The Research Problem

Soft robotics, as actuators with a high degree of freedom and durable for brittle damage, have a wide range of application prospects. Previous studies have exploited this technology a lot. For clinic usage, for example, helping gout patients regain control of their end limbs as digital gloves or serving as an extra skeleton as air muscle. For more general applications, both bionic soft robots and rigid-free soft robots have been developed to create more room for development in robotics.

Actuators are indispensable parts of soft robots, and most soft robots need to control many actuators. However, since the control of soft structures is not linear, most existing controls use multiple air chambers in series to simplify the control. However, this method can only control the curvature of the overall structure, which will lose many degrees of freedom compared to independent control.

This project proposes a method to control multiple gas chambers independently, namely, a pneumatic decoder for soft robotics to solve this controlling difficulty. This decoder is free from rigid structures and minimizes the number of inputs to control many outputs with a small number of inputs, working as a pneumatic decoder, just as its name suggests.

The figure above shows a moving robot with actuators consisting of chambers in series. It has a general schematic (as shown in the figure below) of a soft robotic system, including pressure source, valves system, sensor/drivers, user control, feedback, and actuators. The robot to deliver in this project uses a similar scheme to show the decoder’s function.

Design Criteria


A simple soft robot is designed to test pneumatic decoder-based controlling. It consists of several actuators, a control circuit, a pneumatic decoder, and a pressure source. The robot can have several motions (as shown in the draft below), and it could be helpful in depopulated lands for search and rescue or mapping.


The bellowing schematic of the robot also shows each parts need to be complete in this project.

Main Break Down Tasks

  • Logic of Decoder

Here a decoder is designed to control six actuators with three inputs.

The table shows the working logic of the decoder, where the yellow blocks represent a valve with a higher pressure threshold to close and orange ones having a lower pressure threshold. The valve has two chambers. When the pressure reaches the threshold, the lower chamber will expand till the upper chamber is completely blocked so that the valve is closed. Valves in each row are connected through the lower chamber. Similarly, in each column, valves are connected through the higher chamber. By function, each row is named a control line, and each column is named a flow line.


  • Pneumatic circuit
The input table shows the pressure needed for each control line to open a particular flow line. For example, to open flow line 1, when we set all three control line at the low-pressure threshold, all the rest five flow line will be blocked since all of which has a low threshold valve blocked in the line, except for flow line 1, so that only flow line 1 is left open. A similar method applies to all six flow lines.
  • Pneumatic Elements

The decoder consists of 3 pressure sources for the control line, and each source needs to be able to input at least three levels of pressure, low pressure, low threshold pressure, and high threshold pressure. Then two kinds of valves, whose threshold can be realized by the elasticity of the flexible part. The valve design will be described in later sections. Finally, an air source feeding all the flowlines is needed.

  • Silicone vs PE film

An elastic part is needed not only for the actuator but also in a valve, it should be very extensible and sensitive to pressure change. A spring or an elastic thin film are of options. The silicone layer and a PE film are compared, and the former is proved to be more sensitive to pressure and very extensible, but the latter is prone to be broke. The elasticity of the material is shown in the video.
  • Structure and Function

The extend of actuator is also used to calculate the pressure change, so a strain gauge is to be embedded in each actuator for measurement. With the data of pressure change of each actuator, the most direct result is to know which actuator is touching the ground. Later, with the order of actuators was touched, we can use a non-direction graph to map the route our robot travels. The code of mapping is linked below.


  • Valve Control

To realize the essential function of the decoder, use PWM and Mosfet to control pressure proportionally.

  • Signal Analyzing

Collecting data for each actuator is necessary for feedback control or further applications. This project introduced how to map the route by measuring the expansion of each actuator.

  • Feedback

For future study, an oscillation circuit can be added to the system. It can store a periodic trigger of actuators, which will give the robot a specific movement like moving forward automatically or turning around when facing an obstacle.

  • Sizing Down

This project aims to embed a flexible printed circuit (FPC) as the control system. In this case, the rigid part could be considerably reduced. The FPC design is described in the following sections.

Fabrication: Structures

Valving with Hybrid Materials

Two relatively rigid parts (still soft) sandwiching a soft layer described above composed a hybrid valve. Similar to the valve in the figure above, but instead, the lower chamber will expend to block the upper one, and both two rigid parts can be of an identical shape. 

The simplest solution using the most accessible method is to bond the PE film with a FDM printed abs part. But as previously discussed, the PE film is not ideal in this case, the low resolution of FDM printing is also prone to get a blocked channel. The two materials are also hard to bond. So this is more of a sketch.

Figures of this version are shown in the gallery bellow.

An alternative design eliminated the bonding problem, where a silicone layer both between the two rigid layers and surrounding the whole structure is cured at the same time. The bonding is done by curing the 3 parts together. A holder (yellow model) is designed to cure the bottom and middle layer, and the opposite layer is cured after the other half is cured. The upper and lower rigid parts are of identical shape, and the chamber is in a smooth surface, which will be easier to fit the expanded curve shape of the soft layer.

The rigid parts are printed by LCD printer using resin substrate. Many other curing processes is tested in for this concept. But in every case, the silicone won’t cure.



In version 1.2, I made a thin layer of silicone to test the feasibility of a thin layer. It succeeds by curing between two cardboards. At the same time, I tried to cure the valve in a larger piece of silicone, planning to cut the valve off entirely. Again, the cure is thriving in cardboard made by laser cutting. However, the cure failed again around the valve.
A laser-cut box model can be generated easily on this site. Upload this file to Glowforge laser, and you can have a custom box.

Alternative Rigid Part

The rigid design is easier to make, consisting of a valve with four vents, a piston, and a spring. As shown in the picture above, a spring will connect the cap on the bottom and the piston, and rubber rings can be put in the circular crater around piston. When the upper chamber has a greater pressure, the piston will be pushed down till the lower chamber is blocked. 

Version 2.0 has the leakage problem in the area pointed by red arrows. 

The leakage is solved in version 2.1, but the fraction between each part needs much testing.

Final version

The half in the lower-left corner is in flow lines, having a lower threshold. The one with a deeper and smaller trough has a higher-pressure threshold. Both designs can also be used in control lines.

A maximum of 9 parts can be printed in an LCD printer with resin substrate. It has a much higher resolution compared with the FDM printer since it was decided by screen rather than the nozzle. It is recommended to do the CAD on OnShape, where files are stalled in the cloud. It will be helpful, especially when the CAD is reviewed by a group, so that everyone can work on it simultaneously. You can find a the instruction of the LCD printer here.

Then, Chitubox is the slicer for LCD printers. It is important that the supports should reach the far end of the part to prevent falling off during curation. Also, a smaller project area is better. It is recommended when parts are rotated to a vertical position.

The final version is gluing a silicone layer between two resin halves then the three parts are sealed with para films or cling film. From the video, we can see through the translucent chamber that the flexible layer expanded with an increase in pressure.

Fabrication: Circuit

A draft assembled ESP32, pressure source controlling, pressure measurement, primary air source, power regulation. This is the basis for selection and modification.

A Solution for Higher Voltage Input (no longer needed)
Jettisoned high-voltage appliances 0%

  • D4 avoids reverse connection
  • C12 and C13 is for energy storage and decoupling
  • R66 Slightly pull-up potential, triggering D4
  • R65 and R67 is feedback resistor
  • MP2359 is a DC-DC module with a buck circuit, where a DC voltage is transformed into a high frequency power supply through an oscillation circuit, and then the required DC voltage is output through the pulse transformer and rectifier filter circuit.
  • L2 and C11 forms a filter, for the ripple current from MP2359. The higher the switching frequency, the smaller the inductor can be.

  • C2 is a bootstrap capacitor that boosts potential by charging. It can work as clamp protection for ripple current form pin 6.
Pressure Source Controlling
PWM + mosfet 25%
  • D9 is used to protect components from induction voltage breakdown or burnout
  • Q9 is for widening current from I/O to drive motors

  • Pin 1 and 2 is connected to the motor (vacuum pump) 
Monitor Pressure by Measuring Extension
Strain Gauge 50%
  • The strain gauges arranged in Whetstone Bridge Amplifier Circuits is to be set into each actuator. Two repeated units are shown to the right. 
  • C3 prevents self-excitation of the chip can also be used as a low-pass filter.

  • Extension is a variable resistor that is a long conductive fiber inside each actuator. Its deformation affects the resistance change.

  • The voltage measured is amplified by U1A. The scale is set as 1, and can be adjusted by changing the resistance of R32, R33, and R34.

  • Then the data is collected by ADC CH.

Main Air Source for Flow Lines
  • The main pressure source for flow lines is control by TIP120, a Darlington transistor.
  • The dimension of circuit can always be scaled down by integrate modules into FPC (like the red part to the left), instead of using  commercial chips. 
Distribute the Power and Signal
Voltage, ESP32 90%

At this time, a schematic of FPC for decoder functioning (three control lines with one line standing by) , pressure measurement and data collection from six actuators (for mapping),  and wireless control via WIFI is completed.

Draw the FPC Circuit
A 3D Preview 100%

Finally, a small-size circuit for FPC  is ready to be printed on a flexible substrate like PTFE or Teflon.

Prototyping with Larger Elements


Sizing Down or Using Soft PCB


Prototyping Process & Trouble Shooting

A tutorial for PWM control based on HTML is recommended for beginners.

Before using html-based control, the hardware can be tested using serial communication. The code can be downloaded by clicking green words.

This page talks about setting up an webserver with ESP32. 

The PC should be in the same WIFI as ESP32 connects to.

Test the control method above with motors instead.

MPX5500DP is a pressure sensor with a range up to 500 kPa.


  • D4 to D7 on LCD is connected to D2 to D6 on Arduino
  • RS and E is connected to D11 and D10

Trouble Shooting:

Quantitively Measuring the PWM controlled Pressure Input

Combining the methods above, wireless PWM control through WIFI is quantitively tested with a manometer.

Result: Assemble the Entire System

  • Print out all valves needed
  • Make sure no blockage and leakage for each one
  • As previously discussed, the bottom chambers in control line are of an identical shape, the difference in the upper chamber distinguished the thresholds.

A pneumatic circuit is connected according to the logic table in the design criteria section.

Flow lines are highlighted in white, and control lines are marked in white in the photo to the right. 

Connect vacuum pumps to each line, rule out leakage issue in each line.

The video showed that every valve is responding to the pressure change.

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.