Robotics undergoes constant evolution thanks to the steady emergence of new and better technologies. The common conception of robotics is that of the large, rigid mechatronic systems implemented in modern manufacturing, humanoid robotics and remote exploration systems. However, robotics also applies to advanced prosthetics that will allow us to replace lost or damaged limbs. The main limitation is the availability of small, discrete devices that can provide mobility to these robotic limbs. Most people are familiar with servos, DC motors, pneumatic and hydraulic systems. Applying these devices to prosthetics is not as easy as one would think, especially when large hydraulic pumps are required for their operation.

Research has been done on soft bio-mimetic robotic systems that offer more of the benefits of biological systems regarding efficiency, durability, and combinability with existing biological substrates. Some examples of such bio-mimetic actuator systems are Electro-Active Polymers (EAP), Dielectric Elastomer Actuators (DEA) and Pneumatic Artificial Muscles (PAM). The Ionic-EAP depends on the movement of ions by an electric field to produce a large bending motion under a very low voltage (Jain et al. 2014; Tang et al. 2016). A significant problem with these devices is their relatively low force output, which limits their usefulness in loaded applications like an arm or leg. The DEA, on the other hand is another type of EAP actuator that gets very close to the requirements of ideal biomechatronic systems. This device is placed between electrodes and actuated by an electric field. By combining multiples of these devices in a rolled bundle, it is possible to achieve virtually any desired motion (Pei et al. 2004; Kovacs et al. 2009). However, these systems require substantial electrical input and currently do not scale well to large assemblies, not to mention the relative fragility of their active electrodes.

Research on artificial muscle actuators driven by air pressure were shown to have a wide variety of output forces and motion (Martinez et al. 2012; Luo et al. 2015). From the results of the published work, it is evident that these PAMs are ideal actuators for bettering prosthetic systems in terms of actuation force, flexibility, and scalability. However, these pneumatic systems rely on bulky, loud and energy demanding hardware making them impractical for portable prostheses.

A research group from the University of Texas at Dallas (Haines et al. 2014) demonstrated that by using inexpensive polymer fibers such as store-bought monofilament fishing line, and sewing thread it is possible to build artificial muscles that are fast, strong, and reliable. These actuators are constructed by applying extreme twisting to the filament until a helix is formed. According to Haines et al. (2014) these coiled muscles can contract up to 49% of their original length and lift loads 100 times heavier than human muscle of the same length and weight. These results have sparked much interest in researching the characteristics and behavior of these novel, simple, and affordable devices. Aziz et al. (2015) analyzed and characterized twisted yarns and fibers and presented different test methods for these coiled polymer actuators, and the theoretical results were compared against the experimental results. Arjun et al. (2016) presented the design and fabrication of 3D printed hand prosthesis actuated using twisted nylon 6–6. Their results show that it is possible to implement twisted artificial muscles to develop strong, flexible, and affordable prosthetics. However, more research is required to reduce their operating temperatures to make them viable for practical applications.

In order to achieve reliable results from the analysis it is necessary to consistently fabricate these devices. For that reason, a machine that will aid in the fabrication of Twisted Coiled Polymer Actuators (TCPAs) was designed and constructed. By implementing a PID control system in conjunction with a force sensor, it is possible to maintain the tension of the fiber at a constant value and control the muscle parameters. This paper describes the fabrication of the device and the different elements that compose it. The ultimate goal is to disseminate this design and make it available to every research group interested in fabricating their own.

The Muscle Twister.

The initial TCPA manufacturing process in this work was the same as the one used by other researchers (Cho et al. 2016; Masuya et al. 2017; 2018). A piece of Nylon 6,6 was suspended vertically with a weight attached at the end of the fiber. Using a stepper motor the fiber was twisted until a helix started to appear creating the coiled polymer. This is a very delicate process. If the weights are not properly selected, the fiber will either break or create knots. Another problem that occurs with the twisting process is the up and down jerking of the weight. If the twisting speed is not adjusted properly, knots are generated in the fiber, and the TCPA is basically destroyed. Once a working procedure was developed, we were able to fabricate and test our own muscles. For experimentation purposes, it was necessary to fabricate multiple samples with consistent parameters that will allow for proper characterization of the device. The diameter of the coil and its pitch angle affects mainly the performance of the device (Abbas & Zhao 2017; van der Weijde et al. 2017). These parameters can be controlled by adjusting the weights and the twisting speed. On the other hand, there are a number of parameters that cannot be controlled during the manufacturing including human intervention. Therefore, an automated twisting machine was designed.

The device or ”Muscle Twister” (Figure 1) as it was named by the research team, is composed of the following elements: 1) end plates and motors, 2) carriage and rails, 3) slider, 4) sensor board, and 5) control system. Most of the parts of the muscle twister were fabricated using a repurposed 3D printer and scrap aluminum.

Figure 1.

Muscle twister full assembly.

Figure 1.

Muscle twister full assembly.

End plates and motors.

The two end plates of the muscle twister are the main structures of the device. They are slightly different, because of their very specific role. Figure 2a presents a 3D rendering of both end plates. The support end plate is designed to hold the linear rails and the lead screw. Set screws are used to fix the rails to the plate. The motor end plate has an extra set of holes to attach the stepper motors. Figure 2b shows how the motors are attached to the end plate. The motor on the top is the twisting motor, while the motor at the bottom is the carriage displacement motor. The control system was designed to adjust the tension of the fiber by adjusting the position of the carriage, while the twisting motor was set to a fixed speed. The assembly of the structure needs to be done very carefully to keep everything square and prevent jamming of the carriage.

Figure 2.

a) Muscle twister end plates, b) Stepper motors mounted to the end plates.

Figure 2.

a) Muscle twister end plates, b) Stepper motors mounted to the end plates.

Carriage and rails.

The carriage and rails are essential to keeping the tension of the fiber constant. The carriage is attached to the linear rails using linear bearings and the lead screw is connected to the carriage using a nut block. As the lead screw is turned by the stepper motor, the carriage will move in or out, decreasing or increasing the tension on the fiber. The sensor board and slider are placed on top of the carriage. Figure 3 shows a 3D rendering of the carriage.

Figure 3.

Front and back of the muscle twister carriage.

Figure 3.

Front and back of the muscle twister carriage.

Slider.

The slider is the element of the system that is used to transfer the fiber tension into the sensor board. Figure 4 shows how the slider is made. On the front part of the slider a slit was machined and a set screw was used to hold the fiber. By using polished circular rods, the slider delivers pressure directly into the sensor, allowing the system to measure the tension on the fiber in real time.

Figure 4.

Muscle twister slider.

Figure 4.

Muscle twister slider.

Sensor board.

The sensor board was designed to measure the tension on the fiber by using a Honeywell Low Profile Force Sensor (FSS-SMT), and a High Accuracy Instrumentation Amplifier (INA101). A Molex terminal block was used to connect the sensor board to the control board, and a trimmer resistor was used to zero out the no load state. A schematic of the sensor board is shown in Figure 5. The printed circuit board was designed to accept surface mount components to reduce space, except for the terminal block which is a through-hole component. Surface mount components were also selected to prevent a short circuit when the board is in contact with the support place.

Figure 5.

Sensor board schematic.

Figure 5.

Sensor board schematic.

An Arduino Mega with a stepper motor shield was used for the control system. A PID control system was used to maintain the tension on the fiber as constant as possible. A block diagram of the control system is shown in Figure 6.

Figure 6.

Feedback system used to control the tension by displacing the carriage

Figure 6.

Feedback system used to control the tension by displacing the carriage

The user defines the desired tension. This value is equivalent to the amount of weight placed at the end of the fiber when it is placed vertically. Once the tension has been set, the control system will adjust the carriage position until the desired tension is achieved. When the system is initially powered up, the twisting motor will be idle, and only the carriage will be adjusted. Once the desired tension is achieved, the twisting motor will start working at a set speed. As the fiber gets twisted, the tension on the fiber will start to increase. As a result, the carriage will move in, to keep the tension constant. As the fiber gets super-twisted, a helix will start to form. This will be observed as a rapid increase in tension. The combination of these helices and the movement of the carriage produces the desired TCPAs. Once the first coil is created, the other coils will start to form. Figure 7 shows an image of the oscilloscope used to monitor the tension during the fabrication process. This shows how the PID controller is maintaining the tension at a constant value. The most difficult part to control is the coiling process. This generates a sudden spike in tension right before the coil is completely formed. Once the coil has settled the tension drops, and the controller adjusts the tension as needed. At this moment, the effects of different twisting speeds hasn’t been analyzed. However, the twister is fully capable of achieving different speeds for the twisting motor, and the carriage. The main limitation will be how fast the system will be able to react to the changes in tension during the twisting process.

Figure 7.

Force sensor output voltage. The image on top represents the setting of the initial tension. The image at the bottom represents the coiling process.

Figure 7.

Force sensor output voltage. The image on top represents the setting of the initial tension. The image at the bottom represents the coiling process.

Attaching the muscle.

When trying to fabricate a muscle, ideally the same amount (length) of material needs to be used to achieve similar results. On a first attempt, butt splices were cut in half, a loop was created and crimped in place. However, when the fiber was twisted it always snapped in the same place. After multiple attempts and review of the results, it was observed that at the point where the butt splices were crimped, the nylon had no room to move freely. As a result, a lot of stress was concentrated at a single point during the twisting. The solution to the problem was to create a loop knot at one end of the fiber, and use a butt splice at the other end to adjust the length. It seems that having a knot at one end allows for better distribution of forces.

Once both loops were made, one of them was inserted into the slit of the slider plate and secured with the screw. The second loop was secured to the twisting motor using a custom-made clamp designed by the machinist. Figure 8 shows an image of a piece of nylon 6,6 attached to the muscle twister ready to be twisted into a TCPA.

Figure 8.

The image on the left shows a piece of nylon 6,6 attached to the muscle twister. The image on the right show the finished TCPA attached to the muscle twister.

Figure 8.

The image on the left shows a piece of nylon 6,6 attached to the muscle twister. The image on the right show the finished TCPA attached to the muscle twister.

Sensory board characterization.

Once the muscle twister was finished the next step was to create a relationship between the voltage obtained from the sensor board, and the equivalent weight that will be attached to the fiber. This characterization was achieved by placing the Muscle Twister vertically, and attaching different weights directly to the slider. Figure 9 show how these measurements were performed. Table 1 and Figure 10 show the results obtained from the characterization. As expected, the force sensor had a linear behavior.

Figure 9.

Setup for the force sensor characterization using known weights.

Figure 9.

Setup for the force sensor characterization using known weights.

Figure 10.

Characterization of the force sensor as a function of mass and voltage.

Figure 10.

Characterization of the force sensor as a function of mass and voltage.

Table 1

Mass vs. digital voltage reading.

Mass vs. digital voltage reading.
Mass vs. digital voltage reading.
Using the measured data, the trend line polynomial from 1 was obtained.  
formula
This equation allows us to predict the ideal digital voltage for any equivalent weight attached at the end of the fiber. For example, to maintain a tension produced by a 250 grams weight, the digital voltage coming out of the sensor board should be approximately equal to 94.

Muscle samples.

The performance of the muscle twister was characterized by fabricating samples at different tensions. Table 2 shows the number of samples for each tension and the equivalent weight that will need to be hung from the end of the fiber to produce the same results. A 5N force sensor was used for low tensions and a 10N force sensor was used for the higher tensions.

Table 2

Settings used to fabricate the different samples.

Settings used to fabricate the different samples.
Settings used to fabricate the different samples.

Once the samples were fabricated, several measurements were made of the muscles to characterize the performance of the muscle twister. The diameter of the coil was measured using digital calipers. The values from the six samples from every weight group were averaged together. Table 3 shows the average diameter and standard deviation from measuring the different samples. It is evident that as the tension applied to the fiber increases, the diameter of the coil gets reduced as the coils get tighter. The second test was made by measuring the pitch angle of the coil. This process was done by placing all samples under a camera with a macro lens, and manually measuring the pitch angel using image-processing software. The pitch angle for each sample was measured at three different positions, and then averaged. The averaged values for all six samples in the weight group were averaged one more time to obtain a pitch angle for the group. Table 3 shows the results from measuring the pitch angle for all weight groups. Clearly, as the coil gets tighter and smaller, the pitch angle increases.

Table 3

Results obtained from measuring the characteristics of the different samples.

Results obtained from measuring the characteristics of the different samples.
Results obtained from measuring the characteristics of the different samples.

All the previous results are consistent with what was expected from the fabrication process. However, one of the most interesting results is the physical shape of the coils. Figure 11 shows the picture of a coil for each weight group. It can be seen that for low tensions, the coils are not uniform in diameter, but as the tension increases their diameter becomes more consistent. Also, notice how the coils with lower tension have a larger diameter compared to the coils made at higher tension. Based on observations made during the twisting process, it is hypothesized that during the creation of a loop the tension increases by the same amount regardless of the twisting tension setup in the system. As a result, this increase in tension has a more pronounced effect on lower twisting tensions, creating variation in the diameter of the TCPA coils.

Figure 11.

Samples of the five different weight groups. The coils are organized from left to right starting with the ones made with the lowest tension going up to the highest tension.

Figure 11.

Samples of the five different weight groups. The coils are organized from left to right starting with the ones made with the lowest tension going up to the highest tension.

The performance of the muscles fabricated with the muscle twister was analyzed by measuring their contraction as their temperature raises from room temperature to 100 degree Celsius. This was achieved by placing the muscles inside an environmental chamber, and using a Linear Variable Differential Transformer (LVDT). The LVDT reading for six samples fabricated using the same tension were collected and plotted. Figure 12 shows the contraction of the six samples. It can be seen how four of the six samples are very similar, and the other two are very close. The research team believes that these small differences are due to variations during the annealing process of the muscles, and more research needs to be done to identify the ideal annealing process. In future research, the team will work on a detail analysis of the performance of these devices, using mathematical models that will involve the different characteristics found on TCPAs.

Figure 12.

Characterization of TCPAs fabricated using the muscle twister with a tension of 600 grams.

Figure 12.

Characterization of TCPAs fabricated using the muscle twister with a tension of 600 grams.

Conclusion.

This paper presented the design and implementation of an automated muscle twisting machine. The system implemented a PID control system to maintain the tension during the twisting process. In order to test the performance of the device, multiple samples were made using different tension values. The pitch angle and diameter of the different samples were measured to quantify the performance of the muscle twister. The results showed that it is possible to control the pitch angle and diameter of the coil with high accuracy. Been able to control these parameters allow researchers to analyze these devices more accurately, and as a result better mathematical models can be developed.

Acknowledgments.

The project was funded by a Stephen F. Austin State University Faculty Research Pilot Studies Grant.

LITERATURE CITED

LITERATURE CITED
Abbas,
A.
&
Zhao.
J.
2017
.
A physics based model for twisted and coiled actuator
.
IEEE International Conference on Robotics and Automation (ICRA)
,
Singapore
, pp.
6121
6126
.
doi: 10.1109/ICRA.2017.7989726.
Arjun,
A.,
Saharan
L.
&
Tadesse
Y.
2016
.
Design of a 3D printed hand prosthesis actuated by nylon 6–6 polymer based artificial muscles
.
IEEE International Conference on Automation Science and Engineering (CASE)
,
Fort Worth, TX
, pp.
910
915
.
doi: 10.1109/COASE.2016.7743499.
Aziz,
S.,
Naficy
S.,
Foroughi
J.,
Brown
H. R.
&
Spinks
G. M.
2015
.
Characterization of torsional actuation in highly twisted yarns and fibers
.
Polym. Test.
46
(
1
):
88
97
.
Cho,
K. H.,
Song
M.,
Jung
H.,
Yang
S. Y.,
Moon
H.,
Koo
J. C.,
Nam
J.
&
Choi
H. R.
2016
.
Fabrication and modeling of temperature controllable artificial muscle actuator
.
6th IEEE International Conference on Biomedical Robotics and Biomechatronics (BioRob)
,
Singapore
, pp.
94
98
.
doi: 10.1109/BIOROB.2016.7523604.
Haines,
C. S.,
Lima
M. D.,
Li
N.,
Spinks
G. M,
Forough
J.,
Madden
J. D. W.,
Kim
S. H.,
Fang
S.,
de Andrade
M. J.,
Göktepe
F.,
Göktepe
Ö.,
Mirvakili
S. M.,
Naficy
S.,
Lepró
X.,
Oh
J.,
Kozlov
M. E.,
Kim
S. J.,
Xu
X.,
Swedlove
B. J.,
Wallace
G. G.
&
Baughman
R. H.
2014
.
Artificial muscles from fishing line and sewing thread
.
Science.
343
(
6173
):
868
872
.
Jain,
R. K.,
Datta
S.
&
Majumder
S.
2014
.
Biomimetic behavior of IPMC using EMG signal for micro robot
.
Mech. Based Des. Struct.
42
(
3
):
398
417
.
Kovacs,
G.,
Dring
L.,
Michel
S.
&
Terrasi
G.
2009
.
Stacked dielectric elastomer actuator for tensile force transmission
.
Sensor. Actuat. A-Phys.
155
(
2
):
299
307
.
Luo,
M.,
Skorina
E. H.,
Tao
W.,
Chen
F.
&
Onal
C. D.
2015
.
Optimized Design of a Rigid Kinematic Module for Antagonistic Soft Actuation
.
IEEE International Conference on Technologies for Practical Robot Applications (TePRA)
,
Woburn, MA
, pp.
1
6
.
doi: 10.1109/TePRA.2015.7219694.
Martinez,
R. V.,
Fish
C. R.,
Chen
X.
&
Whitesides
G. M.
2012
.
Elastomeric Origami: programmable paper-elastomer composites as pneumatic actuators
.
Adv. Funct. Mater.
22
(
7
):
1376
1384
.
Masuya,
K.,
Ono
S.,
Takagi
K.
&
Tahara
K.
2017
.
Nonlinear dynamics of twisted and coiled polymer actuator made of conductive nylon based on the energy balance
.
IEEE International Conference on Advanced Intelligent Mechatronics (AIM)
,
Munich Germany
, pp.
779
784
.
doi: 10.1109/AIM.2017.8014112.
Masuya,
K.,
Ono
S.,
Takagi
K.,
&
Tahara
K.
2018
.
Feedforward control of twisted and coiled polymer actuator based on a macroscopic nonlinear model focusing on energy
.
IEEE Robot. Autom. Lett.
3
(
3
):
1824
1831
.
Pei,
Q.,
Rosenthal
M.,
Stanford
S.,
Prahlad
H.
&
Pelrine
R.
2004
.
Multiple-degrees-of-freedom electroelastomer roll actuators
.
Smart Mater. Struct.
13
(
5
):
N86
N92
.
Tang,
Y.,
Xue
Z.,
Xie
X.
&
Zhou
X.
2016
.
Ionic polymer-metal composite actuator based on sulfonated poly(ether ether ketone) with different degrees of sulfonation
.
Sens. Actuat. A-Phys.
238
(
1
):
167
176
.
van der Weijde,
J.,
Smit
B.,
Fritschi
M.,
van de Kamp
C.
&
Vallery
H.
2017
.
Self-sensing of deflection, force, and temperature for joule-heated twisted and coiled polymer muscles via electrical impedance
.
IEEE-ASME T. Mech.
22
(
3
):
1268
1275
.