Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
CA 02947625 2016-11-03
POWERED LOWER LIMB DEVICES AND METHODS OF CONTROL
THEREOF
BACKGROUND
The present disclosure relates to powered prosthetic and orthotic devices,
and to bipedal robotic devices.
Every year, more than 7800 people incur a spinal cord injury in North
America. Although around 44% will end up with complete loss of locomotion, the
fortunate ones can benefit from clinical rehabilitation to help recover motor
control. Therapy methods make use of a powered exoskeleton to support the
impaired limbs, and guide them through the human walking cycle numerous
times. Powered orthoses, such as powered exoskeletons, offer the promise of
low-cost rehabilitation, together with empowering wheelchair bound individuals
to
walk, climb stairs, and stand again. Current commercially available
exoskeletons
fail to replicate the human walk due to their non-powered ankle design or
inadequate control scheme.
In order to fully realize their potential, improvements in such devices are
needed to allow for truly natural gait and reduced energy cost for the user.
These
improvements largely fall into two categories: actuator design, and control
system design. Advances in electric motor power density are allowing for more
robust and powerful actuator designs to be realized. Additionally, natural
gait has
been achieved in powered prostheses using various forms of autonomous
control. Similarly, autonomous virtual constraint control in bipedal robotics
has
1
CA 02947625 2016-11-03
allowed for robust control with guaranteed stability. Autonomous gait
controllers,
such as virtual constraint control, have yet to be applied to a hip-knee-ankle-
foot
system, such as a powered orthosis.
Current control systems of powered rehabilitation orthoses are typically
high performance in the sense of natural gait replication, or highly safe in
the
sense of user stability. However, due to their control structures, they often
fail to
achieve both goals.
SUMMARY
In some aspects, methods are provided for controlling a powered lower
limb device. A stance phase control method is disclosed in which the required
joint torque is determined based on the difference between two joint angles,
such
as the knee joint and the ankle joint. A swing control method is also
disclosed
that employs feedback-based minimum jerk trajectory control. In other
embodiments, a joint assembly for use in modular lower limb devices is
provided.
The joint assembly includes a reconfigurable slider-crank mechanism that is
configurable to provide a plurality of different ranges of rotational travel,
rotational
speeds, and torques, for customization according to different anatomical
joints.
The joint assembly may include a compact coupling device for coupling a ball
screw of the slider-crank mechanism to an output shaft of a motor. When
employed to form a modular orthosis, the joint assembly may be adapted for
self-
alignment during setup.
Accordingly, in a first aspect, there is provided a method of controlling a
2
CA 02947625 2016-11-03
hip joint of a lower limb device during a subphase of a stance phase, the
lower
limb device further comprising a knee joint and an ankle joint, the method
comprising:
receiving angle sensor information from angle sensors respectively
associated with the knee joint and the ankle joint;
receiving inertial sensor information from one or more inertial sensors
associated with the lower limb device;
processing the angle sensor information to determine a knee joint angle
and an ankle joint angle;
determining a phase variable dependent on a difference between the
knee joint angle and the ankle joint angle;
processing the inertial sensor information to determine a walking speed of
the lower limb device; and
controlling a hip joint actuator such that a hip joint torque applied to the
hip joint is constrained according to a pre-determined one-to-one dependence
on
the phase variable and the walking speed.
In another aspect, there is provided a method of controlling a lower limb
device during a subphase of a stance phase, the lower limb device comprising a
hip joint, a knee joint and an ankle joint, the method comprising:
receiving angle sensor information from angle sensors respectively
associated with the knee joint and the ankle joint;
receiving inertial sensor information from one or more inertial sensors
associated with the lower limb device;
3
CA 02947625 2016-11-03
processing the angle sensor information to determine a knee joint angle
and an ankle joint angle;
determining a phase variable dependent on a difference between the
knee joint angle and the ankle joint angle;
processing the inertial sensor information to determine a walking speed of
the lower limb device;
controlling a hip torque applied by a hip joint actuator to the hip joint
according to a first pre-determined dependence of hip torque on the phase
variable and the walking speed;
controlling a knee joint actuator such that a knee joint torque applied to
the knee joint satisfies a second pre-determined dependence of the knee joint
torque on the phase variable and the walking speed; and
controlling an ankle joint actuator such that an ankle joint torque applied
to the ankle joint satisfies a third pre-determined dependence of the ankle
joint
torque on the phase variable and the walking speed;
such that the hip joint actuator, the knee joint actuator and the ankle joint
actuator are all controlled according to the same phase variable.
In another aspect, there is provided a method of controlling a selected
joint of a lower limb device during a subphase of a stance phase, the lower
limb
device comprising a plurality of joints, the method comprising:
receiving angle sensor information from angle sensors respectively
associated with two joints of the plurality of joints;
receiving inertial sensor information from one or more inertia sensors
4
CA 02947625 2016-11-03
associated with the lower limb device;
processing the angle sensor information in order to determine a phase
variable dependent on joint angles of the two joints;
processing the inertial sensor information to determine a walking speed of
the lower limb device; and
controlling an actuator associated with the selected joint such that a joint
torque applied to the selected joint satisfies a pre-determined dependence of
the
joint torque on the phase variable and the walking speed.
In another aspect, there is provided a method of controlling a hip joint of
an orthosis during a subphase of a stance phase, the method comprising:
receiving angle sensor information from angle sensors respectively
associated with a knee joint and an ankle joint of a wearer of the orthosis;
receiving inertial sensor information from one or more inertia sensors
associated with the orthosis;
processing the angle sensor information to determine a knee joint angle
and an ankle joint angle associated with the wearer of the orthosis;
determining a phase variable dependent on a difference between the
knee joint angle and the ankle joint angle;
processing the inertial sensor information to determine a walking speed of
the orthosis; and
controlling a hip joint actuator such that a hip joint torque applied to the
hip joint satisfies a pre-determined dependence of the hip joint torque on the
phase variable and the walking speed.
CA 02947625 2016-11-03
In another aspect, there is provided a joint assembly for use with a
modular powered prosthesis or a modular powered orthosis, the joint assembly
comprising:
a support frame;
a motor supported by said support frame;
a screw coupled to an output shaft associated with said motor, said screw
comprising a nut;
a crank pivotally connected to said support frame, thereby forming a joint;
a coupling link for mechanically coupling said nut to said crank, such that
linear actuation of said nut by said motor responsively produces rotation of
said
crank about the joint;
wherein a distal portion of said coupling link is connectable to said crank
at a plurality of selectable anchor points; and
wherein each selectable anchor point is configured to provide a
respective range of rotational travel, rotational speed, and torque that is
customized for a respective anatomical joint.
In another aspect, there is provided a coupling assembly configured to
connect an output shaft associated with a motor to a secondary shaft, the
coupling assembly comprising:
a coupling for securing the secondary shaft to the output shaft; and
a coupling housing connectable to a non-rotating portion of the motor;
wherein said coupling is rotatably supported within said coupling
housing by a first angular contact bearing and a second angular contact
bearing;
6
CA 02947625 2016-11-03
and
wherein said first angular contact bearing and said second angular
contact bearing are arranged in a back-to-back configuration and reside
between
an inner surface of said coupling housing and an outer surface of said
coupling
A further understanding of the functional and advantageous aspects of the
disclosure can be realized by reference to the following detailed description
and
drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
Embodiments will now be described, by way of example only, with
reference to the drawings, in which:
FIG. 1A shows an example of a lower limb device.
FIG. 1B shows an example system including a lower limb device and a
controller.
FIGS. 2A and 2B show the dependence of hip torque on local hip angle
and the knee-ankle angle difference, respectively.
FIG. 3 shows an example stance phase control method.
FIGS. 4A and 4B show example swing phase control methods.
FIG. 5 shows an example gait control method involving stance and swing
phase control
FIGS. 6A and 6B show an example joint assembly for forming a modular
lower limb device.
FIGS. 6C and 6D show an example guidance member for guiding a nut
7
CA 02947625 2016-11-03
and bearings.
FIG. 6E shows an example crank with multiple anchor points.
FIGS. 7A-B, 8A-B and 9A-B show joint assemblies configured for the
ankle, knee and hip joints.
FIGS. 10 and 11 illustrate an example coupling device for coupling an
output shaft of a motor to a ball screw.
FIGS. 12A-B shows an example alignment system for aligning and
adjusting the length of a connection rod between joint assemblies.
FIG. 13 shows an assembly view of an example connection rod alignment
system.
FIG. 14 shows a cross-sectional view of an example connection rod
alignment system.
FIG. 15 shows an example isometric view of a modular exoskeleton
system, including an attachment to the hip actuators.
DETAILED DESCRIPTION
Various embodiments and aspects of the disclosure will be described with
reference to details discussed below. The following description and drawings
are
illustrative of the disclosure and are not to be construed as limiting the
disclosure.
Numerous specific details are described to provide a thorough understanding of
various embodiments of the present disclosure. However, in certain instances,
well-known or conventional details are not described in order to provide a
concise discussion of embodiments of the present disclosure.
8
CA 02947625 2016-11-03
As used herein, the terms "comprises" and "comprising" are to be
construed as being inclusive and open ended, and not exclusive. Specifically,
when used in the specification and claims, the terms "comprises" and
"comprising" and variations thereof mean the specified features, steps or
components are included. These terms are not to be interpreted to exclude the
presence of other features, steps or components.
As used herein, the term "exemplary" means "serving as an example,
instance, or illustration," and should not be construed as preferred or
advantageous over other configurations disclosed herein.
As used herein, the terms "about" and "approximately" are meant to cover
variations that may exist in the upper and lower limits of the ranges of
values,
such as variations in properties, parameters, and dimensions. Unless otherwise
specified, the terms "about" and "approximately" mean plus or minus 25 percent
or less.
It is to be understood that unless otherwise specified, any specified range
or group is as a shorthand way of referring to each and every member of a
range
or group individually, as well as each and every possible sub-range or sub-
group
encompassed therein and similarly with respect to any sub-ranges or sub-groups
therein. Unless otherwise specified, the present disclosure relates to and
explicitly incorporates each and every specific member and combination of sub-
ranges or sub-groups.
As used herein, the term "on the order of', when used in conjunction with
a quantity or parameter, refers to a range spanning approximately one tenth to
9
CA 02947625 2016-11-03
ten times the stated quantity or parameter.
Unless defined otherwise, all technical and scientific terms used herein
are intended to have the same meaning as commonly understood to one of
ordinary skill in the art. Unless otherwise indicated, such as through
context, as
used herein, the following terms are intended to have the following meanings:
As used herein, the phrase "virtual constraint" refers to a rule or constraint
applied or enforced by a control system to cause an orthotic (e.g.
exoskeleton),
prosthetic or robotic device to adjust its joint elements, such as position,
velocity,
or torque, in response to the value of a monotonic phase variable.
As used herein, the phrase "holonomic virtual constraint" refers to a
kinematic relationship between links of a device that is enforced via
feedback,
where a single phase variable is employed per constraint.
As used herein, the phrase "function" or "function relationship" refers to a
relation between a set of inputs and a set of permissible outputs with the
property
that each input is related to exactly one output.
Referring now to FIG. 1A, an example of a lower limb orthosis is shown.
The example lower limb orthosis 100 includes three powered joints ¨ a powered
ankle joint 110A, a powered knee joint 110K, and a powered hip joint 110H,
each
actuated by respective motors 120A, 120K and 120H. The device attaches on
the outside of the operator's lower limbs via adjustable cuffs 150, and is
linked
together around the back of the user's waist by a support brace (not shown in
the
figure, but shown in FIG. 15).
The powered knee joint 110K is connected to the powered hip joint 110H
CA 02947625 2016-11-03
and the powered ankle joint by respective connection rods 130 and 135. The
joint
angles of the ankle joint, the knee joint and the hip joint are measurable
based on
signals received from respective angle sensors 140A, 140K and 140H. One or
more inertial sensors (not shown in FIG. 1A) may be employed to generate
signals that can be processed to determine kinematic measures such as velocity
and acceleration. One or more additional sensors may be employed to provide
signals that can be processed in order to determine the instantaneous gait
phase
(stance or swing) during use of the lower limb device, as well as to detect
subphases of the stance phase.
Referring now to FIG. 1B, an example powered lower limb system is
shown that includes a powered lower limb device 100 that is operatively
coupled
to a control and processing unit 200. The figure schematically illustrates the
connection of the control and processing hardware 200 to various joint
actuators
and sensors of the lower limb device 100, including connections to a hip
actuator
and hip angle sensors 280H, a knee actuator and knee angle sensors 280K, and
an ankle actuator and ankle angle sensors 280A (although not shown in the
figure, the control and processing hardware is also connected to one or more
inertial sensors associated with the lower limb device 100, and to one or more
additional sensors for gait phase determination). Although the lower limb
device
100 is shown as powered orthosis having powered hip, knee and ankle
actuators, it will be understood that this orthosis is merely provided as an
example, and that the control and processing system 200 may be employed to
control a wide variety of powered devices for gait control, such as orthoses
(e.g.
11
CA 02947625 2016-11-03
exoskeletons), prostheses, and robotic devices.
As shown in the example embodiment illustrated in FIG. 1 B, control and
processing hardware 200 may include a processor 210, a memory 215, a system
bus 205, a data acquisition and control interface 220 for acquiring sensor
data
and for sending control commands to the lower limb device 100, a power source
225, and a plurality of optional additional devices or components such as
storage
device 230, communications interface 235, display 240, and one or more
input/output devices 245.
The example methods described herebelow for controlling a powered
lower limb device can be implemented via processor 210 and/or memory 215. As
shown in FIG. 1B, the processing of sensor signals in order to provide gait
control the powered lower limb device 100 is performed by control and
processing hardware 200, via executable instructions represented as swing
control module 250, stance control module 255, and gait phase (and subphase)
detection module 260, and a walking speed detection module 265. Various
examples of these control algorithms are described in detail below.
It is to be understood that the example system shown in the figure is not
intended to be limited to the components that may be employed in a given
implementation. For example, in one example implementation, the processing
hardware 200 may be provided on a computing device that is supported by the
lower limb device 100. Alternatively, one or more components of the processing
hardware 200 may be physically separate from the lower limb device 100. For
example, the processing and computing hardware 200 may include a mobile
12
CA 02947625 2016-11-03
computing device, such as a tablet or smartphone that is connected to a local
processing hardware supported by with the lower limb device via one or more
wired or wireless connections. In another example implementation, a portion of
the control and processing hardware 200 may be implemented, at least in part,
on a remote computing system that connects to a local processing hardware via
a remote network, such that some aspects of the processing are performed
remotely (e.g. in the cloud).
The methods described herein can be partially implemented via hardware
logic in processor 210 and partially using the instructions stored in memory
215.
Some embodiments may be implemented using processor 210 without additional
instructions stored in memory 215. Some embodiments are implemented using
the instructions stored in memory 215 for execution by one or more
microprocessors. A computer readable storage medium can be used to store
software and data which when executed by a data processing system causes the
system to perform various methods. The executable software and data may be
stored in various places including for example ROM, volatile RAM, nonvolatile
memory and/or cache. Portions of this software and/or data may be stored in
any
one of these storage devices. As used herein, the phrases "computer readable
material" and "computer readable storage medium" refers to all computer-
readable media, except for a transitory propagating signal per se.
As indicated in FIG. 1 B, the lower limb device may be controlled according
to the phase of the gait cycle, where separate control algorithms are employed
during the swing phase and the stance phase. As shown at 260, a gait phase
13
CA 02947625 2016-11-03
detection module may be employed to process signals from sensors associated
with the lower limb device 100 in order to determine a gait phase thereof.
Depending on the type of swing and stance phase control algorithms employed,
the signals may also be processed to determine a subphase (e.g. early, mid and
late) of the stance phase.
In one example implementation, the gait phase may be determined based
on force or pressure sensors located on the lower limb device 100. For
example,
force sensing resistors may be provided at the heel and toe to detect heel and
toe contact with the ground. Each force sensor can have a threshold associated
therewith for detecting contact. The early-, mid-, and late-stance subphases,
and
the swing phase, can then be determined as follows:
HEEL ON HEEL OFF
TOE ON Mid Late
TOE OFF Early No ground contact
(swing)
This example approach allows for the differentiation between the three
subdivided states (subphases) of stance phase, and also allows for forward
(clockwise through table) and backward transitions (counter-clockwise through
table) between the phases and subphases. Such an implementation is suitable
for gait phase tracking for both forwards and backwards walking. It will be
understood that the aforementioned gait phase detection implementation is
provided merely as an example, and that other methods of gait phase detection
may be employed. For example, other methods of gait phase segmentation are
disclosed by Lenzi (Lenzi et al., IEEE Robotics & Automation Magazine,
14
CA 02947625 2016-11-03
December, 94-107, 2014).
Method of Stance Control for Lower Limb Device
According to various example implementations of the present disclosure,
in which separate control algorithms are employed for swing and stance phases,
a stance control method may be employed to control the actuation of one or
more joint actuators during a series of stance subphases through the use of a
virtual constraint in which the torque applied to each joint is enforced
according
to a pre-determined dependence of torque on a plurality of joint angles and
walking speed.
In previous reported studies involving the use of a virtual constraint, such
as the method disclosed by Lenzi (Lenzi et al. , IEEE Robotics & Automation
Magazine, December, 94-107, 2014), a virtual constraint stance control method
was employed for the control of a powered prosthesis involving the knee joint
and the ankle joint. According to the stance control method of Lenzi, each
joint
actuator was controlled based on a local joint angle and the measured walking
speed ¨ i.e. the torque applied to the knee joint was determined based on the
local knee angle and measured walking speed, and the torque applied to the
ankle joint was determined based on the local ankle angle and the measured
walking speed.
Although the local joint control method of Lenzi was successful in
controlling a prosthesis having a knee joint and an ankle joint, the present
inventors discovered that such a stance phase control scheme is inoperable
when applied to control the hip joint actuator of a lower limb device having a
CA 02947625 2016-11-03
powered hip joint. Indeed, as clearly shown in FIG. 2A, the dependence of
required hip joint torque on local hip joint angle is not a function. This
lack of a
function between hip joint torque and local hip joint angle precludes the use
of
the local hip joint angle as a phase variable for controlling the torque
applied to
the hip joint.
The present inventors discovered that if the difference in between the
knee joint angle and the ankle joint angle is used as a phase variable, a not
a
function of hip joint torque with respect to this phase variable is obtained.
For
example, FIG. 2B shows the dependence of hip joint torque on knee-ankle angle
for a range of walking speeds, clearly illustrating a functional relationship
between hip joint torque, knee-ankle angle, and walking speed, for the late
stance subphase. The inventors found that such function profiles could be
obtained for each subphase of the stance phase.
It was also found that the knee-ankle angle was a suitable phase variable
for controlling the torque applied to both the knee joint and the ankle joint
of a
lower limb device. Accordingly, the knee-ankle angle and the walking speed may
be employed as global phase variables for controlling the torque applied to
each
joint of a lower limb device having powered hip, knee and ankle joints.
Accordingly, in one example embodiment, a method for controlling the
torque applied to the hip joint involves controlling the hip joint torque by
enforcing, for each stance subphase, a pre-determined dependence of hip joint
torque on knee-ankle angle difference and walking speed.
In another example embodiment, a pre-determined relationship (e.g.
16
CA 02947625 2016-11-03
virtual constraints) may be determined for each joint of a lower limb device,
for
each subphase of the stance phase, where each pre-determined relationship
prescribes the torque required for each joint to produce a suitable gait
according
to the knee-ankle angle difference and the walking speed. For example,
separate
virtual constraint relationships (e.g. look-up tables) may be provided for
each of
the early, mid-, and late stance phases. Through the use of such a virtual
constraint controller, the lower limb device is decoupled from following a
predetermined gait trajectory. This advantage allows the lower limb device to
be
more able to respond to changes in environment such as terrain
conditions/slopes, and changes in the user such as various joint velocities.
The dependence of torque on joint angle difference and walking speed
may be determined, for example, via experiments with subjects in which joint
angles, walking speed, and torques are measured, or, for example, based on
published biomechanical gait data, such as the data provided in Winter
(Winter,
D., The Biomechanics and Motor Control of Human Gait: Normal, Elderly and
Pathological, 1991). The pre-determined dependence on torque may be
normalized by bodyweight, thereby providing a relationship that may be
customized to a particular patient upon input of the patient's body weight.
It is noted that while the preceding example embodiment pertains to the
use of the knee-ankle angle difference as the phase variable, it will be
understood that other phase variables, generated based on the knee-ankle angle
difference, may alternatively be employed. For example, other functions may be
generated, that are dependent, at least in part, on the knee-ankle angle
17
CA 02947625 2016-11-03
difference, while providing a functional relationship with joint torque.
Referring now to FIG. 3, an example control algorithm is provided for
stance control according to a virtual constraint that employs joint angle
difference
as a phase variable. A pre-determined dependence of applied joint torque on a
an angle-difference based phase variable and the walking speed is provided as
a
constraint on torque, as shown at 300. The inputs to this constraint are the
phase
variable 305, which is determined based on joint angle sensors 310, and the
walking speed estimation 315, which is determined based on input from inertial
sensors measurements as shown at 320. As shown in the figure, the joint angle
sensor measurements may be employed, along with the inertial sensors, in order
to estimate the walking speed at 315. The torque setpoint, determined
according
to the constraint 300, is compared at 325 to the measured joint torque 330,
and a
control system (e.g. a PID control system as shown at 335) is employed to
control the motor controller 340 in order to produce the desired applied
torque via
actuation of the joint actuator, shown at 345. The joint torque may be
measured,
for example, using sensors such as strain gauges. Alternatively, the joint
torque
may be indirectly determined by monitoring a current applied to the actuator
motor.
The constraint 300 may be provided as a lookup table for interpolation
during implementation. For example, a lookup tables may provide torque values
for each controlled joint that are valid for walking speeds ranging between
0.5
m/s and 1.75m/s, thus allowing for natural preferred gait of 1.4 m/s along
with
slower and faster walking speeds. The use of the walking speed as a lookup
18
CA 02947625 2016-11-03
variable allows for a wide range of applicable walking speeds through torque
modulation.
Although the preceding example implementation pertained to the control
of all joints of a lower limb device having powered hip, knee and ankle
joints, it
will be understood that the preceding example method may be adapted to control
a wide variety of lower limb devices.
In one example embodiment, the aforementioned stance control method
may be adapted to control only the hip joint (or hip joints) of a powered
lower limb
device. For example, if a lower limb device includes a powered knee and a
powered ankle joint in addition to a powered hip joint, then the stance
control
method may be employed to control the hip joint, while other control methods
(such as based on local angle measurement) may be employed to control the
knee and ankle joints. In another example embodiment, a lower limb device may
include a powered hip joint in the absence of powered knee and ankle joints.
In
such an implementation, local angle sensors associated with the unpowered
joints may be employed to provide the angle sensing inputs to the virtual
constraint controller for the powered hip joint.
In some example embodiments, the aforementioned stance control
methods may be adapted to control the torque applied to a first joint of a
lower
limb device according to a virtual constraint that is dependent on the walking
speed and phase variable constructed from the local angles associated with a
second joint and a third joint. The phase variable may be dependent on the
difference between the joint angles of the second and third joints.
19
CA 02947625 2016-11-03
As noted above, the present method was discovered based on the need to
solve a technical problem in which the absence of a functional relationship
between a local angle of a given joint and a torque applied to the given
joint,
where a functional relationship between torque and a phase variable was
obtained by employing a phase variable involving a difference between two
joint
angles. The local angles may be associated with joints other than the given
joint,
as in the case of the example implementation in which the hip joint torque is
controlled according to the difference between the knee and ankle joint
angles.
A benefit of the aforementioned stance control method is that patient-
specific tuning, other than inputting the patient's weight, is reduced or
eliminated
through the angle-torque relationship, which is independent of joint
velocities that
vary across patients. Furthermore, by interpolating the profiles based on
walking
speed, the requirement for speed-specific tuning is eliminated. This secondary
feature of the control system allows for a more involved rehabilitation
session in a
similar time frame to that of other exoskeletons, as a therapist should not
need to
spend significant time tuning the exoskeleton.
The present example stance control method varies in several key aspects
from the predetermined gait trajectory controllers currently used for
rehabilitation
exoskeletons. Notably, the present control methods provide an autonomous
control strategy. This autonomous structure is inherently more robust to
disturbances as there is no push to re-synchronise with time. Secondly, the
joint
level controllers enforce torque, as opposed to position, during stance phase.
As noted above, the choice of a joint angle difference as virtual constraint
CA 02947625 2016-11-03
for this controller is also unique when compared with other controllers. The
angle-difference constraint is the mathematical difference and requires only
local
joint data. In comparison, the centre of pressure constraint implemented in
prosthesis control by Gregg et al. (Gregg et al., IEEE Transactions on
Robotics
30, 1455-1471, 2014 ) requires Cartesian coordinates and interaction forces
for
implementation. Local angles are significantly easier to sense compared to
Cartesian coordinates, and interaction forces would require a 3-DOF load cell
at
the minimum. The simplicity of sensing the required data and the mathematical
simplicity of calculation are thus benefits of the present virtual constraint
method
based on a joint angle difference as a phase variable.
The present control method differs greatly from conventional practice in
commercial powered orthoses, which presently employ predetermined gait
trajectory control. Such conventional implementations track a recorded motion,
typically from known able-bodied "normal" walking, in an attempt to reproduce
natural gait. In contrast, the example control methods disclosed herein employ
a
phase variable to determine the torque required to continue motion during
various subphases of the stance phase of the gait cycle. The present
autonomous control methods may be beneficial in providing more robust control,
as the requirement for synchronization with time is eliminated.
Moreover, the present example methods of stance control differ from
previous implementations that employ holonomic virtual constraints. In
previous
implementations involving virtual constraints in the field of bipedal
robotics, a
kinematic relationship (i.e. only position information) between an input
variable
21
CA 02947625 2016-11-03
and multiple output parameters, such as joint angles, is employed. Such
methods
have been employed by Westervelt et at., for example, to determine joint
angles
based on the angle of a virtual link connecting the hip and ankle (Westervelt
et
al., Feedback Control of Dynamic Bipedal Robot Locomotion, 2007). However,
the present example stance control methods differs from such approaches, as
the present example methods employ a direct relationship between joint angle
(input) and joint torque (output), as opposed to an indirect relationship
mediated
through a holonomic virtual constraint.
Finally, as noted above, that the present control method differs
significantly from the control method disclosed by Lenzi et al. (Id.). In
Lenzi's
stance phase control, the local joint angle (i.e. ankle or knee) is employed
to
determine the corresponding local joint torque. As described above, the
present
inventors have found such an implementation to be inoperable for control of
the
hip joint, due to a many-to-one relationship between the hip joint torque and
the
local hip angle. In contrast, when the present stance phase control method is
applied to control a hip joint actuator, the difference in the knee-ankle
angle is
employed as a phase variable to provide a virtual constrain that is a function
for
controlling the hip joint torque. Moreover, as described above, the phase
variable
based on the difference between the knee angle and the ankle angle may be
employed as a global phase variable for controlling the torque applied to the
knee and ankle joints.
Method of Swing Control for Lower Limb Device
A minimum jerk trajectory is often used in path planning as it is analogous
22
CA 02947625 2016-11-03
with the smoothest possible path for an object to travel though.
Traditionally, a
minimum jerk trajectory is calculated beforehand (a priori) according to the
initial
position and desired final position and time duration as shown in equation (1)
from Flash & Hogan (Flash & Hogan, The Journal of Nneuroscience, Vol 5(7),
pp. 1688-1703, 1985:
P(t) = Pi + (Pi ¨ Pf) * (15*-1^4 ¨ 6*TA5 ¨ 10*-1^3) (1)
where: P(t) ¨ position, Pi ¨ initial position, Pf ¨ final position, T ¨
normalised time
= (t/ff). This path can then be used to control the motion of an object, such
as the
joint(s) of a lower limb device, using a control system. Such known jerk
minimization swing control methods therefore employ a precalculated minimum
jerk trajectory.
Referring now to FIGS. 4A and 4B two example algorithms are shown for
controlling the swing phase of a lower limb, in which a feedback-based minimum
jerk trajectory control scheme is employed for control of the swing phase of
gait.
This method enable the controller to respond to disturbances in real-time with
a
minimum jerk trajectory.
Unlike previously known swing phase control methods, the example
method illustrated in FIG. 4A involves the calculation of a swing trajectory
based
on a minimum jerk calculation at each time point in real time. The general
equation (2) from Liu & Todorov (Liu & Todorov, The Journal of Nneuroscience,
Vol 27(35), pp. 9354-9368, 2007), shown below, demonstrates this calculation
of
23
CA 02947625 2016-11-03
jerk according to input of current position, velocity, and acceleration of the
joint,
as well as the remaining time of the trajectory:
J(t) = ( 60 / (DA3) )* (Pf ¨ P(t)) ¨ ( 36 /(D"2) )* V(t) - ( 9 / D )* A(t)
(2)
where: J(t) ¨ calculated minimum jerk, D ¨ remaining time = (if - t), Pf ¨
final
position, P(t) ¨ current position, V(t) ¨ current velocity, A(t) ¨ current
acceleration.
This equation is derived from a minimisation problem and is thus produces the
minimum jerk at each time point. In order to produce a trajectory, this
equation is
integrated three times in order to determine the position setpoint for the
trajectory.
An example implementation of this method is shown in FIG. 4A, as
explained herebelow. As shown at 400, the minimum jerk equation, such as
equation (2) above, which is dependent on position, velocity and acceleration,
is
provided with the desired position 405, the desired time duration 410, and the
measured inputs of position 415, velocity 420, and acceleration 425. The
minimum jerk equation is processed to calculate the minimum jerk at 400, and
integrated via the three integration operations 430 in order to generate a new
real-time position setpoint.
This position setpoint is compared at 435 with the measured positon 415
to determine a position error, which is employed, via a controller 440 (such
as a
PID controller), to control the motor controller 445. The motor controller 445
employs an internal PID loop (not shown) to determine ensure either the
required
24
CA 02947625 2016-11-03
velocity or torque to move the joint to the setpoint is realized by the joint.
The
internal PID loop on the motor controller enables a determination of the
velocity/torque required to move towards the desired angle, and the main
feedback loop will reduce this velocity/torque according to the error between
current and desired positions ensures that the correct torque/velocity is
applied to
the joint, while the main PID loop calculates the required torque/velocity to
move
the joint towards the desired setpoint. In this way, the main loop ultimately
controls the position of the joint, while the internal loop regulates the
torque
applied during motion.
FIG. 4B illustrates an alternative example implementation, in which a
velocity sensor 460 is employed to measure the velocity of the joint. This
method
is expected to allow for higher fidelity acceleration calculation 465, as the
second
numeric derivative tends to be of low fidelity due to propagation of error.
Unlike minimum jerk swing control methods known in the art, such as the
method disclosed by Lenzi et al. US Patent Application Publication No.
US20160058582, jerk is minimized while traversing between a given starting and
ending point using a feedback loop that regulates jerk, rather than position.
More
specifically, the method disclosed by Lenzi et al. regulates jerk at a given
point in
time according to a specific position that is computed prior to beginning of
the
swing phase. After selecting a given starting and ending point and a required
duration of time, Lenzi et al. calculates the position vs. time profile that
minimizes
jerk, thereby providing a position profile which is enforced during the swing
phase
via a feedforward and feedback loop. In contrast; according to the present
CA 02947625 2016-11-03
example swing phase control method, a given start and ending point and a
required duration of time are selected, but a position profile for the swing
phase
is not calculated. Instead, a feedback loop is employed to determine, in real-
time,
the appropriate torque that will minimize the jerk for the remainder of the
swing
phase, while still achieving the required end-point at end-time, based on the
current point/velocity/acceleration measurements.
By using a minimum jerk trajectory, the swing controller is able to provide
a smooth motion independent of the stance phase controller. At the end of
stance phase, the current joint angle, velocity, and acceleration are simply
used
as inputs to swing phase controller, thus allowing the swing controller to
function
regardless of the previous stance phase performance. In order to allow for
proper
foot clearance, a maximum knee angle may be enforced as a target point in the
trajectory. The swing duration is calculated in proportion to the stance phase
duration based on able-bodied biomechanics. The minimum jerk trajectory,
generated according to the real-time feedback method described above, is
capable of responding to external disturbances in a minimum jerk trajectory
thereby ensuring smooth motion throughout swing.
In some example implementations, the aforementioned stance phase
control method may be employed with the aforementioned swing phase control
method, in order to provide control system suitable for controlling the
complete
gait cycle. An example implementation of such a combined control method is
illustrated in FIG. 5. In other example embodiments, the aforementioned stance
control method may be combined with a different swing phase control method, or
26
CA 02947625 2016-11-03
the aforementioned swing phase control method may be combined with a
different stance phase control method.
Although the preceding example implementations pertained to the control
an orthotic lower limb device, it will be understood that the present stance
control
embodiments may be readily adapted for the control of prosthetic and robotic
devices.
Modular Actuator for Lower Limb Device
Referring again to FIG. 1A, the example lower limb actuator 100 is formed
from a plurality of joint assemblies 110A, 110K and 110H in a modular fashion.
As can be seen from the figure, a common actuator mechanism is employed to
power each of the hip, knee, and ankle joints in the sagittal plane.
The knee joint assembly 110K is linked to the ankle joint assembly 110A
and the hip joint assembly 100H by respective connection rods 130 and 135.
Such connection rods may be formed, for example, as hollow tubes formed from
carbon fiber. As described in further detail below, the joint assemblies are
configured to receive and secure ends of the connection rods 130 and 135 such
that the inter-joint distance is adjustable, thereby allowing adjustment of
the
modular lower limb device so the axis of rotation of each actuated joint is
aligned
with each respective anatomical joint of the device wearer. Once adjusted, the
joint-to-joint length may be locked, after which the lower limb device may
transfer
its weight to the ground.
FIG. 6A illustrates a joint assembly 600 for use in forming a powered lower
limb device, such as the lower limb device shown in FIG. 1A. Accordingly, a
27
CA 02947625 2016-11-03
plurality of such joint assemblies 600 can be employed for different joints,
such
as the hip, knee, and ankle joints. The joint assembly 600 employs a
reconfigurable offset slider-crank mechanism, which takes advantage of the
varying range of motion between each joint to provide for more speed or torque
as appropriate.
As shown in FIG. 6A, the example joint assembly 600 is powered by a
motor 605, such as a brushless DC motor. The non-rotating portion of the motor
605 is supported on a rigid support frame 650, and a crank 660 is pivotally
coupled to the support frame 650 through a joint pin 665, thereby forming a
joint.
An output shaft (not shown in FIG. 6A) associated with the motor 605 is
coupled,
through a coupling housed within housing 700, to a screw 610, which, through
its
associated nut (not shown in FIG. 6A), converts rotary to linear motion. An
example implementation of the coupling is described in further detail below.
The
crank 660 is mechanically coupled, either directly, or indirectly (as shown in
FIG.
6A and 6B), to a nut provided on the screw 610 via a coupling link 640, such
that
linear actuation of the nut by the motor responsively produces rotation of the
crank about the joint. The output shaft associated with the motor 610 may be a
motor shaft or an output shaft of a gearbox.
In the example embodiment shown in FIGS. 6A and 6B, the screw-based
linear actuator is implemented as a ball screw and a ball nut. The ball nut
actuates the slider joint, generating axial forces with response to input
torques on
the 610 ball screw. The high transmission efficiency renders the mechanism
backdrivable, such that that an input axial force on the ball nut will result
in the
28
CA 02947625 2016-11-03
rotation of the ball screw. It will be understood, however, that a wide range
of
alternative configurations may be employed, such as lead screws, roller
screws,
and differential roller screws.
In the non-limiting example embodiment shown in 6B (which shows a
partially transparent view), the ball nut 615, designed to handle axial loads
exclusively, is also supported with two linear ball bearings 620 on two linear
shafts 625 parallel to the ball screw 610. According to this example
implementation, the slider joint consists of a guide member 630 (a ball screw
nut
attachment) that is shown in FIG. 6A, and in FIGS. 6C and 6D. The ball nut 615
is attached to the guide member 630 and within aperture 632, and the two
linear
ball bearings 620 are respectively retained within apertures 634 and 636 of
the
guide member 630 by retaining rings. As shown in FIG. 6A, the coupling links
640 of the slider-crank mechanism are also joined to the guide member 630, and
to the crank 660, by the combination of shoulder bolts and corresponding
diameter bushings (such that the coupling link 640 is indirectly coupled to
the ball
nut via the guide member 630). These provide proper bearing interfaces for the
rotating surfaces between coupling link to slider joint, and coupling link to
crank.
The two coupling links 640 in turn transmit forces generated forces in the
slider
joint, down to the crank 660. The crank 660 is preferably positioned, during
use,
to be aligned with or near the user's joint axis, in order to prevent, reduce
or
minimize undesired joint reaction forces. In the example embodiment shown in
FIGS. 6A and 6B, the two cranks 640 are provided, with one on each side of the
mechanism, and are fixed to rotate together through the joint shaft. In the
29
CA 02947625 2016-11-03
example embodiment shown in FIG. 6A, the joint interface 665 consists of a
double D shaft and accordingly shaped hole in the crank 640. Rotation of the
crank 660 and joint shaft relative to the static frame 650 is facilitated
through two
flanged bushings.
One or more sensors may be provided on (attached to) the joint assembly
600, such as angle sensors, force sensors, and inertial sensors. As shown in
FIG. 6B, an angle sensor 670 is integrated with the joint assembly to provide
an
angle sensing signal.
The crank 660 may include a plurality of selectable anchor points for
selective, and optionally reconfigurable, connection of the distal portion of
the
coupling link 640. For example, as shown in FIG. 6E, the crank 660 includes
three selectable anchor points 662, 664 and 666, where FIG. 6B shows an
example configuration in which the coupling link 640 is connected to anchor
point
662. These anchor points vary the lever arm associated with the slider-crank
mechanism, thereby permitting the selection of a different range of rotational
travel, rotational speed, and torque. Such changes the effective lever arm
provide more or less rotational travel, at the expense of torque.
In one example embodiment, the selectable anchor points of the crank
660 may be configured to provide a respective range of rotational travel,
rotational speed, and torque that is customized for a respective anatomical
joint.
Accordingly, the joint assembly 600 can be reconfigured to permit the
selection of
a suitable anchor point of the coupling link to the crank in order to
configure the
joint assembly for different use in actuating a desired anatomical joints. A
set of
CA 02947625 2016-11-03
such joint assemblies, each having a common mechanical design, may thus be
individually configured for actuating different anatomical joints, and
assembled in
a modular fashion to form a powered orthotic (e.g. exoskeleton), prosthetic,
or
robotic device.
FIGS. 7A-B, 8A-B, and 9A-B illustrate the configuration of the joint
assembly of FIGS. 6A-B according to the actuation properties and requirements
of different anatomical joints. In the example configuration shown in FIGS. 7A
and 7B, the joint assembly 600 is customized for the actuation of an ankle
joint,
in which the coupling link 640 is connected to anchor point 662, where FIGS.
7A
and 7B show the two limits of the range of rotational motion. In the example
configuration shown in FIGS. 8A and 8B, the joint assembly 600 is customized
for the actuation of a knee joint, in which the coupling link 640 is connected
to
anchor point 666, where FIGS. 8A and 8B show the two limits of the range of
rotational motion. In the example configuration shown in FIGS. 9A and 9B, the
joint assembly 600 is customized for the actuation of a hip joint, in which
the
coupling link 640 is connected to anchor point 666, where FIGS. 9A and 9B show
the two limits of the range of rotational motion.
As the ankle joint requires larger torque requirement than the hip and
knee, while also a smaller range of motion, the anchor point 662 employed in
FIGS. 7A-B is located closer to the joint axis than the anchor points in FIGS.
8A-
B and 9A-B. In the ankle configuration, the lever arm is 50 mm allowing for
160
Nm of torque and a rotational travel of 65 degrees. While for the knee
configuration, a maximum travel of 110 degrees is achieved with torques of up
to
31
CA 02947625 2016-11-03
90 Nm. The hip is kinematically similar to the knee but with a 10 degree
offset.
The design is inherently safe due to the fact that each joint is mechanically
limited within the human range of motion. In some example implementations, the
anchor points may be configured as follows:
Max Torque Range of
Min Torque (Nm)
(Nm) Motion (deg)
22.5 35
Ankle -0.2 - 175
dorsiflexion plantarflexion
0 110
Knee - 0.1 - 100
extension flexion
Hip - 0.1 - 17.5 98
extension flexion
5
In one example implementation, these maximum torque values all
translate to the same thrust load of 3500 N, which is established as the
maximum
operating load. The actuator is actually limited by the bearings which can
support
up to 4400 N (the ball screw is up to 5800 N), while the motor could produce a
10 maximum of -5500-6000 N.
Compact Coupling Mechanism
In one example implementation, in order to protect the weak bearings
inside the brushless motor from the axial loads generated, a ball screw
support
mechanism may be employed. Such a ball screw support solution may be made
15 compact through the design of a mechanical coupling (between motor shaft
and
ball screw), which allows for angular contact bearings to be placed on the
outer
diameter of the coupling. This configuration allows for the use of larger
diameter
angular contact bearings, as opposed to conventional coupling methods in which
the bearings are placed along the ball screw shoulder. An example
32
CA 02947625 2016-11-03
implementation of such a coupling is shown in FIG. 10, and an exploded view in
FIG 11.
The brushless DC motor 605, given appropriate feedback, allows for
speed and torque control of its output shaft 680. The output shaft 680 is
directly
coupled to the miniature ball screw 610 through a mechanical coupling 710,
having an internal bore. The coupling 710 offers two different shaft to hub
interfaces. On the motor (proximal) side, the proximal end 705 of the coupling
710 is split, and two bolts are used to clamp the coupling 710 to the motor
shaft
680 for torque transfer through frictional force. The head of the two bolts
rest
within counter-bores in the coupling shoulder 715, which play a beneficial
role
during assembly, as noted below.
On the ball screw (distal) side, the distal end of the coupling 710 features
an 8 degree tapered bore, for use with a collet 720, such as an ER collet.
With
the ball screw shoulder inserted, tightening a standard collet nut 730 against
the
ER collet 720 pushes it further into the tapered cavity, greatly increasing
the
interface pressure and frictional force. This locking mechanism ensures proper
shaft to hub interface for the transfer of torque, axial load, while
maintaining ball
screw alignment.
As the motor torque is transmitted to the ball screw through the coupling,
the ball nut then converts this torque to a linear force. This bidirectional
axial load
is then transmitted back through the collet 720 and to the coupling 710. When
the
ball screw is in tension, the coupling shoulder 740 resting against the inner
ring
of the angular contact bearing 750A transmits the load to the bearing housing.
33
CA 02947625 2016-11-03
When in compression, the bearing lock nut 735 that rests against the inner
ring of
the other angular contact bearing 750B transmits the load to the motor
mounting
plate 760, which, together with the distal bearing housing portion 770, forms
the
coupling housing 700 shown in FIG. 6A. The motor mounting plate 760 is
secured with four socket head screws that clamp the motor mounting plate and
distal bearing housing 770 to the frame. The alignment of the support assembly
to the frame is assured through the geometric extrusion on the bearing
housing.
The counter-bores (with no bolts inside) may be used to prevent rotation
of the coupling 710 during tightening of the bearing lock nut 720 and collet
nut
725. This is achieved by placing two solid shafts within them (e.g. having a
diameter of 5.95mm), which extends between the bearing shoulder extrusions of
the motor mounting plate.
In some example embodiments, angular contact bearings are employed to
handle high axial load. The example embodiment shown in FIG. 10 (2 bearing on
coupling) provides a compact, small to medium load application. In some
example embodiments, the angular contact bearings could be replaced by deep
groove ball bearings for cost saving reasons.
Although the present example coupling embodiment is illustrated in the
context of driving a ball screw assembly for use in a lower limb device, it
will be
understood that the present coupling device may be employed for a wide variety
of other uses, and to drive different types of secondary shafts. For example,
the
coupling device could be employed to drive a device other than a screw & nut,
thereby supporting only radial loads (the coupling could use a different
secondary
34
CA 02947625 2016-11-03
shaft-hub interface at this point, i.e. without a collet). The collet locking
interface
is preferred for its integrity in transmitting high loads and maintaining good
alignment. A clamping interface is only need on the motor side, in order to
transmit the motor torque.
Self-Alignment of Modular Lower Limb Device
In some example embodiments, the joint assembly may be adapted to
include a joint alignment mechanism, in which anatomical reaction forces are
employed to automatically adjust the length and rotation between two joint
assemblies (z-alignment + rotation about z ¨ where z is the direction along
the
axis of a connection rod connecting two adjacent joint assemblies).
Referring to FIG. 1A, once a modular lower limb device, formed from joint
assemblies as described above, is fastened to the user through a cuff
interface,
thereby providing x & y alignment, the length degree of freedom between the
joint assemblies (z-alignment) may be left unconstrained, such that the
connection rods 130 and 135 are permitted to slide relative to the support
frames
of the joint assemblies.
The relative slidable translation of a connection rod 130 relative to a given
joint assembly 600 is illustrated in FIGS. 12A-B, where FIG. 12A shows the
connection rod fully inserted into the support frame, through a locking
assembly
800, in a configuration that minimizes the inter-joint separation. FIG. 12B
shows
the connection rod 130 in a configuration that maximizes the inter-joint
separation. The user and device are then moved through some of the operating
range of motion, and if any misalignment is present, the length of the
connecting
CA 02947625 2016-11-03
rod between the two joint assemblies will vary. If the x & y alignment from
the cuff
interface is correct, the rod will be coaxial to the anatomical joint axis,
and the
length between the joint assemblies will naturally converge to its anatomical
neighbor. The length and rotation degree of freedom can then be locked; joint
alignment is complete. If convergence does not happen and the joint length
keeps varying throughout the range of motion, it would indicate that the x & y
alignment is incorrect. In other example embodiments, alignment mechanism
may also be provided to permit self-aligning capabilities also in the x and y
directions.
The locking of this degree of freedom (z) is realized by a clamping collar
810 that can be tightened to apply a clamping force around the connecting rod
130, as shown in FIGS. 13A-B and FIG. 14. Adjacent to both sides of the collar
are two flanged bushings 820 and 825, which provide the sliding interface for
the
connecting rod 130. The flanges of the bushings 820, 825 provides a bearing
surface against the collar if the rotation degree of freedom is desired to be
kept
free while the collar is locked (if the collar is unlocked, the length +
rotation
degrees of freedom will be free). To lock the rotation of the connecting rod
130,
the two top screws 832, 834 can be tightened to clamp both flanges of the
bushings against the collar via pressure from the length adjustment top plate
830
and the linear shaft top plate 840. During tightening of the screws, the four
screws 842 against length adjustment side plate 840 are kept loose. A slot in
length adjustment side plate 840 allows for small vertical displacements of
length
adjustment top plate 830. This assures proper friction force between the
collar
36
CA 02947625 2016-11-03
810 and flanged bushings 820, 825, and prevents unwanted rotation of the joint
assembly.
In one example implementation, this connecting rod support within each
joint assemblies is configured to permit a possible inter-joint length
variation of
84mm between the ankle and the knee (to account for differences in anatomical
joint lengths), and 160mm between the hip and the knee (two mechanisms with
full range). This range is sufficient to accommodate a range of patients
spanning
a 1 percentile male to a 99th percentile male by variation of the separation
between each joint.
The specific embodiments described above have been shown by way of
example, and it should be understood that these embodiments may be
susceptible to various modifications and alternative forms. It should be
further
understood that the claims are not intended to be limited to the particular
forms
disclosed, but rather to cover all modifications, equivalents, and
alternatives
falling within the spirit and scope of this disclosure.
37
