2017 International Conference on Rehabilitation Robotics (ICORR)

QEII Centre, London, UK, July 17-20, 2017.

MoBio: A 5 DOF Trans-humeral Robotic Prosthesis

R. Achintha M. Abayasiri, D.G. Kanishka Madusanka, N.M.P. Arachchige, A.T.S. Silva, R.A.R.C. Gopura

Abstract— In this paper, a 5 DOF trans-humeral robotic wrist. For example if a person writes a word using a trans-
prosthesis: MoBio is proposed. MoBio includes 2 DOF at humeral prosthesis with absent U/R deviation, shoulder and
wrist which is rare in other trans-humeral prostheses. Through elbow should take over the particular motion. This causes
anthropometric features MoBio prosthetic arm can achieve
elbow flexion/extension, forearm supination/pronation, wrist limiting the full use of the shoulder mobility and discom-
radial/ulnar deviation, wrist flexion/extension and compound fort to the amputee. Therefore it is important for a trans-
motion of thumb and index finger. An EMG based control humeral prosthesis to have both wrist motions. Even though
method which uses EMG signals of the biceps brachii and researches have been carried out to achieve these functions
triceps brachii, is used with a motion switching mechanism using parallel manipulators and motors [8] and using hybrid
to control the prosthesis. Experimental results have verified the
usability and effectiveness of MoBio in performing Activities of (Electrically powered+body powered) prostheses [6], they
Daily Living. have failed to achieve the shift between two axes of two
Index Terms— trans-humeral, prosthesis, electromyography wrist motions. Intersected two axes make the wrist motions
of the prosthesis deviating further from the natural motions.
I. I NTRODUCTION Moreover, when designing a prosthesis weight is a crucial
factor. Prostheses with over weight limit the ranges of mo-
Trans-humeral prostheses are used to replace the missing tions of the prosthesis and cause musculo-skeletal disorders.
upper limb segment after an amputation between shoulder Thus, the weight of the prosthesis should be similar or lesser
and elbow [1]. In order to make these prostheses more com- to the actual human upper limb. Though the fact remains as
plaisant for amputees, researchers are working on developing such, due to the components added to achieve the required
functional, anthropomorphic and dexterous prostheses. How- motions, prostheses tend to exceed the weight limit. Hence,
ever, these prostheses stay way back when compared to the keeping the co-relation between the functions of prosthesis
natural limbs: in the sense of functionality and anthropometry and the weight of the prosthesis has become a challenging
[2]. Therefore, developing active upper limb prostheses with research.
proper hardware design and human motion intention based As a solution, this paper proposes a 5 DOF prosthesis
control has become a widely researched area nowadays. named MoBio. It has novel 2 DOF wrist which has 20mm
Due to trans-humeral amputation, elbow, forearm, wrist axes shift between two wrist motions. Moreover, it can
and hand motions are lost. Therefore, it is expected to achieve full ranges of motions as of a human upper limb.
replace those lost motions as much as possible through Furthermore, MoBio is fabricated considering the weight of
a trans-humeral prosthesis. Among lost motions elbow the prosthesis so that its weight becomes closer to the actual
flexion/extension (F/E), forearm supination/pronation (S/P), arm. In order to evaluate the controllability of the prosthesis
wrist ulnar/radial (U/R) deviation, wrist F/E and finger an EMG based control algorithm with a motion switcher [9]
motions are very much important for the Activities of Daily is used. Control algorithm uses EMG signals of bicep brachii
Living (ADL) [3], [4]. Therefore, a trans-humeral prosthesis and triceps brachii for the controlling purpose. MoBio can be
should be able to perform these motions while having sequentially controlled according to the motion intention of
dexterity and anthropomorphic aspects. the user, by 1 DOF at a time through a switching mechanism.
It is evident that existing prostheses are rarely being The paper is structured as follows. Section II proposes
able to achieve required DOF of a trans-humeral prosthesis the design and actuation mechanisms of MoBio. Section III
while achieving dexterous hand motions [5]–[7]. Therefore, explains the prosthetic controller of the proposed prosthesis.
a prosthesis with simultaneously working major upper limb Section IV describes the experiments and results. Section V
motions while having the relevant anthropometric aspects, presents the conclusion of the paper.
is required to achieve human like motions. However, almost
all the available trans-humeral prostheses do not include two II. D ESIGN AND ACTUATION M ECHANISMS OF M O B IO
wrist motions at the same time [5]. If one of the motions is MoBio is a right arm trans-humeral prosthesis. In order
absent then the shoulder has to perform the motions for the to realize human like motions, MoBio is designed with 5
DOF: Elbow F/E, forearm S/P, wrist F/E, U/R deviation and
*This work was supported by Senate Research Council (SRC) grant (no.
SRC/LT/2013/07) compound motion of thumb and index finger. The prosthesis
R. Achintha M. Abayasiri, D.G. Kanishka Madusanka, N.M.P. is shown in Fig. 2(a). The prosthesis can be attached to the
Arachchige, A.T.S. Silva and R.A.R.C.Gopura are with the stump arm of the amputee via osseointegrated implant [10]
Bionics Laboratory, Deptartment of Mechanical Engineering,
University of Moratuwa. achinthamihiran@gmail.com, or through socket and straps. Considering the high strength
gopura@mech.mrt.ac.lk to weight ratio, Aluminum is selected to build most of the

978-1-5386-2296-4/17/$31.00 ©2017 IEEE 1627

 Bevel Gear pair (1:1)
Elbow Housing Forearm Motor Forearm base
Forearm Wheel plate U/R Motor U/R motor plate
Elbow Motor

L brackets

Forearm base U/R Deviation Hand base

Elbow Shaft holding Spur Gear Wheel Forearm bars Shaft
plate
Gear pair Internal Gear
(1:1) Motor with
Gear pair Wrist F/E
Wrist connecting screw
Harmonic (23:16) Motor
Forearm Connecting plate
Drive U bracket
Plates
Elbow Shaft
Force sensor

(a) (b) (c) (d) (e)

Fig. 1. (a) 3D Model of Elbow (b) 3D Model of Forearm Part 1 (c) 3D Model of Forearm Part 2 (d) 3D Model of Wrist (e) Inside of the Hand

TABLE I
D-H PARAMETERS OF M O B IO FROM ELBOW TO HAND

Link αi 1 ai 1 di θi
1 π/2 l1 0 θ1
2 π/2 0 l2 π/2 + θ2
3 π/2 lw l3 π/2 + θ3
(a) z4 Palm
4 π/2 lh 0 θ4
x4
lw y3
𝑙1 𝑙𝑓 x3 𝑙𝑒2
x1 z2 located at the two ends of the shaft. The forearm connecting
z1 𝑙3 𝑙𝑒1
plates (see Fig. 1(b)) rigidly connects with elbow shaft so
𝑙2 x0 x2
that the forearm assembly can rotate relative to the elbow
y0 assembly. Thus the elbow F/E motion is achieved.
Elbow (b)
Gear wheels of the elbow are fabricated using cast iron
Fig. 2. (a) MoBio Prosthetic arm (b) Kinematics model of MoBio since elbow requires high torques and applies high dynamic
tooth loads on the gear wheels.
components of the prosthesis. Fabricated prosthesis: MoBio
weighs 3.2kg which is same as the weight of actual human B. Forearm
arm [11]. Actual human arm’s weight of 75Kg human is The forearm design of MoBio is shown in Fig. 1(b) and
3Kg. MoBio has four major sections: elbow, forearm, wrist Fig. 1(c). A DC motor is used (DCX22S, Maxon motors)
and hand in its design. The kinematic model is shown in to achieve the forearm S/P motion. To transmit the output
Fig. 2(b) and table I shows the D-H parameters [12] of the power of the motor to the forearm, output shaft of the motor
prosthesis where l1 = 75.6mm, l2 = 16.8mm, l3 = 4.9mm, lf is connected to the internal gear with a spur gear. This
= 162.2mm, lw = 20.0mm, le1 = 69.5mm and le2 = 15.6mm. internal gear is rigidly fixed to the forearm base wheel to
enable the forearm S/P motion. In this design, forearm base
A. Elbow wheel lies inside the forearm wheel. In order to support the
The design of the elbow is shown in Fig. 1(a). A brush relative motion between two wheels a needle roller bearing
less DC motor (BLDC)(EC 4 Pole, Maxon motors) is used is used. Needle roller baring is press fit to the forearm wheel
to initiate the rotary motion of the elbow. BLDC is fixed and base wheel is press fit to the bearing. Forearm base
to the elbow housing. Elbow housing can be connected to a plate is connected to the base wheel. The forearm base plate
socket or to a osseointegrated implant. The BLDC motor is can rotate with the forearm base wheel and achieve relative
connected to a harmonic drive gear box (100:1) through a motion between two forearm parts [see Fig. 1(b) and 1(c)].
spur gear pair (ratio 1:1). The output of the harmonic drive For the purpose of achieving high load bearing ability to the
gear box is coupled to the elbow shaft through a spur gear forearm compared to previous researches [8], forearm bars
pair with gear ratio of 23:16. Elbow shaft can rotate relative have used to connect the forearm base plate to the wrist
to the Elbow shaft holding plates with the aid of bearings connecting plate.

1628
 20mm Axis shift
𝑅1, 𝑅2
N
U/R Deviation
N
250 270 𝑅𝑇 < 𝑅𝐸𝑋𝑃 𝑅T > 𝑅𝑀𝐴𝑋

Y Y

Proportional Controller Motion Switcher

Low Level Controller Prosthetic Arm

Fig. 4. EMG based controlling algorithm. RMS values of bicep brachii and
triceps brachii (R1 and R2 ) are processed to calculate their addition (RT )
in order to decide whether to move the prosthesis or to switch the current
motion. REXP and RM AX are experimentally achieved according to
700
amputee’s EMG RMS. In order to reduce discrepancies in motion switching
due to accident motions, RMS values between REXP and RM AX are
Wrist F/E neglected.
700

Fig. 3. Wrist design of MoBio. It can achieve 2 DOF wrist motions the amputation and signals are similar to that of a able-
while maintaining the 20mm axis shift between U/R deviation and wrist body person. The human can see the prosthesis in operation
F/E motions. Shape of the U/R connecting plate allows the wrist to achieve
full range of motion of U/R deviation.
and accordingly he might be able to change the muscle
activation, hence required motions are achieved. Since this
controlling method uses human vision feedback, the system
C. Wrist can be considered as a Human In The Loop (HITL) system
[14].
U/R motor (DCX22S, Maxon Motors) of the wrist [see EMG based controlling algorithm is shown in Fig. 4.
Fig. 1(d)] is connected through the wrist connecting plate. The algorithm consists of proportional EMG controller and
Bevel gear pair rigidly attached to the U/R motor shaft and motion switcher. Motion switcher can sequentially switch the
U/R deviation shaft enables the U/R deviation. Bevel gear motion.
attached to the motor shaft transmits rotary motion generated RMS of EMG signals for a sample size of 100 is taken
by the motor to the U/R deviation shaft through the bevel as of (1).
gear attached to the U/R Deviation shaft. This mechanism
gives the flexibility to the design, to adjust the distance s
Pt+N
between two perpendicular axes of the two wrist motions. i=t Ei 2
Rn = (1)
Wrist of MoBio has the same feature with 20 mm shift N
between these two axes [see Fig. 3]. L brackets of the wrist
where, Rn , Ei , N and t are the RMS of channel n (n=1,2),
is connected to the wrist connecting plate. Shaft for U/R
raw EMG of channel n at the given time, sample size and
deviation is supported on L brackets with two bearings,
the beginning of the sample respectively.
where the shaft can rotate relative to the L bracket. The shape
These RMS of EMG signals of biceps brachii and triceps
of the U/R motor plates is designed to achieve the full range
brachii are taken as the inputs to the EMG based controlling
of motion of the U/R deviation. U bracket is connected to the
algorithm. The addition of two RMS of EMG values (RT )
wrist F/E motor (DCX22S, Maxon Motors) through bearing,
[refer (2)] has been checked for a isometric contraction [15].
which provides the hand-wrist connection option and the F/E
motion of the wrist.
RT = R1 + R 2 (2)
D. Hand
Motion switcher is activated if an isometric contraction
The hand [Fig. 1(e)] comprises of a hand base which is detected[see Fig. 4]. It switches the motion sequentially
enables the connection with the U bracket of the wrist. DC between 5 motions. The wearer can select desired motion by
motor is located inside the hand. A screw is fixed coaxially sequentially switching through the motion sequence.
with the DC motor shaft. A ball which moves along the screw The prosthesis controller controls the selected motion from
is fit to the fingers. Therefore, when the DC motor starts to the motion switcher, if RT is below a specified threshold. The
rotate, thumb and index fingers are pushed apart or towards EMG proportional controller uses the agonist−antagonist
due to the ball and screw mechanism [13]. nature of biceps brachii and triceps brachii. The elbow
flexion motion makes higher signals on biceps brachii and
III. P ROSTHETIC CONTROLLER
lower signals on triceps brachii. Elbow extension makes
The prosthetic controller controls the prosthesis using triceps brachii signals high and biceps brachii signals low.
EMG signals from biceps brachii and triceps brachii. It is This phenomena is used in the EMG proportional controller.
assumed that the parts of two muscles are available after EMG proportional controller is given as of (3).

1629
 brushed DC motors of the prosthetic device and EPOS2
PERSONAL COMPUTER motor controller is used for the controlling of the BLDC
DAQ card motor. In the experiments, EMG signals are sampled at
EXTRACTION SYSTEM
Micro-controllers 2000Hz and band pass filtered to be within 50Hz to 450Hz.

Micro-controller B. Experiments
EMG SIGNAL

Motor Drives
In the first experiment, to monitor the joint angle re-

PROSTHESIS
sponses according to the PID controller, a sinusoidal wave
Angle measuring is generated through micro-controller as the desired motion
device
to PID controllers for the brushed DC motors located in the
forearm joint and the wrist of prosthesis. Sinusoidal input
Human arm Encoders Motors and encoder feedback (output motion) values are collected
using the micro-controller.
Fig. 5. Experimental Setup Second experiment is carried out to validate the EMG
based proportional controller. Angle measuring device was
attached to the human arm. It was then connected to a micro-
controller to record the real time angle measurements. At
∆θ = K1 (K2 × R1 − K3 × R2 ) (3) the same time encoder feedback of the elbow motor was
recorded to determine the angles of the elbow motion. EMG
where ∆θ, K1 , K2 , K3 are the angular change of the joint signals extracted from biceps brachii and triceps brachii of a
and constant gains depend on the person respectively. [16] healthy subject were used to control the prosthesis. The joint
angle values given by the micro-controller output, are plotted
θ(t) = θ(t 1) + ∆θ (4) against the time according to the motions of both prosthetic
where θ(t) is current joint angle and θ(t 1) is the previous arm’s elbow and the healthy subject’s elbow motions.
joint angle. Third experiment is conducted to monitor the prosthetic
Joint angles found from (4) are sent to the low level arm’s ability to reach towards a destination. Prosthetic arm
controllers of the respective joints. was controlled by a healthy human arm. First the path of
Motion sequence of MoBio has to be remembered by the the human arm was tracked. Then the path of the prosthesis
user. However this controlling method gives the flexibility was tracked while controlling by the healthy human hand.
for the user to skip motions if necessary. User has to Markers were pasted on both human hand and MoBio to
perform “n+1” number of motion switchers, in order to track their paths. The path of the hands in the both occasions
skip “n” number of motions in the sequence. The video were tracked and recorded using a High Definition camera.
attached with this paper summarizes and explains further The recorded path was plotted using the data extracted
about the sequence of motions and skipping the motions in using Kinovea software. Motion paths of both human arm
the sequence. and robotic arm in the X-Y plane and robotic arm’s path
Proportional-Integral-Derivative controllers (PID) are im- along 2 axes with respect to time were plotted afterwards.
plemented [17] as low-level controllers at micro-controllers Furthermore, EMG signal variation at the same time period
(ATmega 2560, Atmel). Low level controller is used by the is also recorded.
relevant prosthesis joint according to the angles received
C. Results
from the high level controller. Furthermore, each and ev-
ery joint is driven by a separate micro-controller. Motor Fig. 7 and Fig. 8 depict the output responses of the motors
commands are sent through micro-controllers to the motor for S/P, wrist F/E motions and U/R deviation. In each graph it
controller (L298, H-bridge). can be seen that output motions are not reached to the same
peaks as the desired motion. Since the ranges of motions
IV. E XPERIMENTS prosthesis are limited to the values as shown in Table II,
MoBio prosthetic limb was evaluated for its effectiveness prosthesis cannot achieve desired peak to peak ranges of
and usability. Sequence of experiments are carried out in this motions.
regard. Fig. 9 shows the motion angle relationship of both human
arm and MoBio in performing the elbow F/E motion. There
A. Experimental Setup is a lag in the robotic arm motion compared to the actual
Experimental setup (see Fig. 5) consists of MoBio, an- arm motion. This phenomena mainly occurs due to the delay
gle measuring device, 3 microcontrollers (ATmega 2560, generated by the signal processing and communication in the
Atmel), BLDC motor controller (EPOS2, Maxon Motors), micro-controllers and softwares. The distortion which has
3 H-bridges (L298), data acquisition (DAQ) card (6220, occurred at the peak of the robotic arm motion curve, has
National Instruments), EMG extraction system (16 channel generated due to the noise interferences.
BagnoliT M desktop EMG system, DELSYS) and a personal Fig. 6 shows the spatial motion of MoBio in reaching an
computer (PC) with MATLAB. H-bridges are used for the object. Fig. 10 depicts the vertical movement (along Y-axis)

1630
 A B C D E F

Fig. 6. MoBio’s reach towards an object: User controls MoBio through his motion intentions while taking the vision feed back from the eyes, to reach
the hand towards a specific object. The motion is performed in a plane parallel to sagittal plane.

Desired Motion 160

Output Motion Actual arm
50 140 MoBio arm

120
Angle (deg)

0
100

Joint Angle (deg)

-50 80

60
-100
40
0 500 1000 1500 2000
Time (ms) 20

Fig. 7. Motion output of the S/P motor to a desired motion input 0

0 10 20 30 40 50 60
Time (s)

40 Fig. 9. Elbow motion comparison between actual human arm and prosthetic
100 Desired Motion Desired Motion
Output Motion Output Motion
arm
30

50 20 TABLE II
10 R ANGE OF MOTIONS COMPARISON OF ACTUAL ARM AND M O B IO ARM
0
0
Range (deg) Range (deg)
-10 Motion Human Limb MoBio arm
-50
-20
Elbow Flexion/Extension 0 - 145 0 - 150
Supination/Pronation -85 - 70 -85 - 70
-100 -30 Wrist Flexion/Extension -70 - 70 -60 - 60
−10 Wrist Ulnar/Radial deviation -35 - 20 -27 - 25
0 500 1000 1500 2000 2500 3000 100 200 300 400 500 600 700 800
Time (ms) Time (ms)

Fig. 8. (a) Motion output of the wrist F/E motor to a desired motion input
(b) Motion output of the wrist U/R motor to a desired motion input V. C ONCLUSION
This paper proposed a 5 DOF trans-humeral prosthesis
which can mimic the human motions. A novel 2 DOF
mechanism is proposed to realize the wrist motions. Ex-
and horizontal movement (along X-axis) of the prosthesis
perimental results verified that the prosthesis can achieve
during this task. Furthermore, EMG pattern recorded at
the full range of motions of elbow F/E, forearm S/P, U/R
the same time is also included in the graph so that the
deviation and wrist F/E, just like the actual human arm.
shifting regions can be identified. Here the hand motion
MoBio weighs about 3.2kg which is same as grown human’s
is not available in the graph [Fig. 10] because it only
upper limb. Furthermore, results confirm the fact that the
reveal MoBio’s ability to reach an object . Two consecutive
proposed prosthesis can reach a desired object successfully,
isometric contractions are performed during the reaching
which is an essential activity in performing ADL.
process without any visible time gap to skip the wrist U/R
deviation and hand motion at respective intervals. R EFERENCES
A comparison of experimentally obtained ranges of mo-
[1] D. A. Bennett, J. E. Mitchell, D. Truex, and M. Goldfarb, “Design of
tions of MoBio and ranges of motions of human arm [11] a Myoelectric Transhumeral Prosthesis,” IEEE/ASME Transactions on
are given in the Table II. Mechatronics, vol. 21, no. 4, pp. 1868–1879, Aug. 2016.

1631
 400
real-time myoelectric control of multifunction artificial arms,” JAMA,
vol. 301, no. 6, pp. 619–628, 2009.
300 Distance along X−axis [5] “All about the Luke Arm.” [Online]. Available:
Distance along Y−axis http://mobiusbionics.com/the-luke-arm.html
200 A-B Elbow flexion [6] S. K. Kundu, K. Kiguchi, and E. Horikawa, “Design and Control
B-C Forearm pronation Strategy for a 5 DOF Above-Elbow Prosthetic Arm,” International
C-D Wrist extension Journal of Assistive Robotics and Mechatronics, vol. 9, pp. 79–93,
D-E Wrist flexion
100
E-F Elbow extension
2008.
[7] C. Piazza, C. D. Santina, M. Catalano, G. Grioli, M. Garabini, and
A B C D E F A. Bicchi, “SoftHand Pro-D: Matching dynamic content of natural user
0
commands with hand embodiment for enhanced prosthesis control,”
in IEEE International Conference on Robotics and Automation, May
−100 2016, pp. 3516–3523.
[8] D. S. V. Bandara, R. A. R. C. Gopura, K. T. M. U. Hemapala,
and K. Kiguchi, “A multi-DoF anthropomorphic transradial prosthetic
−200 arm,” in IEEE RAS/EMBS International Conference on Biomedical
Robotics and Biomechatronics, Aug. 2014, pp. 1039–1044.
−300
[9] M. P. Smits, “Method and apparatus for switching degrees of freedom
0 1 2 3 4 5 6 in a prosthetic limb,” Aug. 9 1994, uS Patent 5,336,269.
Time (s)
[10] T. Albrektsson and C. Johansson, “Osteoinduction, osteoconduction
and osseointegration,” European Spine Journal, vol. 10, pp. S96–S101,
2001.
[11] P. Helliwell, Biomechanics of the Upper Limbs: Mechanics, Modeling,
and Musculoskeletal Injuries. Taylor & Francis, 2007.
[12] A. A. Hayat, R. G. Chittawadigi, A. D. Udai, and S. K. Saha,
“Identification of denavit-hartenberg parameters of an industrial robot,”
in Proceedings of Conference on Advances In Robotics, 2013, pp. 1–6.
[13] J. F. Lin, “Kinematic analysis of the ball screw mechanism considering
Fig. 10. Spacial motion of MoBio along X and Y directions in reaching variable contact angles and elastic deformations,” 2003.
a object and EMG signal pattern change with time [14] L. F. Cranor, “A framework for reasoning about the human in the
loop,” UPSEC, vol. 8, no. 2008, pp. 1–15, 2008.
[15] J. M. Wilson, C. K. Thompson, L. C. Miller, and C. J. Heckman,
“Intrinsic excitability of human motoneurons in biceps brachii versus
[2] D. G. K. Madusanka, L. N. S. Wijayasingha, R. A. R. C. Gopura, triceps brachii,” Journal of neurophysiology, vol. 113, no. 10, pp.
Y. W. R. Amarasinghe, and G. K. I. Mann, “A review on hybrid 3692–3699, 2015.
myoelectric control systems for upper limb prosthesis,” in Moratuwa [16] D. G. K. Madusanka, L. N. S. Wijayasingha, K. Sanjeevan, M. A. R.
Engineering Research Conference, Apr. 2015, pp. 136–141. Ahamed, J. C. W. Edirisooriya, and R. A. R. C. Gopura, “A 3DOF
[3] D. J. Magermans, E. K. J. Chadwick, H. E. J. Veeger, and F. C. T. transtibial robotic prosthetic limb,” in International Conference on
van der Helm, “Requirements for upper extremity motions during Information and Automation for Sustainability, Dec. 2014, pp. 1–6.
activities of daily living,” Clinical Biomechanics, vol. 20, no. 6, pp. [17] D. G. K. Madusanka, R. A. R. C. Gopura, Y. W. R. Amarasinghe,
591–599, Jul. 2005. and G. K. I. Mann, “A simulation environment for control algorithms
[4] T. A. Kuiken, G. Li, B. A. Lock, R. D. Lipschutz, L. A. Miller, K. A. of transhumeral prostheses,” in International Conference on Emerging
Stubblefield, and K. B. Englehart, “Targeted muscle reinnervation for Trends in Mechanical Engineering, 2015, pp. 190–196.

1632