Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 14

Article

Design of Upper-Body Exoskeleton with Multifilament Muscle


Actuation and Biomimetic Support
Soorya Narayanan, Rahul K S, Hyder Mundoly, Arshed Ahammed, Gangadhara Lacchireddi, and S.
Budhe

Abstract: The main aim of the present study is to develop an upper torso exoskeleton using
Multifilament Mckibben artificial muscles that facilitates load distribution and dynamic load
compensation features while providing an alternate load pathway to transfer the load from the
shoulders to the pelvis, in addition to the human spine. Exo-Suit's limb actuator devices and
sensors in the design will be based on study and analysis of human Musco-skeletal impact and
measured during the person’s interaction with the external environment.

Initially the dimensions of individual muscle strands are determined using the analytical approach.
Finite element method is then used to determine force supplied by individual and total bundle for
each muscle section and load bearing also. Electromyography (EMG) test is conducted to measure
muscle response and establish reference signal for further studies. Through this study we find a
faint balance between sufficient load bearing for human augmentation and sufficient flexibility for
a “best of both worlds” of sorts.

Keywords: Mckibben muscle; Exoskeleton; Multifilament; Biomimetic; Actuator

1. Introduction
Exoskeletons are wearable devices placed on the user’s body to augment, reinforce, or restore human
performance. Factory workers who use exoskeletons have less back and shoulder pain, and they are able
to be more physically active both on the clock and off. Exoskeletons can transfer the weight of a user’s
arms from the shoulders, neck, and upper body to the body’s core, reducing physical stresses. Pneumatic
exoskeletons increase strength and provide stability through human-guided flexion/extension and locking
mechanisms.
Other than the labour reduction capabilities that exo-suits may be designed to have, an exo-suit can
also be designed to augment the users strength and limits. This paper discusses the design of such an exo-
suit that is powered by multifilament Mckibben muscles [1]. Mckibben muscles are pneumatic devices
that work as actuators to enable motion of machine elements. Mckibben muscles can be bundled together
to form what resembles a human muscle and this bundle can be controlled like the earlier to realise a
human like movement. Although in its budding this stage, institutes have been able to produce impressive
motion with “multifilament muscle“fibres. However the use of multifilament muscle in exo-suits is
relatively new and this paper is an attempt to figure out its feasibility. A conceptual design based on
desired requirements are thought of and discussed throughout the length of this paper. CAD models that
represent the ideas are also shown. Further possible sensory methods and control are also looked at.
Research and literature review concludes, multifilament muscles triumph over ordinary Mckibben
muscles when contraction ratio and actuation methods are considered [1]. Thicker bundles does not
proportionally increase contraction force due to dead space. Multifilament muscles are a better choice for
implementing muscles redundancy in robots Formulations have been made without considering hysteresis
and other material factors-So Schutle’s formula [2] stands valid. The basic strand build contains: inner
elastic tube, outer stiffer tube, covering mesh, clamps and actuator. EMG signals have been employed in
most designs although they produce significant noise. AgCl [3] electrodes have been implemented as
input devices as they produce better results as far as surface EMG electrodes are concerned.Proper
filtering of signals [4] should be done to avoid unwanted disturbances in the control .A common element
found in pneumatic designs, is the presence of a binary control that lets the suit work only in the two
extremities, ON/OFF.A better continuous control has been shown to bring better utility to the suit. Hence
this paper also talks on the inclusion of a proportionally controlled solenoid valve and its control. A PID
[3] force control was implemented to add control to the entire suit. A basic Arduino model was added to
install the basic logic loop of the suit.

2. Materials and Methods


2.1 Theoretical methodolgy
2.1.1 Conceptual design of Mckibben Muscles
 A particular direction was taken while working on the design of the Mckibben muscles and the mechanics
of the rigid elements present.
 Mckibben muscles and the mechanics of the rigid elements are presented in figure 1.

Figure 1. CAD of Mckibben muscle fibre


 Dimensions of Mckibben muscle are considered as,
Braiding angle (α ¿ = 180 , Length of Mckibben muscle = 0.2 m, Outer diameter (D 0)= 6 mm, Outer sleeve
diameter =0.4 mm
For inner tube, Outer diameter =4.34 mm, Inner diameter =3mm, Different dimension of mckibben
muscles are shown in the figure 2.

Figure 2. Mckibben Muscle


 Theoretical force formula (Known as Shulte’s formula [2]) to calculate the theoretical force exerted by
Mckibben muscle is,
π R 20 P x (1)
F ( p , ε )=
2
( ( 1−ϵ ) cos
2
α −1 )
sin 2 α
P*=Papp-Pth

 Relation between applied pressure and contraction ratio was determined through the steps given below
Maximum contraction ratio as [1].
1 (2)
ε max=1− =0.3929
√3 cos 18 0
From the research, maximum pressure that can be applied =0.7 MPa [12]
Assuming Mckibben muscle as a pressurized rubber hollow cylinder having thickness t b, Elastic modulus
Eb, then the circumferential strain [9],
△ R ( P−Pth ) R 0 (3)
ε c= =
R0 Eb t b
Pth is estimated by regression method from the data from research papers [8] given in the table 1 and the
estimated value is 0.1441MPa
Table 1. list of Pth values along with wall thickness of Mckibben muscles
X: Wall thickness (mm) 0.28 0.56 1
Y: Pth 0.08 0.21 1.89
Final Relation between length and force was derived from the the equations (1), (2), (3) and is given
below
−4 −10 2 −16 3
F=5.0705 ×10 ( P−Pth )−5.006 ×10 ( P−Pth ) −7.3828× 10 ( P−Pth ) (4)
Maximum Pressure value for biceps Mckibben muscle is determined by considering Maximum required
biceps force as 500 N and number of individual strands on biceps as n=16, there individual strand force
requirement is F= 31.25 N Then required actuation pressure from equation (4), P = 0.652624 MPa
 Maximum required actuaction pressure is determined from the FBD of alignment of back muscles as
shown in the figure 3

Figure 3. FBD of Back muscles along with shoulder. Dimensions


are AB=153.5 mm, CA=217.0817 mm, BD=108.54 mm

Length of back Mckibben muscle = 217.0817 mm


Total force required by back Mckibben muscle F m = 707.106721 N, Considered number of muscles is 22
then Maximum force required by single filament Fsingle= 32.1412 N
Therefore by solving equation (4) required actuation pressure is, P = 0.6510683 MPa

2.1.2 DH modelling of exo-suit:


Kinematic modelling can be done with DH representation method [10]. DH parameter table of robot
manipulator is used as an input to dynamic modelling, Torque requirements for each joint can be
calculated from dynamic modelling. Arm Drawing and DH modelling are shown in the figure 4 and
figure 5
Figure 4. Drawing of one arm of exosuit

Figure 5. DH model representation of exosuit


DH parameter was determined by analysing DH modelling and is tabulated in table 2
Table 2. DH parameter table
Link θ d a α
0-1 θ 1=0 0 0 -90
1-2 θ2-90 0 a2 -90
2-3 θ3-90 a3 a4 -90
3-4 θ4-90 0 a5 0
4-5 θ5-90 a7 a6+a8 0

General transformation matrix and total transformation matrix [11]:


[ ]
C θ n+1 −S θ n+1 C α n +1 S θn +1 S α n+1 an +1 C θ n+1
S θ n+1 C θ n+1 C α n+1 −C θ n+1 S α n+1 an +1 S θn +1
An +1=
0 S θn+1 C α n+1 d n +1
0 0 0 1
0
Total transformation matrix, T 5= A1 A 2 A 3 A 4 A 5

Data from DH parameters and DH modelling are used to estimate Dynamic modelling which determines
the force and torque required to give to a particular joint or link inorder to get a particular motion.

2.2 Sensors
An electromyography (EMG) sensor is used for obtaining the input signal for our exosuit. A potential
difference is created in the neurons of our muscle fibers as it contracts and expands, the emg sensor
continuously measures this potential difference across our muscle. Surface emg electrode is used instead
of needle electrode due to its non-invasive technique and simple setup. The AgCl gel electrodes are
mostly preferred for obtaining accurate value of emg. The AgCl layer in the electrode allows current from
the muscle to pass more freely across the junction between the electrolyte and the electrode [4]. Thus,
only a feeble electrical noise is introduced in the measurement). The useable range of EMG signal is from
50-500 Hz. and the dominant information lies in 20-150 Hz.
An Arduino Uno microcontroller can be used to convert AC to DC using the root mean square method
[5]. A differential amplifier with high CMRR ratio is used as the first stage of processing, its advantage is
that the elements which are common in both the signal inputs are subtracted by this differential amplifier .

2.2.1 Experimental details (validation)


As per these requirements the following hardware was chosen and a circuit was designed ( Figure 6):
1. Advancer technologies EMG Muscle Sensor V3
2. Arduino Uno R3 (ATmega328)
3. EMG Gel Electrodes
4. MB102 Breadboard
5. 9V Batteries
6. Jumper wires
Figure 6. Experiment setup to obtain EMG reference signals
The aim was to obtain the emg values for biceps contractions and extension, the test was carried out with
one of the group member as the subject. The skin was properly cleaned before placing the emg gel
electrodes. Two electrodes were placed across the biceps muscle and the third electrode has to be placed
on a bony area which serves as the reference electrode.
Due to the high cost of AgCl electrodes and its less availability we went with normal gel electrodes which
introduced greater noise in the signal. The emg sensor V3 that we choose has a good filtering band width
and decent amplification factor, so it was able to produce a good rectified signal output. The arduino uno
R3 was appropriately coded to obtain the rms value of signal. Output signals from biceps muscles shown
in the figure 7.

Figure 7. EMG results of biceps extension and contraction

2.3 Modelling
Further the above calculations and the dynamic modelling were used to realise what the suit would look
like on a test user; the following images describe the design conceptualised and designed on CAD.
Figure 8. CAD model of exo-suit
2.3.1 Shoulder Complex

Figure 9. Shoulder Complex

2.3.2 Arm
Figure 10. Upper Arm

2.3.3 Elbow joint

Figure 11. Elbow joint

2.3.4 Forearm

Figure 12. Forearm

2.3.5 Hand complex


Figure 12. Hand complex

2.3.6 Back complex

Figure 13. Back complex

2.4 Control
The control of pneumatic suits has always been a challenge that requires precision in all the systems that
make the suit. Binary control is easily achieved using a simple logic where the suit can be made to power
up when the load increases above the human limit. However a more complicated yet continuous approach
can be implemented with proper choice of valves, and in this case we take the help of proportional
solenoid valves [7]. Although this adds more complexity to the control design the benefits can blind it.
The figure shows a Simulink schematic for a proportional valve control. The mechanical and electrical
segments work on the control variable which is the ”spool position “ based on this , the PID works to
control the suit each muscle at a time..
Figure 14. Proportional Solenoid Valve Control block

Figure 15. Overall Simulink System block


3. Results and Conclusion

The given conceptual design requires further study and analysis through proper experimentation. The
material initially chosen for the Mckibben muscle was silicon rubber, but experimentation might prove
the need for better and sturdier materials that can withstand the required loads and working scenarios.

Using a statistical method, an empirical relationship was found between braid angle, twist angle, and
length of the muscle.
φ=9 ×10−5 l 2−2 ×10−7 θ 2+ 0.1277 l+ 0.0062θ+ 42.295 (5)
Where  = Braid angle (°), l= length of the muscle (mm), = twist angle (°)
The error in the above equation is estimated around ±0.986°. Based on the load conditions, simulation of
the entire assembly was done in Fusion360 and from the results it was observed that the design is safe
under maximum load condition with a minimum FOS value of 2.412. The structural analysis of entire suit
was conducted on the applied range of load, 0 to 50 kg.
7

4
FOS

FOS
3 FOS = 2
Load = 50 kg
2

0
0 10 20 30 40 50 60 70 80 90 100
Load/kg

Figure 16. Variation of FOS w.r.t Load

The study was made on the assumptions that back muscles are always engaged, hence the Mckibben
fibers which facilitate shoulder motions are full engaged. Therefore the study is conducted by varying
load applied to the exoskeleton and support load offered by biceps Mckibben fibers. The following graphs
shows the variation of different strength analysis parameter within the range of load applied to the
exoskeleton
1000
900
800

Von Mises Stress/MPa


700
600
500
400
300
200
100
0
0 10 20 30 40 50 60 70 80 90 100
Load/kg

Figure 17. Variation in Stress w.r.t Load

1000
900
800
700
600
Stress/MPa

500
400
300
200
100
0
0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008
Strain

Figure 18. Stress-Strain Curve of the suit

From the graphs we can conclude that our design is safe throughout our range of load (0-50 kg)
(FOS>2).Strength wise the suit works at the best condition in between 15 to 30 kilograms (FOS>4). But
when structural analysis was done with maximum load condition, load applied to the suit was 50 kg and
the support offered by Biceps Mckibben fibers are to be around 60 kgwt, to accommodate weight of the
suit and frictional loses. Back complex muscle fibers which assists in shoulder internal-external rotation,
will provide a maximum load of 100 N towards back (since the range of motion to back from neutral
positon is -20°). The fibers which assists in shoulder flexion-extension will provide a load of 500 N,
considering all these factors simulation was done on the suit.
Figure 19: Stress Distribution in Exoskeleton

Table 1. Simulation results in Exoskeleton

Minimum Maximum
Safety Factor
Safety Factor (Per Body) 2.356 15
Stress
Von Mises 8.617*10-6 MPa 375.5 MPa
1st Principal -190.9 MPa 248.9 MPa
3rd Principal -597.2 MPa 89.06 MPa

Total Displacement 0 mm 4.012 mm

Reaction Force Total 0N 1268 N


X -121.6 N 102 N
Y -286.8 N 415.6 N
Z -1255 N 560.7 N

Equivalent Strain 5.749E-11 0.002985

Total Contact Pressure 0 MPa 188.9 MPa

The EMG circuit used poses its own issues as surface electrodes can do only so much in gathering
required data from the user. Although current placing does give appreciable and reliable values better
positions can be achieved through further study. The control mechanism should be tested in rigorous
situations to ensure reliable circuitry. As added in other exo-suits, inclusion of a modified version of the
“dead-man’s switch” can be included in the postulated design so that in case of mechanical or electrical
failure , the user can release the switch to prevent injury to themselves.

4. Discussion
In future designs, a load bearing suit that comprises of antagonistic actuators that emulate the structure of
human musco-skeleton can pave the way for superior characteristics. As a further note adding complex
control devices like an FGPA Array can bring in precise control that opens doors to much better control
and augmentation capabilities. The CAD designs developed can be made dynamic to understand the
working of the suit better against actual working environment to better understand the regions of strengths
and weaknesses. In short just as you would picture the human body in its inherent muscular form, the
future holds a design where the addition of a second layer of artificial muscle of Mckibben form bridges
the gap between the human limit and superior strength

5. Patents

Supplementary Materials:
Author Contributions: Conceptualization, Soorya Narayanan; methodology, Soorya Narayanan, Rahul
K.S, Arshed Ahmed and Hyder Mundoly; software, Soorya Narayanan and Rahul K.S; validation, Rahul
K.S, Arshed Ahmed and Hyder Mundoly; formal analysis, Arshed Ahmed ; investigation, Arshed Ahmed;
resources, Soorya Narayanan, Rahul K.S, Arshed Ahmed and Hyder Mundoly; data curation, Soorya
Narayanan, Rahul K.S, Arshed Ahmed and Hyder Mundoly; writing—original draft preparation, Rahul
K.S , Arshed Ahmed and Hyder Mundoly; writing—review and editing, Soorya Narayanan and Rahul
K.S; visualization, Rahul K.S ; supervision, Soorya Narayanan. All authors have read and agreed to the
published version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement: All data assumed for producing results in this paper has been obtained
from the references mentioned. EMG data and other measurable values have been measured in real-time.
Basic data has been taken from certain papers to ascertain a few parameters used in the equations in this
paper.
Acknowledgments: This paper would not have been feasible without the kind support and assistance of
many individuals and organizations. We want to thank every one of them from the bottom of our hearts.
We would like to express our gratitude to the authors of the papers from which we were able to glean
important material. We would also want to thank a variety of technical entities for contributing to our
work by assisting us with their expertise.
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A

Appendix B

You might also like