You are on page 1of 7

Proceedings of 2017 IEEE

International Conference on Mechatronics and Automation


August 6 - 9, Takamatsu, Japan

Research on the Cable-pulley Underactuated


Lower Limb Exoskeleton

Yang Liu, Yongsheng Gao, Feiyun Xiao, Jie Zhao


State Key Laboratory of Robotics and System
Harbin Institute of Technology
Harbin, Heilongjiang , 150080, China
gaoys@hit.edu.cn

Abstract - A wide variety of lower limb exoskeletons have Additionally, the passive and the quasi-passive exoskeleton
been applied in the fields of load-carrying augmentation, walking [2] cannot provide positive assistance power for the load-
assistance, and rehabilitation training. The traditional carrying walking. For the traditional underactuated
exoskeleton based on fully actuated joints can provide powerful exoskeleton [3], fewer actuators are used to reduce the overall
assistance for human motion, while it is bulky requiring a large
energy consumption, which limits its application. The passive and
mass of the exoskeleton and the inertia of the movement of the
quasi-passive exoskeletons are weak in the load-carrying leg bar. The underactuated exoskeleton with only one joint is
condition due to the lack of active actuators. The traditional actuated generally, so the ability of power assistance is
underactuated exoskeleton with only one joint actuated generally limited. A new concept of soft exoskeleton using the Bowden
has limited power assistance. As for soft exoskeleton, cable to drive the ankle joint was proposed by Harvard [4].
uncomfortable feeling resulting from the pressure generated by And it allows only 15% of the total torque required by the
cable tension on muscles are unavoidable. To compensate these joints to be delivered [5].Nonetheless, when a large tension is
drawbacks, this paper presents a lower limb exoskeleton based provided by the cable, uncomfortable feelings generated by
on the cable-pulley underactuated principle. By studying the the cable pressure on muscles will occur. This limits the
synergies of torque and power assistance at the joints and
designing the control strategy, a single cable will be able to drive
assistance performance of the soft exoskeleton.
the hip joint and the knee joint simultaneously, which can Evidence from [6] suggests that the cable-pulley
reduce the number of actuators. Considering the human- mechanism is suitable for low inertia and lightweight robotic
exoskeleton interactive dynamics, a biological method is utilized devices. And the literature [7] shows that the cable-pulley
to evaluate the performance of the exoskeleton. Finally, the mechanism applied in the field of underactuated dexterous
contrast simulation results demonstrate that the exoskeleton hand enables a single motor to drive multiple joints movement
proposed in this paper can effectively provide power simultaneously. It increases the ability of the dexterous hand
augmentation in load-carrying walking without interfering the to adapt to different shapes of objects and reduces the number
human motion and the control strategy is feasible. of actuators and the difficulty of control. This underactuated
concept is different from the traditional underactuated
Index Terms - lower limb exoskeleton, underactuated
exoskeleton concept that a single actuator drives a single joint
mechanism, dynamics, simulation
generally [8]. This approach takes into account both the small
I. INTRODUCTION number of actuators and multiple joints power assistance. And
the successful application in the dexterity hand field provides
Lower limb exoskeleton should be in harmony with us with a new idea to design an exoskeleton which has small
human movement, and it is served as an assistance device to number of actuators with multiple joints actuated by a single
provide power assistance for load-carrying condition. For the cable.
traditional fully actuated lower limb exoskeleton, there are To compensate these drawbacks of the exoskeleton
number of powerful actuators in joints, which allows the user reviewed above, this paper proposes a lower limb exoskeleton
to lift a heavy object. However, the energy consumption is based on the cable-pulley underactuated principle. We focus
large, which brings up the challenge of battery energy storage on the research of synergies of torque and power assistance of
technology or other large energy storage methods [1]. coupled joints. The hip joint and the knee joint can be actuated
Moreover, the inertia force generated by exoskeleton is large simultaneously by only a single actuator in this paper. Firstly,
in the motion, which limits the velocity of the exoskeleton. As the model with coupled joints of the lower limb exoskeleton
for the passive and quasi-passive exoskeleton, the load force is was modeled. Then the dynamics of the mechanism was
transferred to the ground by the exoskeleton structure, which analyzed by the principle of virtual work. Following that, the
reduces the pressure generated by load on human body. At the control strategy of synergies of torque and power assistance in
same time, by adding elastic energy storage components in the the joints was exploited. Finally, the contrast simulation of
joints, the bio-energy economy is improved. However, in the load-carrying walking without exoskeleton and with
experiment, the consumption of oxygen of the wearer exoskeleton indicates that the proposed exoskeleton could
increased by 10% than the condition without exoskeleton.

978-1-5090-6759-6/17/$31.00 ©2017 IEEE 577


effectively provide power augmentation in load-carrying can be measured by the tension sensor. In order to detect
walking. human motion intentions, there are encoders mounted on the
joints and pressure sensors mounted on the foot. These sensors
II. MECHANICAL DESIGN OF LOWER LIMB EXOSKELETON
are combined to detect the current gait cycle percentage to
A. Structure of lower limb exoskeleton provide real-time control of cable tension.
The mechanical structure of the proposed lower limb
exoskeleton is shown in Fig.1. There are thirteen DOFs in the
exoskeleton shown by the dash-dotted lines in Fig.1 with two
DOFs at the hip joint, one DOF at the knee joint, three DOFs
at the ankle joint, and one rotational DOF about the spine.
Among them, the hip joint and the knee joint are actuated
simultaneously by only a single actuator per leg in sagittal
plane. And the other DOFs are actuated passively. The range
of the exoskeleton motion is the same as the range of human
motion and the stopper mechanism is designed to ensure the
safety of the human body. Moreover, the lengths of the thigh
and shank rods of the exoskeleton are adjustable for different
individuals.
Fig. 2 Winding method of the cable-pulley mechanism

B. Principle of cable-pulley udneractuated mechanism


Due to the transmission of the cable, there are some
limiations between the hip joint and the knee joint. Actually,
cable-pulley underactuated mechanism is an evolution of the
metamorphic mechanism [9]. The traditional mechanism has a
fixed topology, with a fixed number of components and
DOFs. In contrast to that, the mobility of the metamorphic
mechanism can be changed in the process of continuous
motion. In other words, the remaining DOFs of metamorphic
mechanism can still be able to motion even though some
DOFs are constrained. For example, when the knee joint
reaches the extreme position, the rest DOFs of the hip joint
can still be in involved in the motion. This characteristic of the
metamorphic mechanism makes it possible to adapt to the
shape and size changes of the objects.
For the lower limb exoskeleton based on the principle of
the cable-pulley underactuated mechanism, the human lower
limb corresponds to the object whose shape and size is
continuously changing. In the motion process, exoskeleton
Fig. 1 Exoskeleton mechanical structure and distribution of degrees of
Freedom (The hip and knee joints have a similar structure ) constantly adapts to the shape and size changes of the human
The thigh rod and the hip pulley rotate about the same lower limb. Once appropriate travel distance of the cable is
axis independently. Similary, the shank rod and the knee given, the exoskeleton motion of the hip joint and the knee
pulley can also rotate about another axis independently. The joint can be distributed by the human motion during the
winding method of the cable is shown in Fig.2. The cable is adaptive process. Thus, the motion synergy of the hip joint
first wound several circles around the drive pulley to ensure and the knee joint is achieved.
the enough elongation of the cable and one end of the cable is At the hip joint and the knee joint, flexor or extensor
fixed on the driving pulley. Then the cable is wound on the torque can be provided. Therefore, there are a total of 22=4
right side of the hip pulley. After being wound one circle combinations, and the combination shown in Fig. 2 is selected
around the hip pulley, the cable is wound on the right side of in this paper. If a suitable time window is selected to provide
the knee pulley similarly. Finally, after being wound one torque assistance for the joints, the synergies of torque and
circle around the knee pulley, the other end of the cable is motion in the coupled joints can be achieved [10]. And a
fixed on the shank rod. single cable is able to drive multiple joints simultaneously.
To achieve the feedback control of cable tension and III. DYNAMIC ANALYSIS OF THE ROBOTIC LEG
position, the force sensors and position sensors are needed. An
encoder is mounted on the end of the motor shaft to measure A. Analysis of the motion constraint of the cable-pulley
the motor speed. And the travel distance of the cable can be mechanism
obtained by the relation calculation of cable drive transmission. The hip joint and the knee joint are actuated by a single
The tension sensor is mounted on the cable, so that the tension cable in Fig. 2-a. Let xs be the travel distance of the cable.

578
  The schematic of the two-link mechanism and the cable-
xs and xs represent the moving speed of the cable and the pulley mechanism system is shown in Fig. 3.Once the driving
moving acceleration of the cable respectively. Generalized tension of the cable is applied in the system, the shape and
coordinates q1 and q2 are rotational angles of the hip joint and size of the cable-pulley underactuated mechanism will change.
the knee joint respectively. rhip is the hip pulley radius. rknee is xs is the travel distance of the cable outside the system. S is
the knee pulley radius. The relation between the travel the remaining length of the cable inside the system. The
distance of the cable and the rotational angles of the joints is relation among the cable travel distance, the cable tension, the
angle and the generalized torque of the joint can be obtained
shown in Fig.2-b. The travel distance xs can be written as:
using formula (4) of the principle of virtual work.
x  AB  AC  CD
s
  DE (1)  W  T  xs    q (4)
In formula (1), AC  0 ( d1 is a constant), DE  0 ( d 2 is a Due to the total length of the cable is a constant, the
constant),     q r . The travel distance x of the
AB  q1rhip , CD 2 knee s
relation between xs and S : xs  S  cons tan t is obtain.
cable is only related to the contact arc length of the cable and Based on the above analysis, we have:
the pulley. The relation among travel distance xs , moving xs S
 0 (5)
  q q
speed xs and moving acceleration xs of the cable is obtained.
 xs xs
 xs  q1rhip  q2 rknee  (6)
  q q
 
 Substituting the formula (5), (6) into (4), the relation
 x s  q1 rhip  q 2 rknee (2)
  among the cable travel distance, the cable tension, the
 
 x s  q1 rhip  q 2 rknee generalized coordinates and the generalized torque of the joint
 can be derived.
B. Dynamic model of the robotic leg x S
  ( s )T T  ( )T T  J S (q)T T (7)
The dynamic equation of exoskeleton leg is established q q
by the Lagrange method. The general dynamic equation is
obtained, IV. LOWER LIMB EXOSKELETON CONTROL STRATEGY
  
D ( q ) q  H ( q, q ) q  G ( q )   (3)
A. Time window selection
T
where q  [q1 , q2 ]T and    1 , 2  . q  R 21 is a generalized A single motor is used to drive the two coupled joints
simultaneously. To achieve the torque synergy and the joint
coordinates vector of rotation angles of the hip joint and the
motion compatibility, a suitable time window should be
knee joint.   R 21 is a generalized torque vector.  1 and selected to provide torque assistance for the joints. And the
2 represent the driving torque at the hip and the knee goal is to maximize the torque and power augmentation by
selecting a suitable time window.
generated by the cable tension T respectively. In the case of Firstly, only flexor or extensor torque of the joint can be
ignoring the friction between the cable and the pulley, the provided by the cable. Whether flexor torque or extensor
tension of cable is equal in every part. D(q)  R 2 2 is a inertia torque is determined by the winding method of the cable.
 Thus, once the cable winding method is determined, the
matrix. H (q, q)  R 22 is a centripetal force and the Coriolis
torque polarity of the hip joint and the knee joint is required
force matrix. G (q)  R 21 is a gravitational torque vector. unchangeable. Secondly, the positive power is required to
provide power assistance. The relation between the cable
tension and the power can be obtained by the formula (8).
From the formula (8), it indicates that when the direction of
the torque generated by the cable coincides with direction of
the link angular velocity, the positive power will be generated
by the cable tension.
P  T v  T r     (8)
The torque peak and the power of the joints are two
important indicators to evaluate the power augmentation
effect. The time window should be selected to cover one or
several torque peaks and a large percentage of the gait cycle.
The time window should also cover the concentrated period of
Fig. 3 Schematic diagram of the two-link mechanism and the cable-pulley
positive power. Under the condition of human load-carrying
mechanism system( xs is the travel distance of the cable outside the system.
walking with 50kg without exoskeleton, the torque curves of
S is the remaining length of the cable inside the system) the human hip and the knee joint of two cycles are shown in
Fig.4.And the hip and knee joints power curves are shown in

579
Fig.5. The time window is marked with several rectangles. joint torque will be proportionally changed. Since both the
Among the time windows in Fig.4, the second and the fourth energy consumption and the torque peak of the hip joint is
time windows account for a large percentage of gait cycle, and larger than the knee joint. The hip torque curve is selected as a
in these windows, the torque directions of the hip joint and the reference to calculate the command torque. The relation
knee joint are unchangeable. Moreover, the Fig.5 illustrates between the command torque curves and human joint torque
that in the second window and the fourth window, large power curves are determined by equations (10) and (11). The
is consumed by the hip. And the second and the fourth time designed command torque curves are shown in Fig.4.
windows cover most of the positive power period of the knee.  command _ hip _ torque     hip _ torque (10)
Consequently, in this article, the second and the fourth time
 command _ knee _ torque    hip _ torque (11)
windows are selected to provide power assistance.
The tension signal can be obtained by the tension sensor.
The driving torque design of the joint can be eventually
converted to the force trajectory design of the cable using the
formula(12).
 command _ hip _ torque  command _ knee _ torque
Freference   (12)
rhip rknee

C. Controller design
Admittance controller is suitable for the small stiffness
environment of the cable transmission. Thus, the admittance
Fig.4 The hip and knee joints torque curves of human load-carrying walking controller based on the position control is applied. The
with 50kg without exoskeleton(The red solid curve represents the hip torque admittance controller structure is shown in Fig.6. The inner
curve. The yellow solid curve represents the knee torque curve. The black loop of the controller structure is the position feedback loop
solid curve represents the command hip torque curve. The blue solid curve
represents the command hip torque curve.)
and the outer loop is the force feedback loop. The designed
force trajectory is pre-stored in the controller. The foot
pressure sensor and the joint encoders are used to detect the
current gait cycle percentage to achieve real-time control of
the force.
The cable tension signal Fexoskeleton is obtained as a
feedback signal by the tension sensor, and the designed
tension curve is used as the reference tension Freference .The
force error F can be calculated by the formula (13), which is
used as the admittance controller input signal.
F  Freference  Fexoskeleton (13)
Fig.5 The hip and knee joints power curves of human load-carrying walking The relation between the force error and the desired
with 50kg without exoskeleton(The red solid curve represents the hip power
curve. The yellow solid curve represents the knee power curve.) position xs _ desired is:
B. Force trajectory design  
M d x s _ desired  Dd x s _ desired  K d xs _ desired  F (14)
The walking process of human body lower limb is a
periodic rhythmic motion, and the joint torque is periodically The transfer function of the admittance controller is
changing. The driving torque generated by the cable is derived,
delivered to the human lower limb to assist the human motion. 1
The designed transmission system allows the   50% total Gd ( s )  2 (15)
M d s  Dd s  K d
torque to be delivered to the human hip and knee joints.
M d is the virtual inertia. Dd is the virtual damping and K d is
The radiuses of the hip pulley and the knee pulley are
designed independently because of the different motion the virtual stiffness.
characteristics. Since the torque peak is an important indicator
to evaluate the power augmentation effect. According to the
torque peak ratio of the hip joint and the knee joint, the pulley
radius ratio  of the hip joint and the knee joint is determined
by formula(9).
r 
  hip  hip _ torque _ peak (9) Fig.6 The structure diagram of admittance controller
rknee  knee _ torque _ peak
V. SIMULATION
The cable tension is nearly equal in every part of the
cable. When the cable tension changes, the hip and the knee A. Biological modeling

580
In this paper, a nineteen-link human model was
established. There are 20 triaxial joints and 60 DOFs
(regardless of hand DOFs) in the model. The number and the
distribution of the degrees of freedom is very closer to the
human. And the biomechanics software LifeModeler is used
to establish the human model of a 65 kg, 20-year-old and 175
cm man. The towing model of human is shown in Fig.7. When
the inverse dynamic simulation is performed, the attachment
points are towed by the towing points. And the motion of the
human model is determined by the towing points motion
spline generated by the motion data captured by the 3D
motion analysis system. During the inverse dynamic process,
the torque changing process is recorded by the joints. When Fig.8 The walking simulation animation of load-carrying model without
the direct dynamic simulation is performed, the torque exoskeleton(The contact force is marked with the red arrows.)
changing process is repeated by the joints.
C. Load-carrying simulation with exoskeleton
Considering the human-exoskeleton interactive dynamics,
the bushing force (force and torque in three dimensions in
ADAMS) is used as the connecting force between the human
model and the exoskeleton. The command force is applied in
the exoskeleton to assist the human load-carrying walking.
And the control strategy introduced above is applied in this
simulation. The walking simulation animation of is shown in
Fig.9. Under the condition of load-carrying model with
exoskeleton, the torque peak of human hip joint is -70.81 Nm
shown by the dotted curve in Fig.10. The power of the human
hip joint is shown by the dotted curve in Fig.11.The torque
peak of the human knee joint is -51.62 Nm shown by the
dotted curve in Fig.12. The power of the human knee joint is
shown by the dotted curve in Fig.13. The contrast results of
the torque peaks are shown in TABLE I.

Fig.7 Towing model of human in LifeModer(The red points represent the


attachment points. The yellow points represent the towing points)

B. Load-carrying simulation without exoskeleton


Based on the biological human model, the 50 kg -load-
carrying model without exoskeleton is established in the
ADAMS. The contact force between the legs and ground
produces a normal force and a transverse friction force based
on parameters input by the user. With the reference of friction
force of the normal level walking in actual environment in
literature [11], [12], the static coefficient is set to 0.8 and the Fig.9 The walking simulation animation of load-carrying model with
dynamic coefficient is set to 0.25. The contact force is marked exoskeleton(The contact force is marked with the red arrows.)
with the red arrows in Fig.8. The direct dynamic simulation is
performed. The walking simulation animation of the load- TABLE I. RESULTS OF THE CONTRAST SIMULATION

carrying model without exoskeleton is shown in Fig.8. Under


Without With Reduce
the condition of load-carrying walking without exoskeleton, Joint Parameters
exoskeleton exoskeleton proportion
the torque peak of human hip joint is 103.24 Nm shown by the
solid curve in Fig.10. The power of human hip joint is shown Hip peak torque(Nm) 103.24 -70.81 31.4%
by the solid curve in Fig.11.The torque peak of human knee Knee peak torque(Nm) -76.01 -51.62 32.1%
joint is -76 Nm shown by the solid curve in Fig.12. The power
of human knee joint is shown by the solid curve in Fig.13.

581
Fig.15 Knee angle curve without exoskeleton and with exoskeleton contrast
Fig.10 Hip torque curve without exoskeleton and with exoskeleton contrast
VI. DISCUSSION
From the whole gait cycle, comparing the condition of
load-carrying walking with exoskeleton to the condition of
load-carrying walking without exoskeleton, the torque peaks
and the power of the human joints are both significantly
reduced. Moreover, the angle curves of the human joints
shown in Fig.14 and Fig.15 are almost coincident in the two
conditions. It demonstrates that the human walking was not
interfered by the exoskeleton.
Fig.11 Hip power curve without exoskeleton and with exoskeleton contrast In other hands, the interactive dynamics between the
human body and the exoskeleton should be considered in the
power augmentation evaluation [13]. In the literature [13], a
human walking gait simulation was made through the
establishment of neuromuscular walking model. The model
was established with only 19 DOFs to reduce the complexity
of the human model. Only the DOFs in the sagittal plane were
actuated. The accuracy of the simulation was possibly reduced
by this simplification. In this paper, a biological model whose
number and distribution of the degrees of freedom is very
Fig.12 Knee torque curve without exoskeleton and with exoskeleton contrast
closer to the human was applied in the simulation.
As a consequence, the results indicate that the
exoskeleton proposed can effectively provide power
augmentation in load-carrying walking without interfering the
human motion.
VII. CONCLUSION
In this paper, a cable-pulley underactuated principle-
based lower limb exoskeleton is proposed. A single cable is
able to drive the hip joint and the knee joint simultaneously
using synergies of torque and power assistance, which reduces
Fig.13 Knee power curve without exoskeleton and with exoskeleton contrast the number of actuators. Moreover, the control strategy of
exoskeleton is designed by selecting a suitable time window,
which can achieve the synergies of torque and joint motion of
coupled joints. The human-exoskeleton interactive dynamics
is considered through the biological method. The biological
method is suitable to evaluate the performance of the
exoskeleton. In the simulation, the torque and power required
by level walking are both reduced by the assistance of the
exoskeleton. The simulation results demonstrate that cable-
pulley underactuated principle-based lower limb exoskeleton
Fig.14 Hip angle curve without exoskeleton and with exoskeleton contrast can effectively provide power augmentation in load-carrying
walking without interfering the human motion and the control
strategy is feasible.
In future, the prototype of the designed exoskeleton will be
built and applied in the rehabilitation of human lower limb.
ACKNOWLEDGMENT

582
This work is supported by Resrach supported by National
Key Technology Support Program of China under Grant
2015BAF01B03.
REFERENCES
[1] Walsh, C. J., Pasch, K., & Herr, H. (2006, October). An autonomous,
underactuated exoskeleton for load-carrying augmentation. In Intelligent
Robots and Systems, 2006 IEEE/RSJ International Conference on (pp.
1410-1415). IEEE.
[2] Walsh, C. J., Paluska, D., Pasch, K., Grand, W., Valiente, A., & Herr, H.
(2006, May). Development of a lightweight, underactuated exoskeleton
for load-carrying augmentation. In Robotics and Automation, 2006.
ICRA 2006. Proceedings 2006 IEEE International Conference on (pp.
3485-3491). IEEE.
[3] Yu, S., Lee, H., Kim, W., & Han, C. (2016). Development of an
underactuated exoskeleton for effective walking and load-carrying assist.
Advanced Robotics(8).
[4] Asbeck, A. T., Dyer, R. J., Larusson, A. F., & Walsh, C. J. (2013, June).
Biologically-inspired soft exosuit. In Rehabilitation robotics (ICORR),
2013 IEEE international conference on (pp. 1-8). IEEE.
[5] Ding, Y., Galiana, I., Asbeck, A., Quinlivan, B., De Rossi, S. M. M., &
Walsh, C. (2014, May). Multi-joint actuation platformula for lower limb
soft exosuits. In Robotics and automation (ICRA), 2014 IEEE
international conference on (pp. 1327-1334). IEEE.
[6] Palli, G., Pan, L., Hosseini, M., Moriello, L., & Melchiorri, C. (2015,
May). Feedback linearization of variable stiffness joints based on twisted
string actuators. In Robotics and Automation (ICRA), 2015 IEEE
International Conference on (pp. 2742-2747). IEEE.
[7] Birglen, L., & Gosselin, C. M. (2006). Geometric design of three-phalanx
underactuated fingers. Journal of Mechanical Design, 128(2), 356-364.
[8] Birglen, L., & Gosselin, C. M. (2006). Grasp-state plane analysis of two-
phalanx underactuated fingers. Mechanism and Machine Theory, 41(7),
807-822.
[9] Dai, J. S., & Jones, J. R. (1999). Mobility in Metamorphic Mechanisms of
Foldable/Erectable Kinds. Journal of Mechanical Design, 121(3), 375-
382.
[10]Bartenbach, V., Schmidt, K., Naef, M., Wyss, D., & Riener, R. (2015,
August). Concept of a soft exosuit for the support of leg function in
rehabilitation. In Rehabilitation Robotics (ICORR), 2015 IEEE
International Conference on (pp. 125-130). IEEE.
[11]Li, K. W., Chang, W. R., Leamon, T. B., & Chen, C. J. (2004). Floor
slipperiness measurement: friction coefficient, roughness of floors, and
subjective perception under spillage conditions. Safety Science, 42(6),
547-565.
[12]Swensen, E. E., Purswell, J. L., Schlegel, R. E., & Stanevich, R. L.
(1992). Coefficient of friction and subjective assessment of slippery work
surfaces. Human factors, 34(1), 67-77.
[13]Lim, B., Hyung, S., Kim, K., Lee, J., Jang, J., & Shim, Y. (2016,
October). Simulating gait assistance of a hip exoskeleton: Feasibility
studies for ankle muscle weaknesses. In Intelligent Robots and Systems
(IROS), 2016 IEEE/RSJ International Conference on (pp. 5664-5669).
IEEE.

583

You might also like