See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/258253986

System characterization of RiceWrist-S: A forearm-wrist exoskeleton for upper

extremity rehabilitation

Article in IEEE International Conference on Rehabilitation Robotics : [proceedings] · June 2013

DOI: 10.1109/ICORR.2013.6650462 · Source: PubMed

CITATIONS READS

15 318

3 authors:

Ali Utku Pehlivan Chad Rose

Rice University Auburn University
22 PUBLICATIONS 955 CITATIONS 39 PUBLICATIONS 407 CITATIONS

SEE PROFILE SEE PROFILE

Marcia Kilchenman O'Malley

Rice University
292 PUBLICATIONS 6,751 CITATIONS

SEE PROFILE

All content following this page was uploaded by Marcia Kilchenman O'Malley on 14 March 2014.

The user has requested enhancement of the downloaded file.

2013 IEEE International Conference on Rehabilitation Robotics June 24-26, 2013 Seattle, Washington USA

System Characterization of RiceWrist-S: a Forearm-Wrist Exoskeleton

For Upper Extremity Rehabilitation
Ali Utku Pehlivan, Student Member, IEEE, Chad Rose, Student Member, IEEE,
Marcia K. O’Malley, Member, IEEE

Abstract— Rehabilitation of the distal joints of the upper

extremities is crucial to restore the ability to perform activities
of daily living to patients with neurological lesions resulting
from stroke or spinal cord injury. Robotic rehabilitation has
been identified as a promising new solution, however, much
of the existing technology in this field is focused on the more
proximal joints of the upper arm. A recently presented device,
the RiceWrist-S, focuses on the rehabilitation of the forearm
and wrist, and has undergone a few important design changes.
This paper first addresses the design improvements achieved
in the recent design iteration, and then presents the system
characterization of the new device. We show that the RiceWrist-
S has capabilities beyond other existing devices, and exhibits
favorable system characteristics as a rehabilitation device, in
particular torque output, range of motion, closed loop position
performance, and high spatial resolution.
Index Terms— Exoskeletons, serial mechanisms, haptic inter-
face design, stroke rehabilitation, spinal cord injury rehabilita- Fig. 1. RiceWrist-S – Forearm and wrist exoskeleton for stroke and spinal
tion. cord injury (SCI) rehabilitation.

I. I NTRODUCTION
In the United States, about 795,000 people suffer a stroke
each year. Stroke, the leading cause of long-term disability,
has a significant social and economic impact with an esti- In order to facilitate effective rehabilitation of the wrist,
mated $38.6 billion yearly cost [1]. There are approximately an exoskeleton needs to possess: i) a functional workspace
12,000 incidences of Spinal Cord Injury (SCI) in the United that matches healthy human capabilities [5], ii) the ability
States each year [2] with an estimated total yearly direct and to apply torques to specific joints [6], and to quantitatively
indirect costs of $14.5 billion and $5.5 billion, respectively evaluate rehabilitation and pharmacological treatment effects
[3]. Due to the large need for rehabilitation and limitations [7], iii) good backdriveability and backlash-free operation
of classical rehabilitation techniques, interest in robotic re- [8], and iv) advanced control capabilities [9]. End-effector
habilitation has increased greatly in recent years. based robots, such as the MIT-MANUS [10] and Mirror
Clinical studies investigating robotic rehabilitation proto- Image Movement Enabler (MIME) [11] possess a large
cols support the idea of employing these devices in treatment functional workspace, but do not have the capability to
of stroke and SCI patients in order to further induce brain apply torques to specific joints or quantitatively evaluate
and spinal cord plasticity and improve patient outcomes. patients like exoskeleton based robots, such as the 5 DOF
Nearly all the activities of daily living (ADL), such as MAHI Exoskeleton [12], 5 DOF Rupert [13], 6 DOF ARMin
eating, drinking, cleaning, dressing, etc., involve distal upper [14] and 7 DOF CADEN-7 [15]. However, there is still a
extremity movement and a certain level of manual dexterity. need for devices for wrist rehabilitation, and therefore the
In order for a stroke or SCI patient to regain the ability RiceWrist-S (Fig.1), first presented in detail in [16] (Fig.
to perform ADL, effective rehabilitation of the upper limbs, 2(b)), was designed. This first iteration of the RiceWrist-
especially the distal joints, is required. This need is very S mainly addresses the design shortcomings of RiceWrist
acute, as it is common for as little as 12% of stroke patients [17], a serial-in-parallel mechanism, which was designed by
to report no difficulty using their affected hand and 38% to our group. In this paper, we will present the mechanical
report major difficulty 3 months after their stroke [4]. design modifications carried out over previous design of
RiceWrist-S and show the system characterization of this
This project was supported in part by Mission Connect, a project of the new device (Fig. 2(c)). The improvements due to these design
TIRR Foundation, and by NSF grant CNS-1135916. The authors are in the modifications will be presented via comparing the system
Mechatronics and Haptic Interfaces Laboratory, Department of Mechanical
Engineering and Materials Science, Rice University, Houston, TX 77005 with the first iteration [16]. The system characteristics will
(e-mails: [email protected], [email protected], [email protected]) also be compared with a same purpose device [18].

978-1-4673-6024-1/13/$31.00 ©2013 IEEE

 (a) (b) (c)

Fig. 2. (a) Kinematic structure of RiceWrist-S, a 3-DOF serial RRR mechanism. (b) The previous design. (c) The new design introduces considerable
design improvements over previous design.

II. D ESIGN D ETAILS

RiceWrist-S is a 3 DOF, electrically actuated, grounded
forearm-wrist exoskeleton. The system is a serial RRR
manipulator, the kinematic structure of which is depicted
in Fig. 2(a). RiceWrist-S is capable of actuating the user’s
forearm pronation/supination (PS), wrist flexion/extension
(FE), wrist radial/ulnar deviation (RU) DOFs separately. In
addition to the actuated DOFs, the system employs a passive
linear degree of freedom on the handle coupling in order to
keep the user’s wrist in an anatomically natural posture.
In order to ensure zero backlash and low friction on the
FE and RU joints, the RiceWrist-S utilizes cable drive trans-
missions, while the PS joint employs a frameless, brushless
motor with direct drive.
The mechanical design of the device presented in this
work introduces considerable design improvements in both Fig. 3. RU joint cable routing detailed demonstration.
RU and FE joint over the device introduced in [16]. The
modifications, rationale, and results of the modifications are
presented in detail below, grouped under RU and FE joint The main benefit of the cable routing system is that it
subsections. enables us to place the RU joint actuator exactly below the
FE rotation axis, decreasing the inertia of the mechanism
A. RU Joint considerably compared to the previous design.
Different from the previous design, the new exoskeleton In the previous design, the capstan arc was driven by the
employs a cable routing system to transmit the actuation on threaded spool which is attached to the motor shaft (Fig.
the RU joint. The actuation is transferred from the motor 2(b)). Because of the thickness of the motor shaft, the spool
shaft to the transmission shaft via two steel cables (see Fig. had to have a certain thickness and we could not achieve a
3). The method, described in [19], requires first winding larger transmission ratio than 1:12.5. In the new design we
cables in opposite directions, then fixing them on both the could specify the thickness of the transmission shaft inde-
motor shaft and the transmission shaft. Pretension has to be pendent from any factor, and achieved a 1:24 transmission
applied to both cables in order to achieve sufficient stiffness. ratio. The increase of the transmission ratio provided both
In order to pretension the cables, a precision threaded steel higher torque outputs and better sensor resolution (see Table
transmission rod was coupled with two threaded aluminum I and II). One potential problem regarding this design is the
cylinders. The cables, which are fixed on the motor shaft difficulty of cable installation.
from one end, are fixed to these two cylinders, rather than to
the transmission shaft itself. So, by screwing the cylinders B. FE Joint
away from each other, pretension can be applied to the The placement of the actuators is very important for
cables. Two nuts are used for each cylinder to prevent them keeping the inertia of the device as low as possible, in turn
from loosening. resulting in better backdrivability. In the previous design,
One end of the transmission shaft is used as a capstan we placed the FE actuator tangentially to the opening of
spool. The capstan arc, coupled to the device handle, is the PS actuator in order to keep the distance of the FE
driven by means of a cable drive system (Fig. 3). We used actuator as close as possible to the PS rotation axis. Another
a Maxon RE-30 brushed DC motor for the actuation and a consideration for the placement of the FE actuator was that
CPT Avago 5540 HEDS optical encoder with 500 counts per a cable drive system was employed for the transmission, and
revolution for position sensing at the RU joint (see Table I). the spool, which was attached to the motor shaft, had to be
 TABLE I
S ENSOR AND ACTUATOR S PECIFICATIONS

Joint Actuator Transmission Sensor Sensor Resolution

Forearm Pronation/Supination Applimotion 165-A-18 Direct-Drive MicroE Mercury 1500 0.002◦
Wrist Flexion/Extension Maxon RE-40 (148877) Cable-Drive (1:18) Avago HEDS 5540 0.01◦
Wrist Radial/Ulnar Dev. Maxon RE-30 (310009) Cable-Drive (1:24) Avago HEDS 5540 0.0075◦

TABLE II
ACHIEVABLE JOINT RANGES OF MOTION (ROM) AND MAXIMUM CONTINUOUS JOINT TORQUE OUTPUT VALUES FOR R ICE W RIST-S.T HE VALUES
GIVEN IN PARENTHESIS IN THE TORQUE COLUMN SHOWS THE CAPABILITY OF THE OLD DESIGN . T HE REQUIRED ROM AND TORQUE VALUES FOR 19
(ADL) AS EXTRACTED FROM [15] ARE ALSO GIVEN FOR COMPARISON .

ADL RiceWrist-S
Joint
ROM (deg) Torque (Nm) ROM (deg) Torque (Nm)
Forearm Pronation/Supination 150 0.06 180 1.69 (1.69)
Wrist Flexion/Extension 115 0.35 130 3.37 (2.80)
Wrist Radial/Ulnar Dev. 70 0.35 75 2.11 (1.10)

III. D EVICE M ODELING AND C HARACTERIZATION

In this section, we will examine the kinematics of the
RiceWrist-S, and present the experimentally determined
characterization of the device, including the spatial resolu-
tion, static friction, bandwidth, viscous friction coefficient
and inertial elements. This system characterization is made
in order to evaluate the device’s potential for rehabilitation
use.

(a) (b) A. Kinematics

Fig. 4. (a) The shaded motor is a representative demonstration of the The RiceWrist-S is a pure rotational manipulator, we are
approximate placement of the FE actuator in the previous design. The interested in the orientation of the user’s wrist-forearm. In the
actuator is moved from A to B, in order to install another support for the neutral position, which is shown in the Fig. 2(a), z0 , z1 and
device handle and get a more rigid device. (b) An idle pulley mechanism
is used to relocate the actuator. z2 coincides with the user’s PE, FE and RU rotation axes
respectively. Frame {1} is coincided with frame {0}, then
rotated − π2 radians around x0 , and the rotation θ1 around z0
corresponds to the user’s rotation around his/her PS rotation
axis. Similarly, frame {2} coincides with frame {1}, then
as close to the capstan arc as possible. These two constraints
rotated π2 radians around x1 , and the rotation θ2 around z1
decided the thickness of the spool, hence we could not apply
corresponds to user’s rotation around the FE rotation axis.
more than 1:15 transmission ration. Also, because the motor
Frame {3} coincides with frame {2} and the rotation around
was placed closely to the capstan arc, installation of a second
z2 corresponds to the RU rotation of the user. The Denavit-
support for the device handle was not possible, which was
Hartenberg parameters of the system can be given as follows:
affecting the rigidity of the device negatively. In the new
design, we kept the distance of the flexion/extension actuator TABLE III
same as the previous design, however, by employing an idle L INK PARAMETERS FOR THE F OREARM AND W RIST JOINTS
pulley mechanism (Fig. 4(b)) we could place the actuator
further back (Fig. 4(a)). By this way we could install another Joint rot(x) tr(x) rot(z) tr(z)
Forearm − π2 0 θ1 0
support for the device handle and get a more rigid device. Wrist F/E π
0 θ2 0
2
Also, the freedom of choosing the placement of the actuator Wrist R/U 0 0 θ3 0
enabled us to increase the transmission ratio to 1:18 (see
Table II). The corresponding transformation matrix from frame {3}
We utilized a Maxon RE-40 brushed DC motor with a to frame {0} is
CPT Avago 5540 HEDS optical encoder with 500 counts  
per revolution for actuation and sensing , and employed a c1c2c3 − s1s3 −c1c2s3 − s1c3 c1s2 0
steel cable for the cable drive transmission system. In order 0
 s1c2c3 + c1s3 −s1c2s3 + c1c3 s1s2 0 
T3 =   (1)
to prevent any slippage, the cable is wrapped around both  −s2c3 s2s3 c2 0 
the idle pulley and the spool three times. 0 0 0 1
 In the above equation, while s and c stands for sine and 0.015

cosine, 1,2 and 3 stand for θ1 , θ2 and θ3 , respectively.

0.01
The Jacobian relating the link velocities to the end effector
angular velocities is given as

Torque [Nm]
0.005

 
0 −s1 c1s2 0

J =  0 c1 s1s2  (2)
−0.005
1 0 c2
−0.01
The kinematic structure shows that, singularity occurs 0 10 20 30 40
Time [s]
50 60 70 80 90

when z0 and z2 are aligned, i.e., when θ2 =± π2 . The θ2 (which

(a)
corresponds to FE rotation) mechanically constrained to be
between ±65◦ (1.13 rad) (see Table II), hence the device is
singularity free. 4
3.5

B. Spatial Resolution 3

Velocity [deg/s]
2.5
We calculated the spatial resolution of the RiceWrist-S 2
by using the sensor resolutions and the Jacobian of the 1.5
device. Any instantaneous change in the joint space causes 1

an instantaneous change in the task space which are related 0.5

0
by the device’s Jacobian. In order to quantify the spatial 5.85 5.9 5.95 6 6.05 6.1
resolution, we created approximately 35,000 scenarios in Time [s]

which to compute the effect of the smallest instantaneous (b)

change detectable in the joint space on the end effector. The
Fig. 5. (a) The commanded torque input to the RU joint actuator at zero
worst-case result of these scenarios in any DOF was a spatial velocity. The spikes occur at the movement initiation instances. (b) The close
resolution of 2.1816 ×10−4 radians. up view of RU joint velocity at one of the ramping up initiation instances.
The desired trajectory imposes movement initiation at 6th second, but the
C. Device Characterization movement starts after the static friction is overcame.

The friction and inertia characteristics of every joint of

1.5
the device are obtained by investigating simple ramp and
step inputs to the system in a similar way presented in [20].
1
A ramp position command which stays constant for 2
Position [radian]

seconds, and ramps up (or down) 5◦ (0.088 radians) in 2

0.5
seconds is sent to the system for each joint separately. The
command starts from 0◦ , goes up to 45◦ (25◦ for RU joint)
0
goes down to -45◦ (-25◦ for RU joint) and comes back
to 0◦ . In this way, we obtained static friction values for
−0.5
0 1 2 3 4 5 6 7
different locations. In order to eliminate gravitational effects Time [s]
as much as possible during the tests, the device was brought
to a configuration at which the axis of rotation of the joint Fig. 6. The response at the PS joint to a step position command with a
proportional control. The actuator is programmed to act as a spring with
being tested was parallel to the direction of gravity, and the spring constant of 54.4 Nm
rad .
handle of the device is fixed and the other two joints are
locked in the neutral position. The corresponding actuator
is programmed as a spring, and the forces applied at zero back to 0◦ (Fig. 6) is commanded. Table III presents the
velocity is recorded. Fig. 5(a) presents the torque values static friction coefficients, viscous damping coefficient and
commanded to the RU actuator while the velocity is zero. inertial element values.
The spikes occur at the instant movement is initiated, when
the device overcomes static friction. The maximum static D. Closed Loop Position Bandwidth
friction values are presented in Table IV. The values are less Depending on the intended control strategy, the control
than 13% of the continuous torque output values for every implementations of robotic devices might employ closed
joint. loop position control. In order to examine the position control
The dynamical properties of the device were determined performance of the device we identified the closed loop
by investigating the response of the system to a step position position bandwidth by tracking a sine position input with
command. We adopted the logarithmic decrement method PD controller for every DOF separately. The amplitude of the
presented in [21], which isolates the inertial and viscous input signal was set to 10◦ (0.175 rad) and the frequency was
effects responsible for the exponential decay of the free increased gradually. The Bode plots for every DOF is shown
vibration of the system. A position input which steps up in Fig. 7. We observed approximately 3.6 Hz, 6 Hz ,and 8.3
from 0◦ to 25◦ (0.436 radians) (45◦ for PS joint), and steps HZ bandwidths for PS, FE, and RU DOF respectively. These
 TABLE IV
D EVICE C HARACTERISTICS . T HE VALUES IN THE PARENTHESIS IN THE FIRST TWO COLUMNS ARE THE RELATED VALUES FOR THE DEVICE
DESCRIBED IN [18], AND THEY ARE GIVEN FOR COMPARISON .

Static Friction Inertia Viscous Coeff. CL Position Bandwidth

Joint
(Nm) (kg.m2 ) ( Nm.s
rad ) (Hz)
Forearm Pronation/Supination 0.221 (0.29) 0.157 (0.0058) 0.428 3.5
Wrist Flexion/Extension 0.198 (0.075) 0.0054 (0.0040) 0.085 6
Wrist Abduction/Adduction 0.211 (0.075) 0.0048 (0.0031) 0.135 8.3

2013 update: A report from the American Heart Association,” Circu-

RU Joint
2
FE Joint lation, vol. 127, p. e8, 2013.
PS Joint
0
[2] Anon., “Spinal cord injury facts and figures at a glance,” National
Spinal Cord Injury Statistical Center, February 2012.
−2
[3] M. Berkowitz, Spinal cord injury: An analysis of medical and social
Magnitude [dB]

−4 costs. Demos Medical Pub, 1998.

−6
[4] P. Duncan, R. Bode, S. Min Lai, S. Perera et al., “Rasch analysis of a
new stroke-specific outcome scale: the stroke impact scale1,” Archives
−8 of physical medicine and rehabilitation, vol. 84, no. 7, pp. 950–963,
−10
2003.
[5] A. Schiele and F. C. T. van der Helm, “Kinematic design to improve
−12
ergonomics in human machine interaction,” IEEE Transactions on
0 1 2 3
Frequency [Hz]
4 5 6 Neural Systems and Rehabilitation Engineering, vol. 14, no. 4, pp.
456–469, 2006.
[6] J. Klein, S. Spencer, and D. Reinkensmeyer, “Breaking it down is
Fig. 7. Bode plot for PS, FE and RU joints. The bandwidth values are 3.6 better: Haptic decomposition of complex movements aids in robot-
Hz, 6 Hz ,and 8.3 Hz respectively. assisted motor learning,” Neural Systems and Rehabilitation Engineer-
ing, IEEE Transactions on, vol. 20, no. 3, pp. 268–275, 2012.
[7] O. Celik, M. K. O’Malley, C. Boake, H. Levin, S. Fischer, and
T. Reistetter, “Comparison of robotic and clinical motor function
bandwidth values are comparable to the human movement improvement measures for sub-acute stroke patients,” in Proc. IEEE
capability, which is desirable feature for robotic rehabilitation International Conference on Robotics and Automation (ICRA 2008),
devices. Pasadena, CA, USA, May 2008, pp. 2477–2482.
[8] V. Hayward and K. MacLean, “Do it yourself haptics: Part i,” Robotics
IV. D ISCUSSION & Automation Magazine, IEEE, vol. 14, no. 4, pp. 88–104, 2007.
[9] E. Wolbrecht, V. Chan, D. Reinkensmeyer, and J. Bobrow, “Optimiz-
In this paper, we presented the system characterization ing compliant, model-based robotic assistance to promote neurore-
of the RiceWrist-S, an exoskeleton for stroke and SCI habilitation,” Neural Systems and Rehabilitation Engineering, IEEE
Transactions on, vol. 16, no. 3, pp. 286–297, 2008.
patient rehabilitation. By modifying the previous design, [10] H. I. Krebs, N. Hogan, M. L. Aisen, and B. T. Volpe, “Robot-aided
the RiceWrist-S was able to provide higher torque output neurorehabilitation,” IEEE Trans. Rehab. Eng., vol. 6, no. 1, pp. 75–
capabilities, lower inertia and a more robust system than 87, 1998.
[11] P. S. Lum, C. G. Burgar, P. C. Shor, M. Majmundar, and M. Van der
the previous design. The presented system characterization Loos, “Robot-assisted movement training compared with conventional
of this novel device shows that the RiceWrist-S has the therapy techniques for the rehabilitation of upper-limb motor function
potential to conduct state-of-the-art rehabilitation regimens. after stroke.” Arch Phys Med Rehabil, vol. 83, no. 7, pp. 952–9, 2002.
[12] A. Gupta and M. K. O’Malley, “Design of a haptic arm exoskeleton
The RiceWrist-S possesses the ability to cover more than the for training and rehabilitation,” IEEE/ASME Transactions on Mecha-
required workspace for ADL. The device provides compara- tronics, vol. 11, no. 3, pp. 280–289, 2006.
ble system characteristics to an existing, same purpose device [13] T. G. Sugar, J. He, E. J. Koeneman, J. B. Koeneman, R. Herman,
H. Huang, R. S. Schultz, D. E. Herring, J. Wanberg, S. Balasubrama-
[18]. Although the static friction values are relatively higher nian, P. Swenson, and J. A. Ward, “Design and control of RUPERT:
for FE and RU joints, the high continuous torque values that a device for robotic upper extremity repetitive therapy,” IEEE Trans.
device can provide, enables to overcome in software the high Neural Syst. Rehab. Eng., vol. 15, no. 3, pp. 336–346, 2007.
[14] T. Nef, M. Mihelj, G. Kiefer, C. Perndl, R. Muller, and R. Riener,
friction. Additionally, these high torque capabilities ,above “ARMin-exoskeleton for arm therapy in stroke patients,” in IEEE 10th
those required for matching the values required for ADL, International Conference on Rehabilitation Robotics (ICORR 2007).
enables us to implement a wide range of control algorithms IEEE, 2008, pp. 68–74.
[15] J. C. Perry, J. Rosen, and S. Burns, “Upper-limb powered exoskeleton
(both assistive and resistive control algorithms). design,” IEEE/ASME Transactions on Mechatronics, vol. 12, no. 4,
The kinematic structure of the RiceWrist-S provides ad- pp. 408–417, 2007.
vantages in terms of application diversity. The RiceWrist- [16] A. Pehlivan, S. Lee, and M. O’Malley, “Mechanical design of
ricewrist-s: A forearm-wrist exoskeleton for stroke and spinal cord
S, by decoupling the wrist joints, enables implementation injury rehabilitation,” in Biomedical Robotics and Biomechatronics
of different movements compared to the stated device [18], (BioRob), 2012 4th IEEE RAS & EMBS International Conference on.
such as backdriving one of the wrist joints while locking the IEEE, 2012, pp. 1573–1578.
[17] M. OMalley, A. Sledd, A. Gupta, V. Patoglu, J. Huegel, and C. Burgar,
other wrist joint. “The ricewrist: A distal upper extremity rehabilitation robot for stroke
R EFERENCES therapy,” in ASME International Mechanical Engineering Congress
and Exposition, 2006.
[1] A. Go, D. Mozaffarian, V. Roger, E. Benjamin, J. Berry, W. Borden, [18] H. I. Krebs, B. T. Volpe, D. Williams, J. Celestino, S. K. Charles,
D. Bravata, S. Dai, E. Ford et al., “Heart disease and stroke statistics– D. Lynch, and N. Hogan, “Robot-aided neurorehabilitation: a robot for
 wrist rehabilitation,” Neural Systems and Rehabilitation Engineering,
IEEE Transactions on, vol. 15, no. 3, pp. 327–335, 2007.
[19] J. K. Salisbury Jr, W. T. Townsend, D. M. DiPietro, and B. S. Eberman,
“Compact cable transmission with cable differential,” Sep. 10 1991,
uS Patent 5,046,375.
[20] J. R. Celestino, “Characterization and control of a robot for wrist
rehabilitation,” Master’s thesis, Massachusetts Institute of Technology,
2003.
[21] J. Liang and B. Feeny, “Identifying coulomb and viscous friction from
free-vibration decrements,” Nonlinear Dynamics, vol. 16, pp. 337–347,
1998.

View publication stats