IEEE ROBOTICS AND AUTOMATION LETTERS, VOL. 6, NO.
3, JULY 2021
Design, Fabrication, and Hysteresis Modeling of Soft
Microtubule Artificial Muscle (SMAM) for
Medical Applications
Mai Thanh Thai , Phuoc Thien Phan , Trung Thien Hoang , Harrison Low, Nigel H. Lovell , Fellow, IEEE,
and Thanh Nho Do , Member, IEEE
Abstract—Robotic artificial muscles (RAMs) are promising
power sources for medical fields such as surgical robotics. However,
existing RAMs are challenged by scalability, material costs and
fabrications. The nonlinear hysteresis in fluid-driven RAMs causes
oscillations in open-loop systems. To circumvent these limitations,
this letter introduces hydraulically soft microtubule artificial mus-
cle (SMAM) that is low-cost and scalable, yet simple to fabricate.
The SMAM, which only requires a flexible silicone microtube and a
hollow micro-coil, is elongated or contracted under a fluid pressure.
The SMAM presents an ideal candidate for flexible robotic systems
such as endoscopic surgical robots. Experiments are conducted to
characterize the SMAMs. Results show that the hysteresis profiles
between the input syringe plunger position and output position
are stable regardless of its configuration, as opposed to the highly
variable responses for the tendon-sheath mechanisms. A new non-
linear model is developed to characterize the asymmetric hysteresis
phenomena of the SMAM. Compared to the Bouc-Wen hysteresis
models, the developed model presents a better capture of hysteresis.
To demonstrate the muscle capability, a SMAMs-driven pulley
and a flexible surgical arm are given. The new SMAM and its
asymmetric hysteresis model are expected to provide a path for
the development of rapidly efficient and low-cost soft actuators for
use in flexible medical devices and surgical robotic systems.
Index Terms—Soft robotics, surgical robots, artificial muscles,
fluid-driven, hysteresis modeling.
I. INTRODUCTION
OBOTIC artificial muscles (RAMs) have potential for
R many diverse applications including industrial robots,
wearable devices and medical instruments [1], [2]. There is
also a need to replace primitive tendons in surgical systems
with RAMs that confer better characteristics [3] for flexible
surgical system [4]–[6]. Due to their small size and low weight,
tendon-driven mechanisms (TDMs) have been used for these
surgeries; allowing surgeons to access complicated and confined
Manuscript received October 15, 2020; accepted March 5, 2021. Date of
publication April 12, 2021; date of current version April 21, 2021. This letter was
recommended for publication by Associate Editor C.-H. Yeow and Editor K.-Y.
Cho upon evaluation of the reviewers’ comments. This work was supported in
part by UNSW start-up under Grant PS58173, and in part by the UNSW Scientia
Fellowship PS46197. (Corresponding Author: Thanh Nho Do)
The authors are with the Graduate School of Biomedical Engineering,
Faculty of Engineering, UNSW Sydney, Kensington Campus, NSW 2052,
Australia (e-mail: maithanh.thai@unsw.edu.au; phuoc_thien.phan@unsw.edu.
au; trungthien.hoang@unsw.edu.au; harrison.a.low@gmail.com; n.lovell@
unsw.edu.au; tn.do@unsw.edu.au).
Digital Object Identifier 10.1109/LRA.2021.3072599
2377-3766 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Florida. Downloaded on March 20,2022 at 23:51:25 UTC from IEEE Xplore. Restrictions apply.
5089
regions of the human internal organs [6]. However, energy loss
caused by nonlinear friction is significant [7]. Furthermore, the
motion and output force of the TDMs are highly dependent
on the distance between its end-effector and its power source,
resulting in poor performance when working against the surgical
target.
Recently, soft artificial muscles (SAMs) have emerged as
a potential candidate for flexible robotic systems [8], [9]. To
drive the SAMs, stimulants such as light, temperature, chemical
composition, fluid, electric and magnetic field sources have
been used [10], [11]. Briefly, light-induced soft actuation is
driven by photochromic molecules to transduce optical en-
ergy to strain/stress. Thermally-driven actuators (TDAs) provide
motion and deformation through the thermal response of the
materials [12]. The TDAs also include shape memory alloys
where contraction motion is achieved under an increase in their
temperature above a threshold [13], [14]. Chemically-responsive
actuators are based on the chemical reaction of substances or
fluid evaporation to transform chemical energy into mechanical
energy [10]. Electrically-driven actuators rely on the respon-
siveness of materials to electrical excitation to generate motion
and force while magnetically-driven soft actuators can generate
motion and force upon exposure to an external magnetic field.
Lastly, fluid-driven artificial muscles (FDAMs) which com-
monly consist of soft fluid-filled chambers arranged in specific
patterns can generate the desired extension, twisting, bending or
contraction under pressure sources, depending on their structural
design [15], [16].
The most well-known FDAM is McKibben muscle [17].
When positive pressure is applied, the McKibben muscle ra-
dially expands and exerts a longitudinal contraction force. By
adding stiff fibers along the length of the cylindrical chamber in
desired patterns, this muscle can contract, extend or bend. The
FDAMs have been recently implemented in surgical systems.
For example, Volder et al. [18] developed a miniature McKibben
muscle to drive an instrument for keyhole surgery. Veale et al.
[19] introduced a hydraulically-actuated Peano muscle with high
flow rates and efficiency peaks of 40%. However, these muscles
exhibit low contraction ratios and bulky structures, which are
ill-suited for medical systems. Li et al. [20] designed negative
fluid pressures-driven origami-inspired artificial muscles. These
muscles could contract over 90% of their original length, gen-
erate stresses of 600 kPa, and deliver peak power densities over
2 kW/kg. Peters et al. [21] demonstrated a new FDAM integrated
5090 IEEE ROBOTICS AND AUTOMATION LETTERS, VOL. 6, NO. 3, JULY 2021

low melting-point material for stiffness control. Fu et al. [22]

introduced a bio-inspired spring reinforced actuator that was
built based on an FDAM and a flexible spring as backbone.
This actuator could enhance robot dynamics while retaining
omnidirectional bending. However, its limitations include large
size, low force, and high weight. In addition, most FDAMs
exhibit viscoelastic properties and this causes hysteresis and
large inaccuracies in muscle performance [23]. Accurate motion
and force control of the FDAMs remains an unsolved problem
due to several challenges such as memory effects, nonlinear
friction, and elastic deformation of soft materials, making the
modeling and model-based approaches difficult. Nonlinear hys-
teresis can also cause oscillations in open-loop systems, poor
tracking results, and potential uncertainties.
Hysteresis modeling for soft actuators has been an active
research area, especially in surgical robots [7]. Several hysteresis
approaches for soft actuators have been studied including the
use of integral modelings such as Preisach, Maxwell-Slip, and
Prandtl–Ishlinskii (P-I) models [24]. However, some of these
methods need more calculation costs in comparison with others.
In contrast, differential modeling such as Bouc-Wen models
and their variants are great candidates to describe the dynamic
hysteresis behaviors due to low complexity in both implementa-
tion and computation. They can also capture a broad variety of
Fig. 1. Soft microtubule artificial muscle (SMAM). (a) Fabrication process
hysteresis curves using continuous state variables and shapes of and working principle. (b) ∼247% in elongation of a SMAM (length 47 mm,
the hysteresis loops [25] However, there is a trade-off between OD = 1.45 mm). (c) SMAM (OD = 1.45 mm, weight = 0.28 g) lifts a load
the number of model variables and accuracy [26]. of 50 g. (d) SMAM (OD = 6.35 mm, weight = 51g) lifts a load of 4.72 kg;
(e) SMAM operates as a tendon sheath mechanism with PTFE tube as a sheath.
To overcome these limitations, this letter will develop a hy-
draulically soft microtubule artificial muscle (SMAM) using
a low-cost fabrication method of insertion, enabling a high
The SMAM diameter is scalable to 0.84 mm or 6.35 mm. It
strength, small size, and high aspect ratio actuator. The SMAM
consists of a flexible inner silicone microtubule and an outer
consists of a flexible silicone microtube and a hollow micro-coil
hollow micro-coil made from inextensible fiber, which covers
made from inextensible fibers. The new architecture of SMAMs
the inner microtubule to prevent radial expansion while pro-
can be implemented into flexible robotic systems such as surgi-
moting axial extension under pressure. Upon pressurization,
cal robots and soft wearable devices. To capture the SMAM
both outer micro-coil and inner microtubule store elastic energy
hysteresis nonlinearity, we introduce a new nonlinear model
(EE), which is released after reducing the pressure, enabling
that closely follows its hysteresis profile under various input
programmable motion and contraction force. The SMAM pres-
signals. To validate the new muscle and the new model, ex-
sure is supplied through a guide tube that joins the muscle to
periments are performed to measure the input-output properties
a syringe driven by a DC motor (Faulhaber, Germany) and a
of the SMAM under different configurations. To demonstrate
linear ball screw (MISUMI, Japan). When the working liquid
the potential benefits of the SMAM over conventional TDMs,
is pressurized, the SMAM is lengthened from Li (position I)
we develop a SMAMs-driven pulley mechanism which is com-
to L (position A) (Fig. 1(a)) or an elongation x. Because of the
mon in many medical robotic systems and a SMAMs-actuated
constraint layer, there is no change in the SMAM diameter under
flexible surgical arm with an integrated gripper that can grip
applied pressure and therefore mechanical energy is converted
desired objects. The rest of this letter is organized as follows:
to the EE, which is released to apply a contraction force against
Section II presents the SMAM developments. Section III gives
a load once the pressure drops. The initial tension of SMAM
the experimental setup and results. Section IV covers a new
highly depends on the initial micro-coil force and elastic force
asymmetric hysteresis model. Section V explores the hysteresis
of the microtubule. These forces will cause the SMAM to return
model validations. Section VI shows how SMAMs can be ap-
to its initial length when working against a load. In [27], an
plied to a pulley mechanism and a robotic catheter. Finally, a
analytical model was developed and the responses of the SMAM
discussion of results, concluding remarks and future work are
were characterized with good durability. However, the SMAMs
presented in Section VII.
with soft materials inherently experience nonlinear memory
effects, friction, and elastic deformation, resulting in a nonlinear
II. DESIGN, FABRICATION AND WORKING PRINCIPLE OF SOFT hysteresis profile.
MICROTUBULE ARTIFICIAL MUSCLE
A. Design and Working Principle B. Fabrication Method and Performance
We here present a new artificial muscle that converts hy- In this letter, the SMAM is fabricated from the insertion
draulic power to force and motion with insignificant energy method (Fig. 1(a)) where an elastic microtubule is inserted
loss regardless of the transmission paths of the driving source. into a hollow coil, enabling a high aspect ratio muscle (up to

Authorized licensed use limited to: University of Florida. Downloaded on March 20,2022 at 23:51:25 UTC from IEEE Xplore. Restrictions apply.
THAI et al.: DESIGN, FABRICATION, AND HYSTERESIS MODELING OF SOFT MICROTUBULE ARTIFICIAL
Fig. 2. SMAM prototypes. (a) SMAMs with different sizes and materials
for the coils and the inner microtubules (stainless steel, fishing line, different
silicone elastomers). (b) SMAM under pressure. (c) A long SMAM (length-
3000 mm, diameter of 1.45 mm) with high aspect ratio of 2069 (length/diameter).
(d) Apparatus for elongation testing.
2069, Fig. 2(c)) while avoiding the non-uniform distribution of
a constraint layer during operation. The stiffness of inner tube
and outer coils does affect the SMAM force. The microtubule
stiffness can be increased by applying a pre-stretch to it during
the insertion process. The outer helical coil can be made from
fishing line, metal wire or nylon wire (Fig. 2(a)) while the inner
microtubule can be any stretchable material fabricated from the
rolling coating process [28] or simply purchased off-the-shelf.
In this letter, we use a stainless-steel micro-coil due to its high
resilience and high energy storage compared to other materials.
Other materials can be used depending on specific applications.
In this letter, we use distilled water as the driving liquid. Exper-
iments are conducted at room temperature and high pressure
(>1.2 bar). Therefore, boiling effects can be ignored. Other
liquids such as medical/cooking/hydraulic oils can be used,
depending on specific applications (medical oils for medical
robots or hydraulic oils for high-temperature environments). We
also degas air bubbles in the muscle using a vacuum chamber
for 30 minutes (Binder - VD115, Binder, USA) until they are
completely removed. This step is vital to the device quality as air
bubbles have considerable effects on the hysteresis shape. We
tie and seal one end of the inner microtubule to the micro-coil
with superglue (LOCTITE, USA) and connect its other end
to a miniature syringe via a guide tube. The guide tube and
microtubule are connected via a short metal tube, with sewing
thread tightly tied around the junctions, to enhance connection
and support.
C. Dimension Choice for the SMAM
The SMAM diameter, length, inner tube materials, and con-
straint coil can be flexibly chosen for specific applications,
depending on force and dimension requirements. To increase the
SMAM force, a stronger inner microtubule with higher Young’s
modulus and stiffer micro-coil is desired. Fig. 2(a) shows various
SMAM prototypes made from different materials for the inner
Authorized licensed use limited to: University of Florida. Downloaded on March 20,2022 at 23:51:25 UTC from IEEE Xplore. Restrictions apply.
5091
tube (Ecoflex-0030, ReoFlex-20 from Smooth-On Inc, Nusil
elastomer, or latex tubing from McMaster-CARR, USA) and the
coil (fishing line, stainless steel). Results from Fig. 1(c) show
that SMAM (silicone inner tube from Saint-Gobain S.A., France,
outer micro-coil from Asahi Intecc Co., Japan, outer diameter
(OD) = 1.45 mm, weight = 0.28 g) could lift a load of 50 g
(178 times its mass) or SMAM (OD = 6.35 mm, weight = 51g)
could lift a load of 4.72 kg (Fig. 1(d)). In addition, SMAM
(OD = 1.45 mm) can elongate up to 245 % of its original length
(Fig. 1(b)). We also demonstrated that the SMAM can be used
as a conventional tendon sheath mechanism (TSM) where it
slides inside an outer PTFE tube simulated as a routing sheath
(Fig. 1(e)). In this letter, the SMAM dimension is chosen in a way
that is well-suited for use in surgical robots or flexible devices.
For example, the small SMAM in Fig. 2(b) could be encircled
under high pressure while the one in Fig. 2(c) has a high aspect
ratio of length/diameter of around 2069. The guide tube can
have any length, depending on the hydraulic source location. For
example, in endoscopic surgical robots, the distance between the
pressure source and the actuator can be up to 2 m, and hence a
SMAM with a long hydraulic transmission is well-suited as it
has a constant energy efficiency regardless of the transmission
paths, as compared to TDMs where a high variation in the used
energy exists (details see next Sections).
III. EXPERIMENTAL SETUP AND RESULTS
An experimental setup (Fig. 2(d)) was developed to fully char-
acterize muscle performance. During the experiment, a SMAM
(OD = 1.45 mm, Li = 50 mm) was used where its proximal end
was firmly locked to a 3D printed fixture while its distal end was
connected to an elastic string (generated a small force of 0.05 N
which can be ignored) to prevent the muscle from buckling)
via an optical encoder (S6S-1000-IB, US Digital, USA) which
recorded the output displacement xout (t). The liquid (water) was
distributed to the muscle by a BD Luer-LokTM 1 mL syringe
(inner diameter (ID) = 4.5 mm) actuated by a DC micromotor
(model 3272G024CR, Faulhaber, Germany) and a ball screw
(MISUMI, Japan). To alleviate the nonlinear hysteresis between
the DC motor and the ball screw system, we used another optical
encoder to record the syringe plunger displacement xin (t). A data
acquisition device (QPIDe, Quanser, Canada) and MATLAB
Simulink were used to acquire the signals. All raw data were
processed through a digital bandpass filter to eliminate unwanted
noise [29], [30].
We here consider SMAM including the syringe and hydraulic
transmission tube as an element. The SMAM motion was excited
by periodic and non-periodic signals, simulating the motion used
in flexible surgical systems. Fig. 3(a) shows 30-second snippets
of a sinusoidal excitation with a frequency of 0.1 Hz. The left
and middle panels depict the time history of the input and output
displacements while the right panel presents the relationship
between the input and output displacements, respectively. Sim-
ilarly, Fig. 3(b) gives results for a dual-frequency sinusoidal
signal that contains two sinusoidal inputs with frequencies of
0.1 Hz and 0.2 Hz, respectively. Fig. 3(c) displays results for a
non-periodic
√ motion with two input frequencies of 0.1 Hz and
0.1 3 Hz. Results revealed that the SMAM hysteresis profile
was asymmetric in the loading and unloading phases. The width
of the hysteresis loops was bigger in the central profile and
5092
Fig. 3. Experimental results with different input signals. From left to right:
Time history of the syringe plunger displacement, time history of the SMAM
output elongation, and its hysteresis profile, respectively. (a) 0.1 Hz sine wave
excitation. (b) a dual-frequency sinusoidal excitation with frequencies of 0.1 Hz
and√0.2 Hz. (c) non-periodic excitation with two frequencies of 0.1 Hz and
0.1 3 Hz.
Fig. 4. SMAM performance. (a) Frequency responses of the SMAM up to 3 Hz
in 10 cycles. (b) The elongation versus the maximum generated force with the
line of best fit where y is maximum generated force and x is output elongation.
(c) Various configurations for the SMAM element. (d) Nearly constant hysteresis
profile of the SMAM with various configurations. (e) Nonlinear hysteresis profile
of the SMAM for various frequencies (0.2 Hz to 1.5 Hz). (f) Durability test.
(g) Input motion vs. contraction force at various initial elongated length. Un-
stretched muscle length = 50 mm.
smaller at the reverse points. The instantaneous elongation of
the muscle in the unloading phase was greater than the loading
phase with no backlash in the hysteresis loops. To show the
reliability of SMAMs, a test was conducted over 2000 seconds
(elongation 80%, f = 0.1 Hz). The result (Fig. 4(f)) reveals that
the elongation of SMAM shrinks only 2% due to the inherent
elastic properties of its components.
We also evaluated the SMAM frequency responses with a
bandwidth of up to 3 Hz, which is an acceptable working range
Authorized licensed use limited to: University of Florida. Downloaded on March 20,2022 at 23:51:25 UTC from IEEE Xplore. Restrictions apply.
IEEE ROBOTICS AND AUTOMATION LETTERS, VOL. 6, NO. 3, JULY 2021
for most surgical robots [7]. Results (average 10 trials and
its standard deviation) disclosed stability although there was a
reduction of 0.4 dB over that frequency range (Fig. 4(a)). We
also conducted experiments to depict the relationship between
SMAM elongation and its maximum generated force. The results
showed extreme linearity (Fig. 4(b)) with the goodness-of-fit
measure (R2 = 0.9962) in 10 trials, suggesting that the muscle
elongation was a good proxy for ascertaining the maximum force
delivered by the muscle. One of the main advantages of the
SMAM over conventional TSMs is the ability to maintain its
hysteresis loop (or constant energy loss) regardless of the trans-
mission paths. To demonstrate this, we conducted experiments
with different configurations between the driving source (muscle
proximal end) and the muscle distal end where the hydraulic
transmission tube was routed inside a flexible body simulated
as a flexible endoscope (Fig. 4(c)). The results (Fig. 4(d)) re-
vealed that the SMAM hysteresis profile was almost unchanged
regardless of the change of the transmission channel. We also
investigated the SMAM hysteresis profile under the change of
applied frequencies. Results (Fig. 4(e)) showed that the SMAM
exhibited similar hysteresis relationships under various frequen-
cies ranging from 0.2 Hz to 1.5 Hz, which are well-suited for
many surgical applications [6], [25]. To evaluate the contraction
force of SMAMs, the experiment setup (Fig. 2(d)) was modified
by replacing Encoder 2 with a FUTEK load cell. The SMAM
(OD = 1.45 mm, Li = 50 mm) was elongated to 65, 75, and
85 mm and connected to the load cell via an inextensible cable.
Noting that the positive direction of input motion is the in
depressurized direction, which is opposite to the motion-motion
hysteresis profiles or the one given in [23]. Cable tension was
0.05 N at the beginning of each case. Three cycles were con-
ducted at each elongated length. The results were repeatable and
there is little hysteresis (Fig. 4(g)). Moreover, the SMAM could
extend up to 175 mm (350%) to exert a maximum force of 6.06
N before plastic deformation.
IV. NONLINEAR HYSTERESIS MODEL
In this section, a new hysteresis model is developed to capture
the SMAM hysteresis loops. The review of the generalized
asymmetric Bouc-Wen model can be found in [25]. We here
reintroduce the conventional symmetric Bouc-Wen model with
six model parameters:
ΦBW (x, t) = γx x(t) + γz z(t) (1)
ż(t) = Kẋ(t) − λ |ẋ(t)| |z(t)|n−1 z(t) − δ ẋ(t)|z(t)|n (2)
where x(t) = xin (t) is input displacement, ΦBW (x, t) =
xout (t) is the output of the hysteresis system, and z(t) is the
internal state. The scale factors γx and γz represent the ratio
of output hysteresis to the input displacement and internal state
respectively. K, λ, δ are dimensionless parameters that control
the shape of the hysteresis loop while n regulates the smoothness
transition of the hysteresis profile. Although the symmetric hys-
teresis Bouc-Wen model is simple for implementation, it cannot
accurately fit the actual data because SMAM hysteresis loops
are asymmetric (see next section). Although the generalized
asymmetric Bouc-Wen model [25] can adequately capture the
SMAM hysteresis profile, it requires ten model parameters to be
identified, resulting in more computation time and complicated
control processes in the case of feedforward compensation.
THAI et al.: DESIGN, FABRICATION, AND HYSTERESIS MODELING OF SOFT MICROTUBULE ARTIFICIAL
Fig. 5. Comparison between hysteresis models and experimental data with
0.1 Hz input signal. (a) The symmetric Bouc–Wen model. (b) The asymmetric
Bouc–Wen model. (c) The new asymmetric hysteresis model.
To overcome these above limitations, we develop here a
new asymmetric hysteresis model which has a fewer number
of model parameters compared to the generalized asymmetric
hysteresis Bouc–Wen model while providing higher accuracy
than that of both the symmetric and generalized asymmetric
Bouc-Wen models. It is given by:
⎧ for multi-periodic inputs consisting of two signals (frequencies
⎨ 0 : x(t) <0
αx1 x4 (t)+αx2 x(t)+(αx1 x4 (t)+αx2 x(t)−1)e−2x(t)
ΦS (x, t) =
⎩ 1+e−2x(t)
: x(t) ≥ 0
(3)
 
ż(t) = |ẋ(t)| A sgn(ẋ(t)) − υ|z(t)|n−1 z(t) + ρ (4)
where ΦS (x, t) = xout (t) is output displacement or contraction
force and x(t) = xin (t) is input displacement. The dimension-
less parameters A, υ, ρ and n adjust the shape and smoothness of
the hysteresis loops while αx1 , αx2 and αz represent the ratio of
the hysteresis output and the input displacement and the internal
state z(t). The new model has seven parameters that are identified
and optimized by minimizing the mean square error (MSE) or F
between the model output and the measured experimental data
based on particle swarm optimization (PSO). Detailed results
including the comparison between the new model and other
Bouc-Wen hysteresis models will be given in next sections.
V. RESULTS FOR HYSTERESIS MODELS
As shown in Fig. 5(a), the hysteresis loops obtained from the
conventional Bouc–Wen model are symmetric in both loading
and unloading phases. As a result, the symmetric Bouc–Wen
model is not able to precisely follow the SMAM hysteresis
profile although there is a close fit in the middle of the hys-
teresis loop. However, the error is greater near the two transition
points. The generalized asymmetric Bouc–Wen model, in con-
trast, shows better results with a smaller value of F = 0.0435
(Fig. 5(b)) because it is inherently able to follow asymmetric
hysteresis loops. This model fits the experimental data not only
near two transition points but also in the middle of the loading
and unloading phases.
Authorized licensed use limited to: University of Florida. Downloaded on March 20,2022 at 23:51:25 UTC from IEEE Xplore. Restrictions apply.
5093
Fig. 6. Comparison between the developed model and experimental data.
(a) With two sinusoidal signals. (b) With a non-harmonic signal.
The new model given by (3) and (4), in contrast, offers the
smallest F = 0.0215 compared to the other two well-known mod-
els (Fig. 5(c)). In addition, it has seven model parameters, which
is less than that of the generalized asymmetric Bouc-Wen model.
The hysteresis curve in the new model is asymmetric for the
loading and unloading phases and it has good tracking properties
in following the experimental data. The identified parameters
of the new model are identified and optimized using the PSO
method:αx1 = −0.082, αx2 = 9.886, αz = −1.809, A = 2.981,
n = 1.035, υ = 2.178, ρ = −0.428. To demonstrate the capability
of the new model, we also re-used these identified parameters
of 0.1 Hz and 0.2 Hz). Fig. 6(a) shows that our model precisely
follows the hysteresis curves with F = 0.1058. The new model
was also √ validated with a non-harmonic sequence of 0.1 Hz
and 0.1 3 Hz (F = 0.0769, Fig. 6(b)). From these results, we
can conclude that our model displays a good agreement with
the actual experimental data for both periodic and non-periodic
input displacements.
To show the capability of our hysteresis model to adapt to dif-
ferent velocities, we tested the SMAM elongation with different
sinusoidal inputs of the same amplitude (1.5 mm) but different
frequencies (0.1, 0.15 and 0.2 Hz) or corresponding to 5.34, 8.07,
and 10.75 mm/s, respectively. We used the same set of identified
model parameters for these cases where the value of MSE of
the proposed hysteresis model for each case calculated to be
F0.1Hz = 0.0215, F0.15Hz = 0.0234, F0.2Hz = 0.0389. From these
results, we can conclude that our new hysteresis model works
well and is consistent across various velocities. However, it is
noted that the MSE value is larger at a higher velocity. This can
be explained by a combination of soft components that exhibit
nonlinear characteristics at different working velocities. This
means that our hysteresis model is beneficial for applications that
require low-speed operation such as surgical robots or flexible
medical devices. Furthermore, the proposed model was also used
to capture the hysteresis relationship between the input motion
and contraction force at the 75 mm initial elongated length of
the SMAM (Fig. 4(g)). The model well fitted the experimental
data (Fig. 7).
VI. SMAMS FOR MEDICAL ROBOTIC APPLICATIONS
We present here two preliminary proof-of-concept medical
applications that utilize our SMAMs as the main mode of
actuation.
5094
Fig. 7. Comparison between the developed model and experimental data for
motion vs. contraction force at the 75 mm -initial elongated length.
Fig. 8. Remote pulley driven by two SMAMs. (a) Diagram for a pair of
SMAMs in a flexible tube simulated as flexible endoscope. (b) The pulley at
various positions under control of two SMAMs.
A. SMAM-Like Tendon-Driven Pulley Mechanism
To control each degree of freedom (DOF) in surgical arms
or medical devices, a pair of tendons are often used to actuate
rotational pulleys [6], [7]. We show that our muscles can be
used as tendon-driven pulley mechanisms. We used two SMAMs
(OD = 1.45 mm, Li = 50 mm) to actuate a pulley via a flexible
cable while the hydraulic transmission tubes are long enough to
operate in a flexible channel simulating as a flexible endoscope
(Fig. 8). The rotational motion of the pulley is generated via
the local extension of each SMAMs regardless of the distance
between the SMAM distal end and the liquid source. While the
liquid transmission tube of each SMAM is attached to a syringe,
the distal end of each SMAM is connected to one end of the cable
that transforms muscle translational motion to the rotational
motion of the pulley. The first SMAM is pressurized and the
second SMAM is depressurized at its syringe end, enabling a
counterclockwise rotation of the pulley. In contrast, when the
first SMAM is depressurized and the second SMAM is pressur-
ized, the pulley will reverse its motion. In this configuration, each
SMAM is firstly elongated to 50% before being installed into
the mechanism. Based on the position of each SMAM’s distal
end, the angle of the pulley is then mechanically regulated.
B. SMAM-Driven Flexible Catheter With Surgical Forceps
This sub-section will develop a SMAM-driven miniature
bending arm and surgical forceps that can be integrated into flex-
ible robotic catheters for delivering treatment in confined spaces.
We made asymmetric and rectangular cutouts in a PET tube by a
surgical knife, forming a compliant bending segment that can be
actuated by a SMAM. We also integrated SMAM-driven surgical
forceps into the bending arm to provide an additional DOF for
the arm such as for gripping or cutting tissues. The new arm
is designed in a way that it can outfit any desired catheter or
surgical arm.
Each SMAM separately controls the bending or gripping
motion by pressure regulation in the two syringes via two
liquid transmission tubes, which travel inside a catheter body
(Fig. 9). In this design, each SMAM is elongated by around
Authorized licensed use limited to: University of Florida. Downloaded on March 20,2022 at 23:51:25 UTC from IEEE Xplore. Restrictions apply.
IEEE ROBOTICS AND AUTOMATION LETTERS, VOL. 6, NO. 3, JULY 2021
Fig. 9. SMAM-driven robotic catheter. (a) Illustration with surgical forceps
in the GI tract. (b) Schematic illustration of the fabrication process for the
asymmetric cutout tube. (i) Teflon tube. (ii) Insert a 3D-printed pattern into
the Teflon tube. (iii) Use a sharp knife to remove the unused parts and then
remove the 3D printed pattern. (iv) A final cutout tube. (c) The catheter working
principle. (i) Initial state. (ii) Gripping state. (iii) Bending state.
70% of its original length where its proximal ends are fixed
to the base while its distal ends are connected to the end of
the cutout tube for bending motion or tied to the forceps for
gripping. A higher elongation will result in a stronger contraction
force or a larger bending angle and stronger holder force of the
bending arm and forceps, respectively. Motions are induced by
the contraction forces generated by the SMAMs. For example,
when the pressure in the first SMAM drops, it becomes shorter
and makes the cutout tube bend towards the side with the cutouts
(Fig. 9(c-ii)) or when the second SMAM contracts under the
hydraulic pressure drop, it applies a force on the forceps that
acts to be close. When the second SMAM returns to its initial
elongation, the tension in this muscle reduces and a rubber string
makes the forceps open (Fig. 9(c-iii)). Noting that the tension
in the second SMAM has a slight effect on the bending motion
of the flexible arm. However, this effect does not degrade the
device performances and therefore this can be neglected.
Our proof-of-concept prototype is shown in Fig. 10 where the
flexible bending arm has a diameter of 3.23 mm. We performed
several experiments to test the bending capability of the flexible
arm and the rotational motion of the forceps. Fig. 10(b) shows
that the flexible tip can induce bending motion while opening and
closing the surgical forceps, all are controlled by the SMAMs.
As an illustration, we adjust the syringe plunger of the SMAM1
to actuate the catheter tip to reach an angle of approximately 45o ,
90o and 135o while controlling the SMAM2 to open and close
the forceps. To demonstrate the usefulness of the flexible arm
when working against an object, we also experimented to grip a
determined object, move it by bending the flexible arm, and then
released it at the desired position. Results (Fig. 10(c)) show that
our SMAMs-driven flexible arm could complete the desired task,
demonstrating that the developed arm with integrated forceps
can be used in surgical procedures such as gripping or cutting a
tissue for biopsy in GI tracts such as the colon or small intestine.
THAI et al.: DESIGN, FABRICATION, AND HYSTERESIS MODELING OF SOFT MICROTUBULE ARTIFICIAL
Fig. 10. SMAM-driven robotic catheter with surgical forceps. (a) Prototype
with bending arm and surgical forceps. (b) Bending motion at 45o and 135o with
opening and closing state of the forceps. (c) Manipulation of the arm to grip,
move and release an object. Scale bar: 10 mm.
As exhibited, the catheter tip motion is not affected by
the length and the configuration of the catheter body due to
the benefits of SMAMs (Section III). Therefore, the catheter
body can be optimized for its target applications, especially
at difficult-to-reach internal organs such as the colon or gall-
bladder. Furthermore, the bending curvature, contraction force
and the forceps actuation can be mathematically controlled by
the motion control of two syringes based on the proposed hys-
teresis model-based feedforward compensation [21] where no
additional onboard sensors are needed for this control process.
VII. DISCUSSION, CONCLUSION AND FUTURE WORK
In this letter, we introduced a new concept of soft muscle
that offers several advantages: low weight, scalability, and a
low-cost fabrication method, making it a well-suited actuator
for flexible robots and medical systems such as surgical robotics.
It is scalable with a high aspect ratio and pre-defined Young’s
moduli based on chosen materials to maximize its capabilities
and working area in specific applications. With its scalability,
a unique hysteresis model and control algorithm can be imple-
mented for different muscles at different sizes. In addition, our
prototype is small and lightweight; SMAM (OD = 1.45 mm,
weight = 0.28 g) could lift a load of 50 g (178 times its mass)
or SMAM (OD = 6.35 mm, weight = 51g) could lift a load of
4.72 kg. Although other pneumatic muscles like the one in [31]
with a weight of 1g could lift a load of 12.2 kg, its contraction
ratio was about 0.3, which is ill-suited for surgical robotics. Our
SMAM, in contrast, can elongate up to 245 % of its original
length and safely return to its original length or up to 350%
before having a plastic deformation. Compared to other soft
fluid-driven artificial muscles such as McKibben muscles [7],
[8], [10], our SMAMs are smaller and scalable down to 0.84 mm
in diameter (Fig. 2(a)) with a high aspect ratio of around 2069,
making them great candidates for use in flexible endoscopic
tools and medical devices [18].
Authorized licensed use limited to: University of Florida. Downloaded on March 20,2022 at 23:51:25 UTC from IEEE Xplore. Restrictions apply.
5095
Although the idea of utilizing the accumulated elastic energy
to contraction force when releasing the pressure has been re-
ported [2], the use of insertion method with outer helical micro-
coil as a constraint layer gives our SMAM a high aspect ratio,
smaller, stronger, more stable, and durable muscle compared to
other conventional fiber-reinforced approaches where manual
wrapping of inextensible fiber are mainly used. In addition,
the helical coil serves as a protective layer that subsequently
increases the fatigue life of SMAM, which is a critical factor
for other muscles such as McKibben-like muscles to be used
in surgical applications. The developed SMAMs also improve
upon conventional TSMs, with minimal performance loss even
in tortuous paths. We also demonstrated that the SMAM hys-
teresis profiles were constant regardless of their configuration
and applied frequencies, evidenced by a constant hysteresis
loop (Fig. 4). TSMs, in contrast, exhibit a strong variation in
the hysteresis profile under these changes, due to non-linear
frictional effects when working against tortuous or complex
paths [32]. The SMAM also has a stable profile across high
frequencies and velocities of the applied input motion. The small
(0.4 dB) loss between input frequencies of 0–3 Hz (Fig. 4(a))
can be explained by the variance of white noise in the frequency
response as well as the elastic deformation of material softness
within the SMAM structure [30].
The asymmetric hysteresis model developed in this letter
provides a better fit to SMAM hysteresis loops when compared
against traditional Bouc-Wen models [20] with fewer model
parameters. In addition, the new model correctly captured the
asymmetric hysteresis relationship not only with periodic and
multi-periodic motions but also for non-harmonic motions (typi-
cal to surgeons’ motions) and at different working velocities. The
model also fitted experimental data of motion versus contraction
force, indicating the potential for force control in future works
using inverse multiplicative structure [22] or feedforward con-
trol [21]. These results showed that our SMAM and its hysteresis
model are crucial for many applications where the integration
of sensors at the muscle tip is complicated and requires extreme
effort [25]. Results from Fig. 8 also proved that the SMAM-like
tendon pulley mechanism can be employed in flexible robotic
systems where pulleys are the dominant candidate to produce
rotational motions. Moreover, the use of the SMAMs to drive a
flexible bending arm and forceps demonstrates that our SMAM
is a great candidate for flexible surgical robots that operate in
tortuous paths. The use of an asymmetric cutout tube has the
most significant advantage of a long moment arm, requiring low
actuation forces for a tight radius of curvature while providing
facile manufacturing with low-cost components.
Despite advances, the SMAM still has several limitations.
First, like all fluid-driven actuators, the SMAM and its pressure
source must be airtight to avoid the introduction of air bubbles.
Second, the use of a helical coil causes high pressure during
its operation, which requires a strong driving pressure source.
However, the fluid volume used is small (normally less than
0.5 ml), and hence enables a compact control unit towards the
achievement of miniature systems. Third, the SMAM shows a
nonlinear hysteresis profile between its input and output under
a known force range. However, the developed hysteresis model
can well capture this phenomenon and therefore benefiting fu-
ture hysteresis model-based control compensation. In addition,
5096
the developed flexible surgical arm with surgical forceps has not
been demonstrated the force capability when gripping against
objects as well as its tracking performance. However, the main
aim of this letter was to demonstrate the capability of SMAMs to
be potentially used with surgical robotic systems and therefore
it is limited in a proof-of-concept. It is noted that the hysteresis
profile will vary under the change of external load or distur-
bances, which requires a nonlinear adaptive control algorithm
with online output feedback. Other unmodelled nonlinearities
will be treated as disturbances and uncertainties where they
are assumed to be bounded. In future work, we will focus on
the development of hysteresis model-based feedforward control
algorithms (when the output feedback is not available) and non-
linear adaptive controller (when the output feedback is available
from soft sensors embedded directly into the SMAM). These
controllers have successfully been developed in our previous
works [21], [28]. These controllers are expected to significantly
improve the tracking capability of the SMAM to follow a desired
motion and force trajectory. We will also re-design the flexible
surgical arm with validated tracking performances, holding force
and bending capability. We will scale up the SMAMs to be used
in emerging applications beyond the sphere of surgical robots,
including soft wearable haptic devices and rehabilitation robots.
REFERENCES
[1] B. Gorissen, D. Reynaerts, S. Konishi, K. Yoshida, J. W. Kim, and
M. De Volder, “Elastic inflatable actuators for soft robotic applications,”
Adv. Mater., vol. 29, no. 43, p. 1604977, 2017.
[2] E. W. Hawkes, D. L. Christensen, and A. M. Okamura, “Design and
implementation of a 300% strain soft artificial muscle,” in Proc. IEEE
Int. Conf. Robot. Automat., 2016, pp. 4022–4029.
[3] E. T. Roche et al., “A bioinspired soft actuated material,” Adv. Mater.,
vol. 26, no. 8, pp. 1200–1206, 2014.
[4] J. M. Pinto, E. A. R. d. Lima, J. C. Pinto, and C. Rolanda, “Natural
orifice transluminal endoscopy surgery: A review,” World J Gastroenterol.,
vol. 17, no. 33, pp. 3795–3801., 2011.
[5] N. Kumar, J. Wirekoh, S. Saba, C. N. Riviere, and Y.-L. Park, “Soft
miniaturized actuation and sensing units for dynamic force control of
cardiac ablation catheters,” Soft Robot., pp. 59–70 2020.
[6] M. T. Thai, P. T. Phan, T. T. Hoang, S. Wong, N. H. Lovell, and T. N. Do,
“Advanced intelligent systems for surgical robotics,” Adv. Intell. Syst.,
vol. 2, no. 8, 2020, Art. no. 1900138.
[7] H. M. Le, T. N. Do, and S. J. Phee, “A survey on actuators-driven surgical
robots,” Sensors Actuators A: Phys., vol. 247, pp. 323–354, 2016.
[8] J. Zhang et al., “Robotic artificial muscles: Current progress and future
perspectives,” IEEE Trans. Robot., vol. 35, no. 3, pp. 761–781, 2019.
[9] P. T. Phan, M. T. Thai, T. T. Hoang, N. H. Lovell, and T. N. Do, “HFAM:
Soft hydraulic filament artificial muscles for flexible robotic applica-
tions,” IEEE Access, vol. 8, pp. 226637–226652, 2020, doi: 10.1109/AC-
CESS.2020.3046163.
[10] L. Hines, K. Petersen, G. Z. Lum, and M. Sitti, “Soft actuators for small-
scale robotics,” Adv. Mater., vol. 29, no. 13, 2017, Art. no. 1603483.
[11] T. N. Do, H. Phan, T.-Q. Nguyen, and Y. Visell, “Miniature soft electro-
magnetic actuators for robotic applications,” Adv. Funct. Mater., vol. 28,
no. 18, 2018, Art. no. 1800244.
[12] C. S. Haines et al., “Artificial muscles from fishing line and sewing thread,”
science, vol. 343, no. 6173, pp. 868–872, 2014.
Authorized licensed use limited to: University of Florida. Downloaded on March 20,2022 at 23:51:25 UTC from IEEE Xplore. Restrictions apply.
IEEE ROBOTICS AND AUTOMATION LETTERS, VOL. 6, NO. 3, JULY 2021
[13] J. M. Jani, M. Leary, A. Subic, and M. A. Gibson, “A review of shape
memory alloy research, applications and opportunities,” Mater. Des.
(1980-2015), vol. 56, pp. 1078–1113, 2014.
[14] H. M. Le, L. Cao, T. N. Do, and S. J. Phee, “Design and modelling of a
variable stiffness manipulator for surgical robots,” Mechatronics, vol. 53,
pp. 109–123, 2018.
[15] F. Daerden and D. Lefeber, “The concept and design of pleated pneumatic
artificial muscles,” Int. J. Fluid Power, vol. 2, no. 3, pp. 41–50, 2001.
[16] T. T. Hoang, P. T. Phan, M. T. Thai, N. H. Lovell, and T. N. Do,
“Bio-Inspired conformable and helical soft fabric gripper with variable
stiffness and touch sensing,” Adv. Mater. Technol., vol. 5, no. 12, 2020,
Art. no. 2000724.
[17] F. Daerden and D. Lefeber, “Pneumatic artificial muscles: Actuators for
robotics and automation,” Eur. J. Mech. Environ. Eng., vol. 47, no. 1,
pp. 11–21, 2002.
[18] M. De Volder, A. Moers, and D. Reynaerts, “Fabrication and control of
miniature mckibben actuators,” Sensors Actuators A: Phys., vol. 166, no. 1,
pp. 111–116, 2011.
[19] A. J. Veale, S. Q. Xie, and I. A. Anderson, “Characterizing the peano
fluidic muscle and the effects of its geometry properties on its behavior,”
Smart Mater. Structures, vol. 25, no. 6, 2016, Art. no. 065013.
[20] S. Li, D. M. Vogt, D. Rus, and R. J. Wood, “Fluid-driven origami-
inspired artificial muscles,” Proc. Nat. Acad. Sci., vol. 114, no. 50,
pp. 13132–13137, 2017.
[21] J. Peters et al., “Actuation and stiffening in fluid-driven soft robots using
low-melting-point material,” in Proc. 2019 IEEE/RSJ Int. Conf. Intell.
Robots Syst. 2019, vol. 2019, pp. 4692–4698.
[22] H.-C. Fu et al., “Interfacing soft and hard: A spring reinforced actuator,”
Soft Robot., vol. 7, no. 1, pp. 44–58, 2020.
[23] T. Vo-Minh, T. Tjahjowidodo, H. Ramon, and H. Van Brussel, “A new
approach to modeling hysteresis in a pneumatic artificial muscle using the
Maxwell-slip model,” IEEE/ASME Trans. mechatronics, vol. 16, no. 1,
pp. 177–186, Feb. 2010.
[24] V. Hassani, T. Tjahjowidodo, and T. N. Do, “A survey on hysteresis
modeling, identification and control,” Mech. Syst. Signal Process., vol. 49,
no. 1-2, pp. 209–233, 2014.
[25] T. N. Do, T. Tjahjowidodo, M. W. S. Lau, T. Yamamoto, and S. J. Phee,
“Hysteresis modeling and position control of tendon-sheath mechanism
in flexible endoscopic systems,” Mechatronics, vol. 24, no. 1, pp. 12–22,
2014.
[26] M. Rakotondrabe, “Bouc–Wen modeling and inverse multiplicative struc-
ture to compensate hysteresis nonlinearity in piezoelectric actuators,”
IEEE Trans. Automat. Sci. Eng., vol. 8, no. 2, pp. 428–431, Apr. 2010.
[27] M. T. Thai, T. T. Hoang, P. T. Phan, N. H. Lovell, and T. N. Do, “Soft mi-
crotubule muscle-driven 3-Axis skin-stretch haptic devices,” IEEE Access,
vol. 8, pp. 157878–157891, Aug. 2020.
[28] T. N. Do and Y. Visell, “Stretchable, twisted conductive microtubules
for wearable computing, robotics, electronics, and healthcare,” Sci. Rep.s,
vol. 7, no. 1, pp. 1–12, 2017.
[29] T. H. Nguyen et al., “An effective method to improve the accuracy of
a Vernier-type absolute magnetic encoder,” IEEE Trans. Ind. Electron.,
Jun. 2020, doi: 10.1109/TIE.2020.3000087.
[30] M.-T. Thai, H. Ahmed, S.-C. Hong, J.-R. Lee, and J.-B. Ihn, “Broadband
laser ultrasonic excitation and Multi-band sensing for hierarchical auto-
matic damage visualization,” Int. J. Aeronautical Space Sci., vol. 20, no. 4,
pp. 913–932, 2019.
[31] N. D. Naclerio and E. W. Hawkes, “Simple, low-hysteresis, foldable, fabric
pneumatic artificial muscle,” IEEE Robot. Automat. Lett., vol. 5, no. 2,
pp. 3406–3413, Apr. 2020.
[32] T. N. Do, T. Tjahjowidodo, M. W. S. Lau, and S. J. Phee, “Adaptive control
for enhancing tracking performances of flexible tendon–sheath mechanism
in natural orifice transluminal endoscopic surgery,” Mechatronics, vol. 28,
pp. 67–78, 2015.