexion while the ankle provides a
plantarflexive torque for push-off. Phase 3 begins as the foot
leaves the ground as detected by the ankle torque load cell and
lasts until the knee reaches maximum flexion. Mode 4 is active
during the extension of the knee joint (i.e. as the lower leg
swings forward), which begins as the knee velocity becomes negative
and ends at heel strike (e.g. as determined by the heel force
sensor).
[0059] In both of the swing phases (Phases 3 and 4), the ankle
torque can be small and can be represented in the controller as a
(relatively) weak spring regulated to a neutral position. The knee
can be primarily treated as a damper in both swing phases.
[0060] Impedance modeling of joint torques was preliminarily
validated by utilizing the gait data from a healthy 75 kg subject,
as derived from body-mass normalized data. Incorporating the four
internal phases described above, along with the motion and torque
data for each joint, a constrained least-squares optimization was
conducted to generate a set of parameters k.sub.1, b and
.theta..sub.c for each phase for each joint for use in Equation 1.
The resulting parameter set can be fit to joint torques and is
shown graphically in FIG. 10. FIG. 10 shows piecewise fitting of
knee and ankle torques during normal speed level walk scaled for a
75 kg adult to a non-linear spring-damper impedance model. The
numbers shown in each phase represent the mean ratio of the
stiffness forces to damping forces predicted by the fit. The
vertical lines represent the segmentation of a gait stride into
four distinct phases. The fit shown in FIG. 10 clearly indicates
that normal joint function can be represented by the use of
piecewise passive functions.
[0061] Controllers according to embodiments of the invention
generally comprise an underlying gait controller (intra-modal
controller). An optional supervisory gait controller (also called
intent recognizer) can also be provided. Both controllers generally
utilize measured information. This information generally comprises
user and ground interaction forces (F) and moments/torques (.tau.),
joint angles and angular velocities from on-board sensors, and can
be used to extract real-time input from the user. The gait control
component utilizes the sensed instantaneous nature of the user
input (i.e., moments and forces) to control the behavior of the leg
within a given activity mode, such as standing, walking, or stair
climbing.
[0062] Two exemplary approaches to intra-modal impedance generation
are described below. The first approach is shown in FIG. 11 and
represents a general form of active-passive decomposition-based
intra-mode control. The second embodiment shown in FIG. 12 includes
the control structure shown in FIG. 11 but adds a supervisory
intent recognizing controller to modulate the intra-modal control
based on inputs from an intent recognition module. As shown in
FIGS. 11 and 12, F.sub.s is the force the user of the prosthesis is
applying, such as a heel force in the case of a heel strike, .tau.
represents joint torque, and .theta. represent joint angles.
.tau..sub.a represents the active component of joint torque which
is roughly proportional to the input force, and .tau..sub.p
represents the passive component of torque. The active joint torque
.tau..sub.a is thus the total joint torque .tau. minus the passive
joint torque, .tau.p. Derivatives are shown using the dot
convention, with one dot being the first derivative (e.g., {dot
over (.theta.)} being angular velocity) and two dots representing
the second derivative.
Intra-Modal Active-Passive Decomposition Control
[0063] In this embodiment of the intra-modal controller, shown in
FIG. 11, the behavior of the prosthesis can be decomposed into a
passive component and an active control component. The active
control component is an algebraic function of the user's real-time
input F.sub.s (i.e., sensed socket-prosthesis interface forces and
moments and sensed ground reaction forces). The controller output
is shown as the active torque (TO minus the passive torque
.tau..sub.p. The controller output .tau..sub.a-.tau..sub.p applied
to the prosthetic leg based on dynamics of the leg responds via
.theta. and {dot over (.theta.)}. The system response, .theta. and
{dot over (.theta.)}, is fed back to the controller.
[0064] Power applied to the prosthesis can be thus commanded
directly by the user through measured interface forces and moments
initiated by user movements. In the absence of these commands from
the user, F.sub.s=0, .tau..sub.a=0 and the prosthesis fundamentally
(by virtue of the control structure) cannot generate power, and
thus only exhibits controlled passive behavior. Due to the
decomposition of energetic behaviors inherent in this control
structure, the prosthesis under it's own control can be generally
stable and passive. Unlike known echo control approaches, the input
can be real-time, based only on the affected leg, and thus the
approach can be equally applicable to bilateral and unilateral
amputees and can reflect the instantaneous intent of the user.
Additionally, unlike echo control that is based on servocontrol,
the prosthesis will exhibit a natural impedance to the user that
should feel more like a natural limb. These combined features
should result in an active prosthesis that will feel inasmuch as
possible like a natural extension of the user. The structure and
properties of both the gait controller and intent recognizer are
described below.
[0065] As described above, since gait is largely a periodic
activity, joint behavior can be functionally decomposed over a
period by decomposing the joint torque into a passive component and
an active component. The passive component can comprise a function
of angle (i.e., single-valued and odd), and a function of angular
velocity passive (i.e., single-valued and odd), such as equation 1
described above. The active component can be a function of the user
input (i.e., socket interface forces). Given a set of data that
characterizes a nominal period of joint behavior, the passive
component can be first extracted from the whole, since the passive
behavior is a subset of the whole (i.e., the passive component
consists of single-valued and odd functions, while the active has
no restrictions in form). The passive component can be extracted by
utilizing a least squares minimization to fit a generalized
singled-valued odd function of angle and'angular velocity to the
torque. Once the passive component is extracted, the residual
torque (i.e., the portion that is not extracted as a passive
component), can be constructed as an algebraic function of the
sensed socket interface and ground reaction forces (i.e., the
direct-acting user input) by incorporating a similar candidate
function, but not restricted to be of passive form. Finally,
superimposing the passive and active components provides a
decomposed functional approximation of the original period joint
torque.
Intra-Modal Locally Passive Event-Triggered Control
[0066] In this embodiment of the intra-modal controller shown in
FIG. 12, a supervisory intent recognizer can be added that utilizes
the same sensed user inputs (i.e., moments and forces) as the
intra-modal/gait controller, but extracts the user's intent based
on the characteristic shape of the user input(s) and system
response (e.g. F, .theta., .theta.-dot). Based on the extracted
intent, the supervisory intent recognizer modulates the behavior of
the underlying gait controller to smoothly transition behavior
within a gait (e.g., speed and slope accommodation) and between
gaits (e.g., level walk to stair ascent), thus offering a unified
control structure within and across all gaits.
[0067] Gait intent recognition can be a real time pattern
recognition or signal classification problem. The signal in this
case is generally the combination of socket interface forces Fs and
the dynamic state of the prosthesis, which in one embodiment can be
a vector of the knee and ankle angles .theta. for a powered leg
prosthesis according to an embodiment of the invention. A variety
of methods exist for pattern recognition and signal classification
including nearest neighbor algorithms, neural networks, fuzzy
classifiers, linear discriminant analysis, and genetic
algorithms.
Sensors
[0068] As described above embodiments of the invention include a
number of sensors for providing signals for adjusting operation of
a leg and ankle prosthesis. A description of one exemplary
arrangement of sensors can be described below with respect to FIGS.
13A, 13B, 14A, and 14B. FIG. 13A is a side view of powered knee and
ankle prosthesis 1300, according to another embodiment of the
invention. FIG. 13B is a front view of powered knee and ankle
prosthesis of FIG. 13A. FIGS. 14A and 14B show perspective and
bottom views of an exemplary sagittal moment load cell suitable for
use in the various embodiments of the invention.
[0069] Each joint actuation unit, such as knee actuation unit 1302
and ankle actuation unit 1304 in FIG. 13A, can include a uniaxial
load cell positioned in series with the actuation unit for closed
loop force control. Both the knee and ankle joints can incorporate
integrated potentiometers for joint angle position. The ankle
actuation unit can include a spring 1305, as described above with
respect to FIGS. 1A-4. One 3-axis accelerometer can be located on
the embedded system 1306 and a second one can located below the
ankle joint 1308 on the ankle pivot member 1310. A strain based
sagittal plane moment sensor 1312, such as sensor 1400 shown in
FIGS. 14A and 14B, can located between the knee joint 1314 and the
socket connector 1316, which measures the moment between a socket
and the prosthesis. In the various embodiments of the invention, a
sagittal plane moment sensor can be designed to have a low profile
in order to accommodate longer residual limbs. The sensor can
incorporate a full bridge of semiconductor strain gages which
measure the strains generated by the sagittal plane moment. In one
embodiment of the invention, the sagittal plane moment sensor was
calibrated for a measurement range of 100 Nm. A custom foot 1318
can designed to measure the ground reaction force components at the
ball 1320 of the foot and heel 1322. The foot can include of heel
and ball of foot beams, rigidly attached to a central fixture and
arranged as cantilever beams with an arch that allows for the load
to be localized at the heel and ball of the foot, respectively.
Each heel and ball of foot beam can also incorporates a full bridge
of semiconductor strain gages that measure the strains resulting
from the respective ground contact forces. In one embodiment of the
invention, the heel and ball of foot load sensors were calibrated
for a measurement range of 1000 N. In addition, incorporating the
ground reaction load cell into the structure of a custom foot can
eliminate the added weight of a separate load cell, and also
enables separate measurement of the heel and ball of foot load. The
prosthetic foot can be designed to be housed in a soft prosthetic
foot shell (not shown).
Microcontroller System
[0070] The powered prosthesis contains an embedded microcontroller
that allows for either tethered or untethered operation. An
exemplary embedded microcontroller system 1500 is shown in the
block diagram in FIG. 15. The embedded system 1500 consists of
signal processing, power supply, power electronics, communications
and computation modules. The system can be powered by a lithium
polymer battery with 29.6 V. The signal electronics require +/-12 V
and +3.3 V, which are provided via linear regulators to maintain
low noise levels. For efficiency, the battery voltage can be
reduced by PWM switching amplifiers to +/-15 V and +5 V prior to
using the linear regulators. The power can be disconnected via a
microcontroller that controls a solid state relay. The power status
can be indicated by LED status indicators controlled also by the
microcontroller.
[0071] The analog sensor signals acquired by the embedded system
include the prosthesis sensors signals (five strain gage signals
and two potentiometer signals), analog reference signals from the
laptop computer used for tethered operation, and signals measured
on the board including battery current and voltage, knee and ankle
servo amplifier currents and two 3-axis accelerometers. The
prosthesis sensor signals are conditioned using input
instrumentation amplifiers. The battery, knee motor and ankle motor
currents are measured by current sense resistors and current
sensing amplifiers. The signals are filtered with a first-order RC
filter and buffered with high slew rate operational amplifiers
before the analog to digital conversion stage. Analog to digital
conversion can be accomplished by two 8-channel analog to digital
converters. The analog to digital conversion data can be
transferred to the microcontroller via serial peripheral interface
(SPI) bus.
[0072] The main computational element of the embedded system can be
a 32-bit microcontroller. In the untethered operation state, the
microcontroller performs the servo and activity controllers of the
prosthesis and data logging at each sample time. In addition to
untethered operation, the prosthesis can also be controlled via a
tether by a laptop computer running MATLAB Simulink RealTime
Workshop. In the tethered operation state, the microcontroller
drives the servo amplifiers based on analog reference signals from
the laptop computer. A memory card can be used for logging
time-stamped data acquired from the sensors and recording internal
controller information. The memory chip can be interfaced to the
computer via wireless USB protocol. The microcontroller sends PWM
reference signals to two four quadrant brushless DC motor drivers
with regenerative capabilities in the second and forth quadrants of
the velocity/torque curve.
Control of Sitting and Standing
[0073] In some embodiments of the invention, additional controls
can be provided for operating the prosthesis when going from a
sitting to a standing position or vice versa. This can be
implemented via the use of a sitting mode controller implemented in
the microcontroller. Operation of the sitting mode controller
consists of four phases that are outlined in the general control
state chart shown in FIG. 16. As shown in FIG. 16, two phases are
primary sitting phases, weight bearing and non-weight bearing. The
other two phases encompass the transition phases, pre-stand and
pre-sit, for standing up and sitting down, respectively. Weight
bearing and non-weight bearing are the primary sitting phases that
switch the knee and ankle joints between high and low impedances,
respectively. The transition phases, pre-stand and pre-sit,
modulate the stiffness of the knee as a function of knee angle, as
shown in FIG. 17, to assist the user in standing up and sitting
down. FIG. 17 shows knee angle modulated knee stiffness during
pre-stand (solid line) and pre-sit (dashed line) phases.
[0074] The modulation allows for smoother transitions near the
seated position. The ankle joint can be slightly dorsiflexed with
moderate stiffness during the standing up and sitting down phases.
Switching between the four sitting phases occurs when sensor
thresholds are exceeded, as depicted FIG. 16. The parameters of the
impedance based controllers are tuned using a combination of
feedback from the user and joint angle, torque and power data from
the prosthesis.
Mechanical Design
[0075] In the various embodiments of the invention, actuation for
the prosthesis can be provided by two motor-driven ball screw
assemblies that drive the knee and ankle joints, respectively,
through a slider-crank linkage. The prosthesis can be capable of
120.degree. of flexion at the knee and 45.degree. of planterflexion
and 20.degree. of dorsiflexion at the ankle. In one embodiment,
each actuation unit consists of a DC motor (such as a Maxon EC30
Powermax) connected to a 12 mm diameter ball screw with 2 mm pitch,
via helical shaft couplings. An exemplary ankle actuation unit
additionally incorporates a 302 stainless steel spring (51 mm free
length and 35 mm outer diameter), with 3 active coils and a
stiffness of 385 N/cm in parallel with the ball screw.
[0076] As described above with respect to FIGS. 1A-4, the purpose
of the spring can be to bias the motor's axial force output toward
ankle plantarflexion, and to supplement power output during ankle
push off. The stiffness of the spring can be maximized to allow for
peak force output without limiting the range of motion at the
ankle. The resulting axial actuation unit's force versus ankle
angle plot can be shown in FIG. 18. FIG. 18 is a plot if axial
force as a function of ankle angle illustrating spring force,
actuator force and total force. FIG. 18 graphically demonstrates
for fast walking the reduction in linear force output supplied by
the motor at the ankle through the addition of the spring. Note
that the compression spring does not engage until approximately
five degrees of ankle plantarflexion. Each actuation unit can
include a uniaxial load cell (such as Measurement Specialties
ELPF-500L), positioned in series with the actuation unit for closed
loop force control of the motor/ballscrew unit. Both the knee and
ankle joints can incorporate bronze bearings and, for joint angle
measurement, integrated precision potentiometers (such as an ALPS
RDC503013). A strain based sagittal plane moment sensor, as
previously described with respect to FIGS. 14A and 14B can be
located between the knee joint and the socket connector, which
measures the moment between the socket and prosthesis. The ankle
joint connects to a foot, which incorporates strain gages to
measure the ground reaction forces on the ball of the foot and on
the heel. The central hollow structure houses a lithium-polymer
battery and provides an attachment point for the embedded system
hardware. To better fit with an anthropomorphic envelope, the ankle
joint can be placed slightly anterior to the centerline of the
central structure. This gives the prosthesis the illusion of
flexion when the amputee can be standing vertically with the knee
fully extended.
[0077] The length of the shank segment can be varied by changing
the length of three components; the lower shank extension, the
spring pull-down, and the coupler between the ball nut and ankle.
Additional adjustability can be provided by the pyramid connector
that can be integrated into the sagittal moment load cell for
coupling the prosthesis to the socket (as is standard in commercial
transfemoral prostheses). In one embodiment of the invention, the
self-contained transfemoral prosthesis was fabricated from 7075
aluminum and has a total mass of 4.2 kg, which can be within an
acceptable range for transfemoral prostheses, and comparable to a
normal limb segment. A weight breakdown of an exemplary device is
presented below in Table 1.
TABLE-US-00001 TABLE II MASS BREAKDOWN OF SELF-CONTAINED POWERED
PROSTHESIS. Component Mass (kg) Battery 0.62 Electronics 0.36 Knee
Motor Assembly 0.72 Ankle Motor Assembly 0.89 Sensorized Foot 0.35
Foot Shell 0.24 Sagittal Moment Sensor 0.12 Remaining Structure
0.90 Total Weight 4.20
Active Passive Torque Decomposition
[0078] Passive joint torque, .tau..sub.p, can be defined as the
part of the joint torque, .tau., which can be represented using
spring and dashpot constitutional relationships (passive impedance
behavior). The system can only store or dissipate energy due to
this component. The active part can be interpreted as the part
which supplies energy to the system and the active joint torque can
be defined as .tau..sub.a=.tau.-.tau..sub.p. This active part can
be represented as an algebraic function of the user input via the
mechanical sensory interface (i.e socket interface forces and
torques).
[0079] Gait is considered a mainly periodic phenomena with the
periods corresponding to the strides. Hence, the decomposition of a
stride will give the required active and passive torque mappings
for a specific activity mode. In general, the joint behavior
exhibits varying active and passive behavior in each stride.
Therefore, segmenting of the stride in several parts can be
necessary. In this case, decomposition of the torque over the
entire stride period requires the decomposition of the different
segments and piecewise reconstruction of the entire segment period.
In order to maintain passive behavior, however, the segments cannot
be divided arbitrarily, but rather can only be segmented when the
stored energy in the passive elastic element is zero. This requires
that the phase space can only be segmented when the joint angle
begins and ends at the same value. FIG. 19 shows the phase portrait
of normal speed walking and the four different stride segments,
S.sub.1, S.sub.2, S.sub.3 and S.sub.4. Thus, the entire
decomposition process consists of first appropriate segmentation of
the joint behavior, followed by the decomposition of each segment
into its fundamental passive and active components.
[0080] The decomposition of each segment shown in FIG. 19 can be
converted to an optimization problem. In each segment of the
stride, 2n data points are selected by sampling the angular
position in equal intervals between its minimum and maximum and
selecting the corresponding positive and negative angular
velocities. In this work, the number of angular position samples
for each segment, n can be set to be 100. The constrained least
squares optimization problem given in Equation 2 below can be
constructed and solved.
min x 1 2 Cx - d 2 2 s . t . 0 .ltoreq. x ( 2 ) ##EQU00001##
where C, x and d are defined in Equations 3, 4, and 5 below,
respectively. The indexing of the joint angular position, angular
velocity and moment samples are explained via the sketch in FIG.
20. FIG. 20 shows a selection and indexing of data samples from a
first segment.
C 4 n .times. 3 n = [ C 1 C 2 C 3 ] T ( 3 ) C 1 = [ diag ( [
.theta. 1 .theta. 2 .theta. n ] n .times. 1 - .alpha. ) diag ( [
.theta. n .theta. n - 1 .theta. 1 ] n .times. 1 - .alpha. ) diag (
[ .theta. . 1 .theta. . 2 .theta. . n ] 2 n .times. 1 ) ] 2 n
.times. 3 n C 2 = [ C 21 C 22 C 23 ] 2 n - 1 .times. 3 n C 21 = [
.theta. 1 - .theta. 2 0 0 0 .theta. n - 1 .theta. n 0 0 0 0 0 ] n
.times. n C 22 = [ .theta. n - .theta. n - 1 0 0 0 .theta. 3 -
.theta. 2 0 0 0 .theta. 2 - .theta. 1 ] n - 1 .times. n C 23 = [
.theta. . 1 - .theta. . 2 0 0 0 .theta. . 2 n - 2 - .theta. . 2 n -
1 0 0 0 .theta. . 2 n - 1 - .theta. . 2 n ] 2 n - 1 .times. 2 n C 3
= [ .beta. .beta. .beta. .beta. ] 1 .times. 3 n x 3 n .times. 1 = [
k 1 k 2 k n - 1 k n b 1 b 2 b 2 n - 1 b 2 n ] ( 4 ) d 4 n .times. 1
= [ .tau. 1 .tau. 2 .tau. 2 n - 1 .tau. 2 n .tau. 1 - .tau. 2 .tau.
2 - .tau. 3 .tau. 2 n - 1 - .tau. 2 n 0 ] ( 5 ) ##EQU00002##
[0081] The matrix C consists of three sub-matrices, C.sub.1,
C.sub.2 and C.sub.3. C.sub.1 can be the main part responsible for
the fitting of the spring and dashpot constants, k and b. C.sub.2
bounds the rate of change of the passive joint torque and ensures
smoothness in the resulting passive joint torque, and C.sub.3 is
basically a row of penalty constants, .beta., which penalizes large
values of the spring and dashpot constants and thus limits the
magnitudes of both. In this work, .beta. is set to 0.1.
[0082] The origin of each virtual spring can be also added to the
optimization problem formulation as a parameter in order to obtain
a tighter passive torque fit. Therefore, the optimization problem
given by (3) can be solved iteratively for a range of values of
spring origin constant, .alpha.. The solution with the least error
norm can be selected as the optimal solution.
[0083] The result of the above stated constrained optimization
problem for segment 1 can be shown in plots (a), (b), and (c) in
FIG. 21. FIG. 21 is the output of the decomposition for s.sub.1 in
FIG. 19 showing the spring and dashpot constants and the active and
passive knee torques (Spring origin, .alpha. is 23 degrees).
[0084] As can be seen from FIG. 21, the decomposed passive part can
be very similar to the joint torque, and thus it can be stated that
the behavior of the joint can be mainly passive. The result of the
decomposition for the segments, can be stored in R.sub.i of the
form given in Equation 6.
R.sub.1=[.theta. {dot over (.theta.)} .tau..sub.pas F.sub.S1
F.sub.S2 .tau..sub.aci].sub.2n.times.6 (6)
where .tau..sub.pas=C.sub.1x.
[0085] The procedure presented above decomposes the joint torques
into active and passive parts. The joint torque references for the
control of the prosthesis are generated by combining this active
and passive torques. There are two major challenges to be solved.
Firstly, the correct motion segment must be selected. Secondly,
after the motion segment is selected at each sampling instant a new
joint torque reference can be generated using the discrete mappings
for the active and passive torque parts.
[0086] A switching system modeling approach incorporating both
discrete and continuous states can be used for the reconstruction
of the torque reference signal. The state chart shown in FIG. 22.
will govern the discrete dynamics of the controller. Since the
sequence of the segments can be ordered (i.e., the direction of the
motion for a specific gait phase does not change), each segment can
transition only to the next one, where the transition guard
function can be written as a inequality in terms of .theta. and
{dot over (.theta.)}. The transitions between segments take no time
and the dynamics of the controller are governed by the
{.intg..sub.pi(.theta.,{dot over (.theta.)});
.intg..sub.ai(F.sub.S)} pair at each sampling instant. The joint
reference torque is
.tau..sub.ref=.tau..sub.a.tau..sub.p=.intg..sub.p.sub.i(.theta.,{dot
over (.theta.)}+.intg..sub.a.sub.i(F.sub.S) (7)
[0087] The decomposition algorithm presented above gives the result
matrix, R, for each segment. The discrete data in R can be used to
construct the joint torque reference for the continuous
measurements of another trial in the same gait phase. At each
sampling instant of the algorithm, the measurement vector
m=[.theta..sub.m, {dot over (.theta.)}.sub.m,
F.sub.S1.sub.--.sub.m, F.sub.S2.sub.--.sub.m].sup.T can be
acquired. For the reconstruction of the passive knee torque part,
the Euclidian error norm between the [.theta..sub.m {dot over
(.theta.)}.sub.m].sup.T and the angular position and velocities of
all the samples in that segment [.theta..sub.i {dot over
(.theta.)}.sub.i].sup.T can be calculated as shown in Equation 8
and stored in the vector e.
e.sub.i= {square root over
((.theta..sub.m-.theta..sub.i).sup.2+({dot over (.theta.)}{square
root over ({dot over (.theta.)}.sub.m-{dot over
(.theta.)}.sub.i).sup.2)} (8)
Then two elements of this vector with the least error norm are
found and the passive knee torque reference can be found as a
weighted linear combination of the passive knee torques
corresponding to these points. The reconstruction of the active
knee torque part is similar where only {.theta., {dot over
(.theta.)}, .tau..sub.pas} is exchanged with {F.sub.S1, F.sub.S2,
.tau..sub.aci}.
Intent Recognition
[0088] The supervisory controller (intent recognizer) switches
among different underlying intramodal controllers depending on the
activity mode the user imposes on the prosthesis. The intent
recognizer consists of three parts: activity mode recognizer,
cadence estimator and the slope estimator.
[0089] The activity mode recognizer detects the activity mode of
the prosthesis (standing, walking, sitting, stair ascent or stair
descent, etc. . . . ). This can be accomplished by comparing the
features which are generated in real time to a feature database
using some machine learning and/or pattern recognition methods. The
present implementation of the gait mode recognizer, which
recognizes standing and walking modes, is described below.
[0090] Firstly, a database which contains all the possible activity
modes (standing and walking in this case) can be generated by
making experimental trials. In the experimental trials, the user
can be asked to walk or stand in different controller modes for 50
second long trials. The socket sagittal moment above the knee
joint, foot heel load, foot ball load, knee angle, knee velocity,
ankle angle and ankle velocity are recorded with 1 ms sampling
period. It should be noted that other sensor signals such as
accelerations and electromyography measurements from the residual
limb can be added to the list of the signals used for intent
recognition. For example, from the recorded experimental trials,
10000 random frames (5000 standing and 5000 walking) of 100 samples
length are generated for all the seven recorded signals. The mean
and the standard deviation of each frame are computed. The mean and
standard deviation of signals are selected as the features since
minimal computation can be required to obtain them. A database
containing 10000 samples with 14 features (mean and standard
deviation of the seven signals) belonging to two classes (standing
and walking) can be generated. After the database is generated, the
dimension of the database can be reduced from 14 to three using
principal component analysis (PCA). Dimension reduction can be
necessary because pattern recognition for high dimensional datasets
can be computationally intensive for real-time applications. After
dimension reduction step, the standing and walking data can be
modeled with Gaussian mixture models. Gaussian mixture models
represent a probability distribution as a sum of several normal
Gaussian distributions. The order of the Gaussian mixture model for
each mode can be determined according to the Minimum Description
Length Criteria.
[0091] As described above, the database generation, dimension
reduction and the Gaussian mixture modeling are explained. For
real-time decision making, overlapping frames of 100 samples can be
generated at each 10 ms interval. 14 features described above are
extracted from these frames and the PCA dimension reduction can be
applied to these features to get a reduced three dimensional
feature vector. The reduced dimension features can be fed to the
Gaussian mixture models for standing and walking and the
probability of the sample vector being standing or walking can be
computed. The mode with the greater probability is selected as the
instantaneous activity mode. Since one decision might give wrong
results in some cases due to noise, disturbance, etc. . . . , a
voting scheme can be used to enhance the results. In the voting
scheme, the controller activity mode is switched if and only if
more than 90 percent of the instantaneous activity mode decisions
among the last 40 decisions are a specific activity mode. Once a
new activity mode is selected by the voting scheme, the underlying
activity controller can be switched to the corresponding mode.
[0092] Such an activity mode recognizer is provided by way of
illustration and not as a limitation. In the various embodiments of
the invention, one or more parts of the algorithm might be
modified. For example, in some embodiments, different features such
as mean, max, kurtosis, median, AR coefficients, wavelet based
features, frequency spectrum based features of the frame might be
generated. Additionally, different dimension reduction techniques
such as linear discriminant analysis, independent component
analysis might be employed. Furthermore, different classification
methods such as artificial neural networks, support vector
machines, decision trees, hidden Markov models might be used.
Cadence and Slope Estimation
[0093] Cadence estimation is accomplished by observing peak
amplitudes in characteristic signal data and then measuring the
time between successive peaks. Since walking is a cyclic activity
each of the sensor signals will be periodic of cadence. The most
relevant sensor signals will contain only one characteristic
amplitude peak per stride such as foot heel load and the ball of
foot load. In the real-time implementations, cadence estimation is
accomplished by recording the foot load after heel strike when it
exceeds 400 N until the load decreases below 350 N. Then, the time
of occurrence of the peak load in this window is found and the
previous peak time is subtracted from the new peak time. This
corresponds to stride time and can be converted to cadence
(steps/min) by multiplying with 120. Once the cadence is estimated,
the intent recognizer selects the corresponding middle layer
controller based on some predefined thresholds as in FIG. 23.
[0094] For example, in some embodiments, a 3D accelerometer capable
of measuring .+-.3 g accelerations is embedded into the ankle joint
coupler where the prosthetic foot is connected. An exemplary
arrangement of such a system is shown by the schematic in FIG. 24.
The accelerometer measurements are used to estimate the ground
slope. In order to estimate the ground slope, the accelerometer
data in tangential direction is used. Assuming the foot is flat on
the ground, the ground slope angle, .theta..sub.s, can be
calculated as in equation (9) below.
.theta. s = sin - 1 ( a t g ) ( 9 ) ##EQU00003##
In Eqn. 9, g is the gravitational constant. In order to find the
ground slope estimate, {circumflex over (.theta.)}.sub.s, the
accelerometer data should be collected while the foot is flat on
the ground as determined by the heel and ball of the foot load
sensors. While the foot is flat on the ground, equation (1) is
computed for the frame of the collected data and the mean of this
frame is outputted as the ground slope estimate, {circumflex over
(.theta.)}.sub.s. Once the slope is estimated, the intent
recognizer selects the corresponding middle layer controller based
on some predefined thresholds. An exemplary state chart for such an
intent recognizer is shown in FIG. 25.
Friction and Cable Drive Based Actuation
[0095] Rather than a ballscrew and slider crank embodiment for the
transmission of torque from a motor to the ankle and/or knee units,
in some embodiments of the invention, the prosthesis can
incorporate a friction and cable drive transmission embodiment.
FIGS. 26A and 26B show front and back views of an exemplary
embodiment of a friction drive transmission 2600 in accordance with
an embodiment of the invention. As shown in FIGS. 26A and 26B, the
shaft 2602 of an electric motor 2604 is preloaded against a first
stage in a housing 2606, such as a larger diameter cylinder or
friction drive gear 2608, which creates sufficient friction to
transmit torque without slip. The shaft 2602 can use one or more
friction rollers 2610 to transmit the torque. The first stage of
the friction drive can also be supplemented with a second stage.
The friction drive gear 2608 drives a smooth pinion 2612 directly,
which is preloaded against a larger diameter cylinder or cable gear
output 2614 in the housing 2606, which in turn transmits torque
directly to the knee or ankle joint.
[0096] In addition to, or rather than a friction drive, the first
or second stage of the transmission can alternatively be embodied
by a cable drive transmission, in which a cable is wrapped around
the circumference of a larger diameter cylinder, such as friction
drive gear 2608, and also around the circumference of a smaller
diameter cylinder, such as pinion 2612. In such embodiments, the
cable is affixed to the friction drive gear 2608, and is
pretensioned, using a tensioning screw 2616 or similar means,
around both the drive gear 2608 and pinion 2612, such that friction
between cable and pinion 2612 enables the transmission of torque
from between the pinion 2612 and drive gear 2608. In one embodiment
of a combined friction drive/cable drive transmission can be used,
in which a first stage of the transmission (i.e., the friction
drive gear 2608 connected directly to the electric motor 2604) is
of the friction drive type, while the second stage of the
transmission (i.e., the cable gear output 2614 connected directly
to the knee or ankle joint) is of the cable drive type.
[0097] Applicants present certain theoretical aspects above that
are believed to be accurate that appear to explain observations
made regarding embodiments of the invention. However, embodiments
of the invention may be practiced without the theoretical aspects
presented. Moreover, the theoretical aspects are presented with the
understanding that Applicants do not seek to be bound by the theory
presented.
[0098] While various embodiments of the invention have been
described above, it should be understood that they have been
presented by way of example only, and not limitation. Numerous
changes to the disclosed embodiments can be made in accordance with
the disclosure herein without departing from the spirit or scope of
the invention. Thus, the breadth and scope of the invention should
not be limited by any of the above described embodiments. Rather,
the scope of the invention should be defined in accordance with the
following claims and their equivalents.
[0099] Although the invention has been illustrated and described
with respect to one or more implementations, equivalent alterations
and modifications will occur to others skilled in the art upon the
reading and understanding of this specification and the annexed
drawings. In addition, while a particular feature of the invention
may have been disclosed with respect to only one of several
implementations, such feature may be combined with one or more
other features of the other implementations as may be desired and
advantageous for any given or particular application.
[0100] The terminology used herein is for the purpose of describing
particular embodiments only and is not intended to be limiting of
the invention. As used herein, the singular forms "a", "an" and
"the" are intended to include the plural forms as well, unless the
context clearly indicates otherwise. Furthermore, to the extent
that the terms "including", "includes", "having", "has", "with", or
variants thereof are used in either the detailed description and/or
the claims, such terms are intended to be inclusive in a manner
similar to the term "comprising."
[0101] Unless otherwise defined, all terms (including technical and
scientific terms) used herein have the same meaning as commonly
understood by one of ordinary skill in the art to which this
invention belongs. It is further understood that terms, such as
those defined in commonly used dictionaries, should be interpreted
as having a meaning that is consistent with their meaning in the
context of the relevant art and will not be interpreted in an
idealized or overly formal sense unless expressly so defined
herein.
[0102] The Abstract of the Disclosure is provided to comply with 37
C.F.R. .sctn.1.72(b), requiring an abstract that will allow the
reader to quickly ascertain the nature of the technical disclosure.
It is submitted with the understanding that it will not be used to
interpret or limit the scope or meaning of the following
claims.
* * * * *