Preliminary Actuator Design

While creating silicone-based actuators using the PnueNets, we encountered issues when attempting to seal the two portions of the mold. Due to the two parts of silicone curing separately, the seal between the bottom layer and the actuator itself was not completely air-tight, resulting in its delamination. In consideration for a simpler process and reliable actuator, we noted features we would like to have for an ideal actuator.

A continuous-pour mold was one of our priorities, as such a process would reduce the time required to fabricate an actuator and eliminate the issue of delamination. However, this brought the challenge of forming the actuator's insert structure, which needed to be removed from the mold in a single motion.

Edward King, a student at our school, conceptualized a system where the inserts inside the mold could be dissolved, creating an air pocket for actuation. King considered using materials such as polystyrene (styrofoam) for such a process. Another benefit we considered in our design of the insert-dissolving method was that the air pockets could now have more complex geometries. A two piece mold limits one’s air pockets to the shape of a rectangular prism. Different geometries such as circles, trapezoids, and parallelograms would now be possible.

Once we formulated the first design and a potential application to test the actuators, a styrofoam insert was manually cut and fit to a simple rectangular mold. We poured a different, transparent material, polydimethylsiloxane (PDMS), into the test cardboard mold so we could physically view the dissolving process. Upon first glance, the large deviation between the densities of each material caused the styrofoam to rotate and float upwards, rendering the actuator unusable in practice. However, as expected, the dissolving of the insert using acetone through a manually poked hole was quick and simple.

Figure 6: Polystyrene insert in clear PDMS mold

Another observation we made was that the manually poked hole (created previously using a Wilton brand cookie flooding etching tool) should, ideally, be built into the mold. This would be done to create consistency from actuator to actuator, to aid in airflow during actuation, and to help dissolve the Insert. After the first test, we considered using fabric on one end of the mold as an inextensible layer to help constrict that side, reducing the overall vertical volume of the actuator. After we finished the simple prototype to gain preliminary information about the style of actuator, we began our iteration on the first version.

Actuator Design Iteration and Fabrication

Version 1:

Once we realized the amount of precision needed when creating these actuators, we switched to using 3D-Printed PLA (polylactic acid) for the base molds, and PVA (polyvinyl alcohol) for the inserts. PVA is water soluble which ideal because a soluble insert allows us to utilize a one-pour process. We can remove the PVA using water after the silicone is cured, leaving the connected air chambers which allow for the inflation and bending of the actuators. Using PLA, we could mark each iteration of mold dimensions with color-coding printer filament. Also, another incentive for using PVA was to test a different material than stryofoam, as our tests with stryofoam were successful. These materials are easy to manipulate as we could 3D-print them and easily change their dimensions. We also transitioned to using EcoFlex™ 00-50 Two-Part Silicone for the actuator structure.

(See the Bill of Materials for more detailed information regarding the materials we used)

Figures 7, 8, and 9: 3D printed molds, a 3D printed insert, and completed V1 actuators

Dimensions:

- Orange Mold: 16.70 x 105.60 x 15.88 mm

- PVA Insert: 12.70 x 88.90 x 9.53 mm

Fabrication Process for First Iteration:

1. Lay down a single layer of fabric

2. Pour the mixed silicone halfway up the depth of the base mold

3. Place the insert centered in the mold

4. Fill rest of the mold with silicone

5. Wait for the mold to cure

6. Remove by pulling up the sides by hand

7. Dissolve the insert

In contrast to the styrofoam inserts, the PVA roughly maintained its position relative to the mold and did not float at all. Another key observation we made was about the dissolving process of the PVA. Before using the compound we tested the water solubility of the PVA in an experiment. Although it was apparent the material required an extensive amount of time to dissolve, the data was not representative of the actual dissolving process inside an actuator. First, it was difficult to expose the entirety of the PVA insert to the heated water through a minuscule hole. In doing so we tore the side walls of the actuator from inserting the nose of the pipet and creating pressure with water.

Version 2:

In the next iteration, we focused on changing the dimensions so there was less height and more width. The extra width added material to the sides so the actuator would not tear, and we compensated for less overall height with more layers of fabric. We also decided to switch back to the polystyrene inserts as they with our current methods could dissolve. However, still challenged with the floating issue of the styrofoam, we noted the inserts and constricting fabric could vertically swap their position such that, if the inserts floated upwards, they would only move towards the side with the fabric. 

Figure 10: Swapping the position of the fabric and inserts

Dimensions:

- Yellow Mold: 23.05 x 105.80 x 12.70 mm

- PVA Insert: 12.70 x 88.90 x 3.18 mm

Fabrication Process for Second Iteration:

1. Fill the mold with mixed silicone

2. Lay two layers of fabric on top

3. Attach strips of velcro

4. Wait for the silicone to cure

5. Dissolve

Version 3:

As we were brainstorming various designs for an actuator to assist one's motion, we decided to use strips of velcro to allow the actuator to be attached to a finger. This would be more versatile than a glove as the as there is no limitation on the user’s hand size.

Figures 11 and 12: Attaching velcro to a mold and velcro attachment testing

(Note: We tested the velcro technique using a PnueNet base in the image above)

Althought the previous actuator iteration was generally successful as it could actuate multiple times, assited in applying force when worn, and did not tear or break while doing so, there were still a few issues. The dissolution hole still had to be manually made, which is counter-intuitive to the goal of simplicity, and makes the dissolving process more tedious. The actuators also, comprising of simply a block with an air channel, had difficulty inflating and bending in the same motion as the PneuNet actuators. What our actuator lacked, which most soft actuators have, are incisions, or slits, to help increase the range of motion. In our case, our actuators had a wide range of motion, but at the cost of the stretching of the silicone. Each air pocket, when inflated, looked and behaved similarly to the inflation of a balloon. The material did not rip due to our solution's elasticity, yet the long-term reliability was questionable. This led us to a design that utilized sectioned air pockets and attachable inserts that could be anchored by the mold.

Figure 13: 3D models of the final mold and insert

In between the iterations one series of the 3D printed mold suffered from scaling issues where the size of the parts did not match, one series lacked the vertical indents, and one packed the air pockets too closely.  

Figures 14, 15, and 16: a 3D printed mold and claws, a fabrication error, and a completed actuator

Of the final generation, we could only completely remove the PVA insert of one the molds through a painstaking process. We decided that if we were to continue, and use the actuator for applications like a motion assisted glove, we should test casting inserts into a mold, and then dissolving said inserts.

Dimensions:

- Slot Mold: 25.40 x 123.83 x 15.88 mm

By the nature of us 3D printing the inserts, we can print inserts that are different shapes and mess with the geometry that way, giving it its modularity. We created actuators using these new geometries, as described in the testing section, but found little difference in actuation.

You can download CAD files for these uniquely shaped inserts here.

Fluidic Control Board Design

We used the fluidic control board instructions with some modifications, as our end application projection required variable PSI output based on information we collected. We controlled the PSI through independent pressure regulators rather than using PWM to have a pressure applied and the end of each fingertip. In addition, we added an LCD to help debug code. As for recording data we tried both Flex and Force sensors, ultimately choosing Force Sensors because we did not care about how much the finger bent, but rather how much force is being applied to the pot.

Figure 17: The schematics and circuit diagram of the fluidic control board

Pressure Sensing Glove Iteration

Using the Pressure Sensing Glove we were able to record the hand motions of an articulate ceramicist. Through trial and error, we created a glove that housed the sensor cable robustly such that it would not interfere with data collection. Originally, we used a Flex sensor hot glued on a latex glove. The glue’s bond was poor and broke near instantly. After switching to the Force Sensor we electrical taped the sensor and extension on the exterior of the glove. Worried that contact between the clay and any electrical components could end up being messy and damage the hardware, we resorted to finally stitching all the components to the interior of the glove. In this manner, all components were hidden and we followed through recording our data.  

Figures 18, 19, and 20: The pressure sensing glove, the sewn on force sensor, and a student testing the sensors

Consumable Materials

Name	Quantity	Image	Purchase Link	(Notes)
EcoFlex™ 00-50 Silicone	~ 10 grams of silicone per actuator		EcoFlex™	We recommend purchasing more silicone than is needed, as it allows for more iteration on the design.
Disposable Dixie Cups	Three (3) per actuator		Dixie Cups
Wooden Stirring Rods	One (1) per actuator		Stirring Rods
Thin Cotton Craft Fabric	Two (2) rectangular 45" by 0.75" sheets per actuator		Craft Fabric	Any relatively thin craft fabric will be sufficient
Disposable Pipettes	Two (2) per actuator		Pipettes	These pipettes will be used to either inject acetone into an actuator made with a polystyrene insert or to funnel hot water into an actuator made with a PVA insert.
1,000 mL beaker	One (1)		Beaker	PVA Insert Only Any size beaker will be sufficient, but make sure that your actuator will fit inside.
Polyvinyl Alcohol (PVA) Filament	~2.2 grams per actuator		PVA Filament	PVA Insert Only Be sure to store this filament in a dry location before printing. PVA is hygroscopic, meaning that it can absorb moisture from the atmosphere, which will effect print quality.
Acetone (Nail Polish Remover)	~ 20 mL per actuator		Acetone	Polystyrene Insert Only Any household nail polish remover will be sufficient.

Scientific Tools / Instruments Used

Name	Quantity	Image	Purchase Link	(Notes)
Craft Knife	One (1)		Craft Knife	Any craft knife (such as a small box cutter or X-ACTO knife) will be sufficient.
Electronic Balance / Scale	One (1)		Balance	Use the balance to weigh the proper proportions of your silicone mixture
PrintrBot Simple 3D-Printer	One (1)		3D Printer	Use a 3D printer to print out the provided Base and Insert molds. Our team used a PrintrBot, but any 3D-Printer with a large enough build volume will be sufficient
Hot Plate	One (1)		Hot Plate	PVA Insert Only You can use the hot plate to dissolve the PVA Inserts inside your actuators. Alternatively, if you do not have access to a hot plate, a standard kitchen stove works just as well.

Electronic and Pneumatic Componenets

See the Soft Robotics Toolkit for more information regarding the Fluidic Control Board

Name	Quantity	Image	Purchase Link	(Notes)
Arduino Mega Development Board	One (1)		Arduino	Any Arduino Development board with a 5V Output and PWM ports will be sufficient
Power Supply	Three (3)		Power Supply	Any power supply capable of supplying up to 12 Volts is sufficient. Ours have a current output of 2A, but you can customize your control board with whatever you have available.
MOSFET 4 v04	One (1)		MOSFET Board
Solenoid Valve	Four (4)		Solenoid
Pressure Splitter Manifold	One (1)		Splitter Manifold
Voltage Monitor	Two (2)		Voltage Monitor
Power Switch	One (1)		Power Switch
LCD Debugging Screen	One (1)		LCD Screen	Use this screen alongside our provided code for debugging and menu selection.
Breadboard Jumper Wires	~50		Jumper Wires	Depending on the setup of your breadboard, you may need mores or fewer jumper wires.
Small Breadboard	Three (3)		Breadboard	The breadboard setup can be customized to the student's preference, but we used small breadboards such as this one.
Pressure Sensor	One (1)		Pressure Sensor	Use this sensor to create a glove that can record pressure data when throwing a pot
Pressure Regulator	Two (2)		Pressure Regulator
Polyester Glove	One (1)		Glove	Use this in the process of creating the pressure sensing glove
⅛” Outer Diameter Pneumatic Tubing	~1 meter used throughout the control board		⅛” Tubing	Use this tubing at your own discretion. Keep in mind that increasing the length of the tube between the pump and the actuator may cause inconsistencies with the pressure outputted.

For a list of necessary materials in spreadsheet format, please see this link.

Modifications to the Fluidic Control Board

Follow the steps for creating the basic Fluidic Control Board on the Soft Robotics Toolkit Page.

The modifications we made to the fluidic control board are represented in this fluid control board schematic. You can use this reference when creating your own control board. Each voltage and data signal wire is color-coded to prevent confusion. Remember to be careful when working with electronics! The low voltages used in this project are still enough to do some harm.

Adding the Pressure Regulators

Instead of connecting the pneumatic tubing from the solenoids directly to the actuator outputs, route the pressure regulators in between. Each regulator requires 5 volts and a PWM data transfer signal to operate. First plug in the ground wire. Then plug in the voltage wire into the 5V source. Finally, plug in the data transfer wires into the PWM ports on your Arduino.

Figure 47: The pressure regulator integrated into the fluidic control board

Adding the LCD Debugging Screen

The LCD Screen requires 5 volts and 8 PWM data transfer ports to operate. First plug in the ground wire. Then plug in the voltage wire into the 5V source. Finally, connect 8 wires from the Arduino into the data ports of the LCD.

Figure 48: The LCD Debugging Screen integrated into the fluidic control board

Sensor Glove

Creating the Pressure Sensing Glove is a simple process requiring only a few key steps.

First, solder a wire onto each pin of the sensor.

Figure 49: Using the Helping Hands to solder wires to the force sensor

Next, place the sensor on the finger of the glove where the pressure will be applied. 

Originally, we had put the sensor on the inside of the glove, believing that the glove would protect it from clay debris. However, we soon realized that because the glove would be protected by an additional nitrile glove, it would be unnecessary to put the sensor on the inside of the glove. Thus, we decided that it would be better to attach the sensor on the outside of the inner-glove for easier access. Special thanks to Mrs. Jacob for helping us sew the glove!

Then, use electrical tape to create a pad on the tip of the finger where the face of the sensor will sit.

Finally, sew the sensor and wires along the length of the finger and down to the bottom of the glove. Once the glove is turned back inside out, it is ready to use! Plug it into the voltage monitor port of your Arduino to start reading values.

Figures 50 and 51: The completed sensor glove freshly sewn and the glove turned inside out

The Infographics depicted below can be used as a quick-reference guide for how to make the original PneuNets and our King Actuators. We have also attached the files for your convenience. 

Why Infographics?

Graphic instructions like this one hold an advantage over traditional written instruction as they allows for more elements of the mind to be stimulated when trying to understand the processes involved. We have found, especially when working with younger students, that using a visual instruction method is much more effective than a written one.

The King Actuator

Solubility Testing

Prior to using the PVA inserts in our actuators, the King Actuator inserts were carved using PS, more commonly known as styrofoam. These inserts had to be cut by hand, a tedious and time consuming process. In addition, the factor of human error made it nearly impossible to ensure that the cells of the insert were identical to each other. This was not ideal because the actuators would never turn out identical, removing the possibility of consistency within our tests. We noticed that PS inserts were less dense than the silicone that covered them, meaning that the inserts would float to the top or shift into undesirable positions during the curing process. However, the PS inserts had one outstanding benefit: they instantly dissolved when they came in contact with acetone. Despite this, the various drawbacks led us to look into a 3D-printed filament that was easily soluble, allowing for a more consistent and reproduceable actuator. After some research, we decided that using PVA as the filament would yield the best results.

A quick comparison of the two materials yields:

Comparative Analysis of Polyvinyl Alcohol and Polystyrene

We conducted a dissolution test on two 3D-printed PVA inserts to gain a better understanding of the rate at which they dissolved. While we knew that the inserts dissolve faster in warmer water due to increased energy in the system, we also wanted to quantify the total time it would take to dissolve and the change in time increased temperature provides. Since the actuators became maleable after exposure to water, gathering quantitaive data such as mass in relation to time was not feasible. Instead, we designed an experiment where we took photos of the two inserts in different temperature solvents during noteworthy occurances and spaced intervals to collect qualitative data. The first hot plate was prehated to 100°C while the other was preheated to 200°C, with the masses of the two PVA internals being nearly identical (2.78 and 2.81 grams, respectively).

Time Elapsed (minutes)	Images	Notes
0		Our testbed with one heatbed preheated to 100°C (Left) and the other at 200°C (Right). The inserts have not been exposed to the water at this point in time.
1		No distinguishable change.
5		No distinguishable change.
10	200C	Visible difference between the 100°C (Left) and 200°C (Right) inserts. The 200°C insert began bubling, dissovling, and slightly arching.
15		The 100°C (Left) insert began arching similar to 200°C insert at 10 minutes. However the water of 200°C (Right) insert became murky, a sign of dissolving.
20		The 200°C (Right) insert arched deeply.
30	100°C	The 100°C (Left) insert began arching similarly.
37	100°C 200°C	The 100°C insert continues to arch. The 200°C however began to arch inwardly, perhaps a sign of little support material the the two points around the clamp.
55	100°C 200°C	The water for both inserts has become translucent. By this time the 200°C insert was completly dissolved.
90		Finally the 100°C insert has dissolved.

This experiment shed light on the benefits of using hotter water, specifically comparing the effectiveness of the two temperatures. However, after we successfully dissolved PVA inserts in our final claw mold, we noted that only the first few cavaties of the insert were exposed to heated water. In essense, only exposed portions of PVA dissolved, forcing us to cycle between dissolving and emptying maleable chunks in order to empty an actuator. While the experiement was successful, the outcome desmonstrated the lack of feasibility of the process in the application. Initially, we judged that the PVA insert was more ideal because of its versatility. However, the dissolution process was more tedious .

If we were to further this project we would do more in depth research in finding a 3D-printable material that completely and instantly dissolves in a solvent. We were considering using Acrylonitrile Butadiene Styrene (ABS) filament instead of the PVA because it dissolves quicker in acetone. The drawback is that we do not have the proper ventilation to safely print the ABS inserts. In addition, we could explore materials that could be cast into a insert mold and later dissolve the insert. This action would ultimately opened our options for more materials aside from 3D-printed ones.

Durability Testing

To confirm the quality of the King Actuator, we performed durability tests comparing various versions of the actuator and a PnueNet. We argued that solid tensile strength and ability to actuate with a high pressure airflow would accurately portray the quality and durability of the actuator. After fabricating actuators of comparable size, we inflated them using a Ryobi Power Inflator until they ruptured or delaminated. Using 240fps slow motion video, we noted PSI changes and when the actuator failed. This data shows us which actuator is able to withstand more pressure and which is the most durable for various applications. 

You can view vidoes of these pressures tests here.

While collecting the data, we experienced inconsistencies in PSI output from the pump because we did not charge the battery between tests for the Final Slit actuator and the PnueNet. As a result, the data may have been scaled differently. To compensate for this, we assumed there was a direct correlation between the maximum recorded PSI and the output PSI, and scaled the data down by a ratio of the maximum PSI of one line to the average of the other two, normalizing the data. From analyzing both data sets and the video, it was evident our single body actuators far outperformed the PnueNets. While the PnueNets delaminated instantly, the other actuators, (The Original King and the Continous Pour with and without slits) survived longer. We were not able to record any data for the PnueNets because it delaminated and burst instantly. However, our actuators still suffered problems. Each of our actuators failed when an air pocket opened between the fabric and silicone. The reason for failure is similar among all of the actuator types; when scaled down, all of our actuators failed at a similar point, further justifying the issue between the seal.

Qualitative Testing

A major component of any pneumatic actuator is its ability to bend with applied pressure. We plotted the curve of the actuators and compared thier general shape.

Original King	Continuous Pour (No Slits)	Continous Pour (With Slits)

All of the actuators arched similarly, although the slotted continous pour expanded within each section, causing the actuator to bend more at the start. In addition, the Continuous Slotted actuator we used had a manufacturing defect where the internal bent the oneside, causing a non-uniform curvature. 

Insert Qualitative Testing

One of the benefits of using a 3D printed insert is that we can change and play with the shape of the cavites formed in the actuator when the PVA is dissolved, giving it its modularity. We 3D printed a set of variable inserts to make actuators out of and see how they impact actuation. We tested some of the internals proposed and noticed no significant visual differences between the different internal shapes in terms of actuation. This is likely due to the fact that as air gets pushed into the cavities it tends to push in the same pattern, thus no matter the shape and size of the cavity the expansion of the cavity will remain the same shape. For the future we would like to create more actuators with these variable shapes and test more volumes out to see if it really has no effect on actuation. However we did notice that the more complicated the shapes got, the harded the actuators were to actuate and remove the inserts from.

Pottery Instruction

As stated, one of the goals of this project was to help teach students to throw a pot more effectively. In order to do this we observed our ceramics teacher throw a pot. The process is highly complicated, including individual finger and arm motion. In order to begin addressing this problem, we worked with Mr. Raeder to simplify the process of throwing a pot to squeezing the edges of a block of rotating clay with two fingers. We figured that if we can replicate the pressure which the teacher is squeezing the clay and output it to actuators attached to the fingers of students, we would be able more effectively teach students how to throw a pot. 

In order to output the pressure to the actuator we had to gather data collected from the teacher throwing a pot. To do that we created a glove with pressure sensors attached the tips of the pointer fingers to read the pressure which the teacher is exerting on the block of clay. Using an arduino control board we created a method we dubbed Active Hand Recording (AHR), in which the the glove, a flex sensor map, is constantly polled at a fixed rate, and the data (the placement of the fingers) from the glove is recorded using the Motion Tracking File/Compressed Motion Tracking File (MCI/CMCI), and the MCI Compression Algorithm (MaRS). The AHR constantly polled data from the glove at a fast and fixed rate meaning that latency and high-level detail were not an issue; however, the AHR system comes at a price.

The Active Hand Recording Analysis

The AHR system works through a three-step process.

1. The Arduino poll information from the glove.

2. The values from the glove are sent to the computer.

3. The computer recording-relatively timestamps the data and save it into a MCI file.

All of this happens at a certain rate, which depends on the Serial baud output rate.

The processing of the AHR data using the MaRS algorithm works through a two-step process.

1. Seek the MCI file for value consistencies within a threshold.

- Example: The hand is motionless for half of a second

2. If a consistency is found, shrink it to one 'average' byte and set the timestamp of the motion(s) accordingly.

- Example: 'The motionless hand at 00:05, 00:05.15, 00.05.43, 00.05.62, 00.06' is shrunk to 'the motionless hand at 00:05-00:06.'

Note: The algorithm can take a long time to work depending on the duration of the motion capture and speed of the processor.

From all of this we are able to retrieve a graph of the pressure as it relates to time (in this case frames).

We built a fluidic control board in order to control the output of air to the actuators. However we struggled with understanding how to control the pressure of the actuators using pressure regulators we purchased. This forced us to simplify the process further which was obviously not ideal. We corrected the noise and the input data into a square wave to make for easier control of actuation. We converted the recorded pressure into a scale with the minimum recorded pressure changed to 0.

 We only used one hand because there was not enough space in the center of the pots to perform controlled actuation. While recording data, sensors were put on both the right and left hands. We noticed that the force applied the outside edge had a greater impact on the straightness of the pot, as on average the right hand, the outside hand, applied more force to the pot than the inside hand. We could not control for the vertical motion of the student, however we believe that after observation, mimicing the vertical position of the hand was intuitive.

This method proved to be ineffective because we oversimplified the process of throwing a pot. Our largest issue was in controlling the fluidic control board's output of pressure for a long duration of time. We experienced difficulties with keeping the pump continuously working over the duration of throwing the put and thus were unable to actually, accurately test the quality of the MaRS algorithm in teaching novices to throw pots.

If we were given more time we would have liked to moved towards a Passive Hand Recording (PHR) system. PHR is a method in which if the glove moves, the data is recorded, unlike AHR which constantly records hand movements, moving or not.

We however did not manage enough time to fully develop this method.

The code used throughout this entire project can be found here.

In order to determine whether or not our MaRS algorithm was more effective in teaching students to throw pots, we had to perform image analysis on the pictures of pots thrown by a teacher, a student after instruction with a teacher, and a student wearing the actuators. We decided that a good metric of analyzing the pots thrown by the individuals was to analyze how straight each of the pots were. To do that we split each pot in half to get a profile view of the edges of the pot. We determined where the base lay and drew a vertical line up from the midpoint of the base. We then, using ImageJ, measured the the distances, in pixels, of various lines to calculate the standard deviation of the lines in relation to the center line. Each side of the pot was split up into different component lines (as shown in the annotated image below) to determine the straightness of the pot. The hypothesis was that if the pot was completely straight then the distance between either edge of the pot from the center line should be the same, measured by standard deviation. Each component line had no more than 100 data points extrapolated from it. Smaller lines had fewer distances measured becuase there was less area to measure. All of our data can be found here.

As stated on our previous page about gathering and outputting data, we were unable to successfully output the pressures and make the actuator inflate consistently and in time to aid the student in throwing a put. Therefore, we, unfortunately, were unable to test the validity of this method of teaching students how to throw a pot. However, the actuators did hold up while they worked, proving that the King Actuator can be used in vigorous applications.