Image Image Image Image Image Image Image Image Image Image

Bartlett School of Architecture, UCL

Scroll to top


No Comments

A Portable Air Driven System for Kinetic Structures

A Portable Air Driven System for Kinetic Structures

One of the essential considerations of kinetic architecture is how to propel the structure. Commonly, kinetic structures are actuated by electric drivers such as servo motors. Application of pneumatics in this area remains far less developed. However, as a soft driver, the air-operated system has unmatched advantages in certain ways. It can, for example, reduce the overall weight of a structure and make the structure have more inherent compliance.

My research presents a design of a new propulsion system for kinetic structures that is based on pneumatic components. It includes a certain kind of soft actuator, the air muscle, and a portable pneumatic kit with an autonomous controller. The research includes the exploration of soft robotic design and explores in which way soft actuators can be applied in stiff mechanical movement. The primary goal of the article is the development of an open source pneumatic toolkit that is portable, affordable and sustainable. The kit can be widely used in kinetic structural research and applications.


1. Introduction

Kinetic structures require actuators to control the mechanisms and generate motions. There are three different types: electrical, hydraulics, and pneumatics. All three have been widely explored by mechanical engineering (Gabriel, 2011). Electrics and hydraulics have been applied in the field of kinetic architecture. For example, the deformable pedestrian bridge at Paddington Basin, designed by Heatherwick Studio (Heatherwick, 2004), and the Morphs developed in the interactive architecture studio (Bondin, 2013). Pneumatics can be a valuable asset for kinetic structures, because they can be simple and powerful in their design. However, research in this area has been neglected.

屏幕快照 2016-08-11 14.24.11

Figure 1. “the Muscle Projects” by Hyperbody, 2004. References were studied to understand the reasons why pneumatics are not wild applied in kinetic structure research.

Apart from the Hyperbody projects, it is hard to find research and applications using pneumatics in kinetic structures. There are a number of potential reasons for this, such as controllability and safety issues. The required technical expertise and the excessive cost are probably the main stumbling blocks restricting development in this field.

In an attempt to overcome these obstacles, the project covered in this report aims to develop a simple home-made pneumatics system, which requires less expertise in the area and a fraction of the cost. In the course of the research in this report, some limitations of pneumatics, like the weight of air compressors, was discovered.

This article focuses on the technical aspects of the project, specifically the design and development of a portable pneumatic kit that can be used in kinetic structural research and applications. Furthermore, the system should potentially overcome some current limitations of pneumatic systems, such as the high cost of the industrial actuators and the inconvenience of air compressors.

1.1 Research Project Overview

1.1.1 Project Golem

屏幕快照 2016-08-11 14.26.47

Figure 2(a) and (b). Project Golem prototypes: Soft walking pyramid(left), Three legged robot(right).


The project covered in this article consists of two parts: Project Golem and Golem-kit. Project Golem explores the possibilities of pneumatically driven structures, to create a kinetic creature with the ability to walk. Inspired by the Morphs from the interactive architecture lab (Bondin, 2013) and Theo Jansen’s walking creatures (Jansen, 2008), project Golem’s structure consists of a rigid framework of rods, which uses soft actuators (pneumatic artificial muscles) to perform movement. The initial idea of Golem was to build a mobile structure that is able to adapt to environmental changes. The structure’s actuators require a portable pneumatic system to work, which was the primary technical challenge faced in project Golem.

1.1.2 Golem-kit

Golem-kit is a toolkit that can be used to create pneumatically powered kinetic structures (Golem-kit, 2016). Due to its modular design it provides a wide range of possibilities. It includes modular joints, soft actuators, and a pneumatic control board. The toolkit was used in the creation of the robot in project Golem. It does however use a different air compressor, which is described in detail in chapter 3.


Figure 3. Golem-kit: A pneumatic multipurpose modular toolkit.


2. Pneumatic System

2.1 Basic Pneumatics

Pneumatics is a branch of engineering that transforms the energy from pressurized air (which usually comes for the atmosphere) to kinetic energy to perform various actions (Parr, 2011). It has long been valued to generate kinetic energy for mechanical work. Due to certain properties of air (e.g. its rapid expansion rates), pneumatic actuators have various merits, such as their light weight and fast reaction times (Parr, 2011). Compared to electric or hydraulic systems, pneumatic ones are more sustainable, since air is an unlimited resource. Furthermore, they are much safer, because the risk of fires is a lot less (Parr, 2011).

Advantages of pneumatics

There are three main reasons why the projects covered in this report use pneumatically actuated systems for locomotion:

  • Pneumatic components do not need to withstand pressures as high as hydraulics do. Therefore, they can be made of high- strength plastics, wood or aluminium, rather than steel or cast iron (NFPA, 2015). This leads to significantly lighter components.
  • Electric motors generally require expensive and complicated electric drive controls (NFPA, 2015). Pneumatic actuators (especially the pneumatic artificial muscles used in the project) are much simpler and affordable, which significantly lower costs for the system.
  • Due to the low pressures required to operate the pneumatic artificial muscles, it provides a natural compliance, which makes the system human-friendly and safe to use (Veneva, 2012).

Limitations and improvements

The application of pneumatics in mobile structures can be limited due to a few factors. One of the biggest issues is the requirement of an air compressor, which are generally very heavy. Additionally, an air reservoir is required, which can be very large. These two components, though vital to the system, can hinder the mobility and affect the aesthetics of the structure. Besides, a reliable pneumatic system might also require specialised components (e.g. air pressure detector and flow rates controller), which could increase the overall cost. Thus, The project had to overcome the following challenges:

  • Design a portable air system for movable structures and optimise the air production to be used with the toolkit.
  • Use low-tech and cheap materials for each component.

Technical terms

The pneumatic system consists of two parts: “the air production and distribution system and the air consuming system” (SMC, 1997). More specifically, air is drawn in from the surrounding atmosphere into a reservoir by a compressor until it reaches the desired pressure. Dust filters and dryers then clean and dry the air, which is important, so the pneumatic components don’t get damaged. The pressurised air can then be distributed via valves, which can control air direction and speed, to the various terminal actuators in the system (Parr, 2011).

Figure 4. Basic pneumatic system by SMC,1997. A pneumatic system consists of an air production and an air consumption part.

Figure 4. Basic pneumatic system by SMC,1997.
A pneumatic system consists of an air production and an air consumption part.

In pneumatics, pressure can be expressed as ABS (absolute pressure). According to the law of pressure, the standard pressure unit is Pa (Pascal).

1 Pa = 1N / m2. Since this unit is minuscule, in order to avoid large numbers, engineers usually use bar instead of Pa (SMC, 1997).

1bar = 100,000 Pa. The standard unit for pressure in the United Kingdom is psi, pounds per square inch (SMC,1997).

14.6 psi = 1 bar. For the sake of simplicity, this report uses psi as the basic unit of pressure.

Most industrial pneumatic systems use pressures of around 100 psi or less (NFPA, 2015). This value depends on the actuators’ capabilities and performance needs. The systems designed for the Golem projects all use pressures of 20 to a maximum of 50psi. These pressures provide enough power to drive the structures, while keeping the system within safe operational modes.

The following sections describes the individual components of a pneumatic system. The first part deals with actuators, specifically pneumatic artificial muscles (PAM), the type that was used for project Golem. The next section describes the optimisation for the air generator, which is of paramount importance of any pneumatic system. As stated earlier, it is also one of the biggest challenges when designing mobile structures. Next, the control system is introduced. The last part in this chapter is a brief discussion about the energy strategy.

2.2 Actuator (the PMA)

Comparison of air actuators

The two most commonly used pneumatic actuators are air cylinders and artificial muscles. The specifications provided in Figure 5 below show the maximum pull/ push force of the different types. When cylinders and air muscles have the same diameter (Ø 20 mm) their respective pull force is almost identical (300N). PAMS, however, can weigh as little as 10 grams, which is less than one-tenth of a standard pneumatic cylinder.

屏幕快照 2016-08-11 14.32.42

Figure 5. Pneumatic actuators comparison. Air muscle by the Shadow Robot Company (left), and Pneumatic cylinder by the Festo (right).

Pneumatic cylinders have one major advantage over PAMs, their tension and compression force are almost equal. PAMs can, however, be modified to provide compression force as well (e.g. by adding springs). Besides, PAMs have the advantage, that they can still work if they are twisted or bent, and are therefore more flexible.

Pneumatic artificial muscle (PAM)

The PAM (known as the air muscle) is an extraordinary actuator that simulates natural human muscles (Minh et al., 2012). A standard PAM is made up of a rubber inner tube (also referred to as the bladder) with a tight plastic netting cover. It works on a simple principle: the inner bladder expands and pushes against the mesh cover when it is pressurised. Due to the deforming restriction of the outside scissor- hinged cover, it can transform the circumferential stress into an axial contraction force (Tondu, 2012). Hence, the muscle contracts when it is actuated and generate tension force.

Figure 6. McKinbben air muscle by Veneva, 2013 How the PAM works when it is inflated.

Figure 6. McKinbben air muscle by Veneva, 2013 How the PAM works when it is inflated.

One of the problems with PAMs is the limited availability for people who do not work in the field. Companies that use them usually have custom ones for specific purposes designed and built, which also makes them relatively expensive. For project Golem a home-made PAM was developed, which is cheap, easily produced and can exert enough force. Its design was kept close to the McKibben muscle (Veneva et al., 2013). A rubber balloon was used as the bladder and an expandable cable sleeve was used for the cover. These resources can easily be purchased to produce a large number of PAMs at a significantly lower price than industrial grade ones. A tutorial on how to produce them can be found on our open source toolkit website,

Contraction ratio experiments

During the development of the PAMs different diameters and materials for the cover were tested. Figure 8 shows the results of these tests. Type A’s cover diameter is 8 mm, and made of medium density braid netting, which has less area of contact with the bladder compare to high density ones, therefore reducing the amount of friction. Type B’s diameter is only 4 mm and made of medium density scissor- hinged netting. In order to calculate the contraction ratio each PAM’s length was measured while being completely stretched, as well as fully inflated.

Figure 8. Contraction ratio test. PAMs with different sizes and materials show different properties when inflated.

Figure 8. Contraction ratio test.
PAMs with different sizes and materials show different properties when inflated.

According to the results in Figure 8, the contraction rate of type A is 30%, much higher than type B’s 17%. Type A therefore shows more potential for project Golem, so it was chosen as the standard PAM for the project. The required pressure for these to operate correctly are in the area of 20 to 30psi. A detailed experiment of this can be found in chapter 3.

2.3 Air Source

The air source consists of two part (SMC, 1997):

  • An air compressor that draws air in from the surrounding atmosphere and compresses it (these are usually electric).
  • An air reservoir that stores the compressed air.

Container that can withstand the maximum pressure of the system can be used as a reservoir. However, the compressor poses a bigger challenge. Because the whole system cannot function without it, it is seen as the most important part of the system. Due to its weight and the requirement of being kept level, it is usually fixed to the ground. Relatively lightweight compressors that are powerful enough to actuate the PAMs are available, but they still require a large battery to run, which is an inconvenience in the context of project Golem.

Comparison and analysis of air compressors

This section deals with possible solutions for this issue. The first approach is removing the air compressor from the design. The connected air reservoir would act like a battery, storing potential energy. This concept would require the reservoirs to be “recharged” regularly using an external compressor. The advantage of this is the obvious reduction of the structures weight. The stored compressed air could still be used to actuate the PAMs. The disadvantage is the evident lack of self- sustainability of the system. In order to increase the maximum running time a larger reservoir would be required. Alternatively, CO2 tanks could be used, but the higher pressure of these causes safety concerns. Because these issues would basically render the structure useless during recharges this method is infeasible for the requirements of project Golem.

The second approach, described in detail in the following section, includes a thorough analysis of the requirements for the compressor, comparing and analysing current air compressors and finding the optimal choice.

There are two types of compressors that are commonly used today: centrifugal and positive displacement compressors (also known as the volumetric pumps) (Matthews, 2014). A centrifugal compressor uses a rotating impeller (a type of rotor) to draw gas in through its inlet (located at the centre of the rotor). The special shape of the impeller and its rotating movement accelerates the gas outwards, thereby creating a vacuum at its centre, which in turn draws in more air. A defuser is used to transform the increased kinetic energy of the gas into potential energy, thereby compressing the gas (Bengtson, 2011).

A positive displacement compressor draws air into a chamber, then decreases the volume of this chamber thereby compressing the air. These types of compressors can be divided into four categories (Hooper, 2013): rotary machine type air compressors (vane and screw) and reciprocating compressors (diaphragm and piston).

Figure 9.Classification of air compressors by Hooper,2013

Figure 9.Classification of air compressors by Hooper,2013

The compressor used in project Golem requires a suitable range of outlet pressure (20 to 50psi) but does not require high flow rates. The rotary machine type air compressors are excluded because they are generally used in industrial settings and provide excessive pressures (150psi) and high flow rates (SMC,1997).

Figure 10. Diaphragm compressor diagram by Bengtson, 2011. The diaphragm draws air in when expanding and forces air out when compressed.

Figure 10. Diaphragm compressor diagram by Bengtson, 2011. The diaphragm draws air in when expanding and forces air out when compressed.

Reciprocating compressors (both diaphragm and piston) on the other hand are used in many different areas. They both work on the same principle with slight differences. The piston compressor uses a piston to increase and decrease the chamber size, while the diaphragm compressor uses a flexible membrane. The reciprocating motion of both piston and diaphragm are usually driven by a rod and crankshaft (Bengtson, 2011). The diaphragm compressor is totally oil free, which is why they are usually used in food and pharmaceutical industries (SMC,1997). Both compressors draw air in when the diaphragm/piston is pulled back and forces air out when they are pushed forward. Two one-way valves are used for the intake as well as the output (Bengtson, 2011).

 Figure 11(a) and (b). Mini diaphragm compressors without air tank, examples by RS (left), and Floret (right).

Figure 11(a) and (b). Mini diaphragm compressors without air tank, examples by RS (left), and Floret (right).

Diaphragm compressors provide pressures in the range of 10 to 60 psi while piston compressors manage 50 to 100 psi or higher (SMC,1997). Therefore the pressure of diaphragm compressors is more appropriate for use in project Golem. Furthermore, the fractional electric motor and smaller reservoir are used in diaphragm compressors, which make them possible be portable. Thus, a small diaphragm compressor, such as a spray painting pump, might be an acceptable choice for pneumatically driven kinetic structures (with power supply).

However, the optimised result of the second way only suits the case that the kinetic structure is not designed to be mobile (for the reason it demands high power that leads to a big power supply), and still has some drawbacks like noise. Hence, the proposal for a third approach, which is the focus of this thesis, that is an autonomous pump prototype designed for project Golem to solve these problems. The detailed design can be found in the third chapter.

2.4 Control System

The control system is part of the air consumption system. It is designed as an independent device that controls compressed air and operates the movements of the PAMs. The control board consists of a set of solenoid valves on the manifold that directs the compressed air and an Arduino board with digital pressure sensors that controls the valves.

The Golem air control board is inspired by the “Fluidic Control Board” from the “soft robotics toolkit” developed by the Harvard University. The “Fluidic Control Board” is an open source hardware platform that is designed for students to test the behaviours of different pneumatic actuators (soft robotics toolkit, 2016). The golem control board simplified the “Fluidic Control Board” by optimising its circuits and removing unnecessary parts to cut down the cost (by more than half). Apart from the control board, an air flow rate controller is embedded in the system which is able to adjust the inflation speed of the PAMs. The resulting system is not only designed for the operation of PAMs but can also be used for other pneumatic applications.

Figure 12. “Fluidic Control Board” by soft robotics toolkit. Reference was studied to evaluate different control ways.

Figure 12. “Fluidic Control Board” by soft robotics toolkit. Reference was studied to evaluate different control ways.

Two different control boards with different sizes, energy supplies, and microcontrollers were designed for project Golem. Control board A, depicted in Figure 13(a), is built on an acrylic board. Four 2-state, 5-way solenoid valves powered by small zinc carbon batteriy are included. An Arduino Uno is used to switch the solenoid valves on and off. This system can control 8 PAMs at the same time and 4 PAMs individually. Control board B is an upgraded version of board A and works on the same principle while being even smaller due to the use of a micro Arduino. The size of control board B is smaller than the size of a conventional smartphone. Additionally, it has a circuit that is designed to add the use of solar energy. Its small size and lightweight design allows this board to be placed on the structure with minimal additional supports.

Figure 13(a) and (b). Pneumatic control board A and B Board A (Left): 150mm by 150mm, Arduino Uno, battery. Board B (right): 100mm by 90mm, Arduino Micro, solar power.

Figure 13(a) and (b). Pneumatic control board A and B Board A (Left): 150mm by 150mm, Arduino Uno, battery. Board B (right): 100mm by 90mm, Arduino Micro, solar power.

As mentioned above, an Arduino board is used which gets the sensor readings as its input and does the necessary calculations to deliver the control systems output. In the computer port, VVVV (a hybrid visual live- programming environment for easy prototyping and development (vvvv, 2016)) is used to test different behaviours because the VVVV can do real-time control. With the help of the Arduino microcontroller, it is possible to add more interfaces, such as additional sensors, to the hardware to improve the systems performance. Project Golem also embeds a GA (genetic algorithm) in its design, which allows the creations to learn how to react in various situations according to its environment.

2.5 Energy Strategy

Project Golem aims to use solar power as an energy source for the air compressor and the control system. Although the energy supply is not strictly part of the pneumatic system, it will still be covered in this report, because of its obvious necessity to power the kinetic structure.

Solar power has a few advantages over more traditional power supplies, e.g. batteries. Apart from being more environmentally friendly, it also allows the system to be self-sustaining, removing the need to recharge or replace the battery regularly. Because Golem is designed as an autonomous system, the necessity for this is apparent. Additionally, the lighter construction of solar panels compared to lead acid batteries reduces the overall weight of the construction. The pneumatic system used in Golem operates on a low voltage of 12V, however the operation current is not small, which means the system requires a powerful energy supply. Conventional 12V lead acid batteries weigh more than 2kg. Equivalent plastic solar panels weigh 3/4 less, which is obviously beneficial for a mobile structure. It is therefore evident, that solar panels are an ideal energy strategy for project Golem. However, batteries are still required during the development phase to ensure efficient testing of the designs.

The next step is calculating the energy consumption of Golem’s pneumatic system. This determines the number of solar panels required. The control board consumes around 4W (0.6W for each solenoid valve and 1.5W for the Arduino), which only contributes a small part to the overall energy consumption. The air compressor requires the largest amount of energy.

The energy consumption per psi of compressed air is unascertainable since air pressure cannot be directly converted into electric power. There are two indicators used to measure the capacity of an air compressor: the air flow rates (CFM) and the air pressure (Chawla, 2016). Both of them need to be considered during the energy consumption calculations. Usually, these two indicators are inverse, which means when one increases the other will decrease. Theoretically the power needed to compress air can be expressed as

“HP = [144NP1Vk / 33000 (k-1)] [(P2 / P1)(k- 1)/N k-1] ”

Figure 15 Horsepower required to compressed air by engineeringtoolbox.

Figure 15 Horsepower required to compressed air by engineeringtoolbox.

According to the “horsepower required to compress air” diagram (Engineeringtoolbox, 2016), 50 psi gauge pressure require around 60W at a flow rate of 1CFM. Project Golems motion frequency is usually very low, which means the air flow rate can be kept low. In practice, the flow rate is lower than 0.3CFM. This means ideally (when no energy loss is considered) the power consumption is around 17W. Therefore, the solar panel specifications for project Golem are 12V and 20W. A more in depth description of the design will be presented in the next section.

3. Autonomous Pump for Golem

3.1 Research design outline

In section 2.2 different air compressors were compared and discussed. Because of its small size and low weight the diaphragm compressor is a reasonable choice for Golem-kit. For example, the TC-20 mini airbrush compressor from Blowtac provides compressed air from 40psi to 60psi at 0.7CFM, while weighing only 4kg (Blowtac, 2016). Additionally, a diaphragm pump can be easily accessed, which is beneficial for Golem-kit.

However, the problems of air compressors discussed in section 2.2 require further investigation. First of all, the weight of the compressor remains an issue. The diaphragm pump mentioned in the last paragraph is quite light compared to other air compressors, but still too heavy for a mobile structure that consists of dowels and lightweight materials, which is one of the advantages of using the PAMs in project Golem. Secondly, the diaphragm pump is an industrial product with a relatively high flow rate, which is designed to compress air at a fast pace. Higher flow rates result in higher energy consumption. For instance, the TC-20 requires around 100W of power, which is difficult to achieve with the selected power supply system. Lastly, the noise issue cannot be ignored. Air compressors generally generate a lot of noise. Silent compressors are available (such as screw type compressors), but they are too expensive and still cannot avoid the weight issue(SMC,1997).

To solve these problems, this section explores possible autonomous pump designs for use in lightweight kinetic structures, especially project Golem. The following sections outline the development of a suitable compressor. Firstly, the requirements for a compressor used in project Golem are analysed. Next, a compressor design, based on a single stage piston compressor, constructed of affordable components is presented. The following section deals with the optimisation of this compressor prototype. The last section of this chapter is an evaluation of the proposed autonomous compressor design.

3.2 Experiments of Golem muscle

The contraction ratio test of the PAM presented in section 2.2 revealed a roughly 30% contraction when it is fully inflated. However, the PAM’s effective power depends on the air pressure. It is therefore important to define the relation between air pressure and tension force. Figure 16 and table 17(a) show the results of experiments with the Ø8 mm PAM using different pressure levels and measuring the resulting force. The results in table 18(b) indicate a linear growth in the relation between air pressure and force.

Figure 16. Tension force experiments.

Figure 16. Tension force experiments.

Fable 17. Experiments result and Linear growth diagram

Fable 17. Experiments result and Linear growth diagram

The purpose of this experiment was to evaluate the pressure that the PAM requires to drive the structure. Air pressure is an essential parameter in the compressor design. The kinetic structure of project Golem requires around 20N tension force. According to the table 16(c), the PAM needs at least 14psi pressure to be actuated and the pressure does not need to exceed 50 psi. Therefore, the requirements for the autonomous compressor are settled to 20 to 50psi.

3.3 Autonomous pump design

The air compressor designed for project Golem is based on a one stage piston compressor. As described in section 2.2 a piston compressor uses a pistons downward movement to get air into the cylinder and its upward movement to compress it (SMC,1997). Hence, the Golem compressor uses a small hand pump as its piston, because of its low price.

The reciprocating motion for the piston is created by an electric motor, which creates rotational movement of a crank. One end of the piston is connected to this rotating crank, which in turn forces the piston to move up and down in the chamber, producing compressed air. This is similar to a cars internal combustion engine. It converts the combustion energy into reciprocating motion, which in turn creates rotational motion using a rod and crankshaft. The Golem compressor functions on a similar principle, but in reverse. Figures 18 shows the principles of the piston compressor.

Figure 18. The piston compressor working principle by SMC,1997.

Figure 18. The piston compressor working principle by SMC,1997.

The main difference between the Golem compressor and a conventional piston reciprocating compressor is the resulting air flow rate. As mentioned before, lower flow rates reduce the unit’s energy consumption and require less powerful motors. Unlike a piston compressor, the Golem compressor is designed to run continuously at a very low speed. When the sunlight is strong enough, the compressor starts running until it reaches the maximum pressure. Because of its low speed, it takes a long time for the Golem compressor to fill up the air reservoir.

3.3.1 Autonomous pump prototype I

Figure 19. Golem compressor prototype I.

Figure 19. Golem compressor prototype I.


Figure 19 shows the first prototype for the Golem compressor. This design kept its main focus on keeping the weight low, in order to make it as portable as possible, and on reducing the rotary speed of the motor the reduce the air flow rate of the pump. As diagram 20(a) and 20(b) show, the Golem compressor consists of three parts:

  • a piston action cylinder that is driven by a 12V electric motor, which uses gears to reduce the rotary motion.
  • a small air tank embedded with a pressure sensor for pneumatic energy storage.
  • the solar power system with power converters and a voltage regulator.
Figure 20(a) and (b). Golem compressor diagram (left), and Air production process (right).

Figure 20(a) and (b). Golem compressor diagram (left), and Air production process (right).

The most essential part of this design is the pressure production component. Diagram 21 and 22 show how the system converts the rotary movement into the reciprocating linear action of the cylinder pump, this mechanism is called a slider-crank. It is easily manufactured and can withstand large loads (Qu and Zhang, 2011). Based on this mechanism, the manual reciprocating cylinder can be transformed into an autonomous pump powered by the electric motor. The Golem compressor prototype 1 uses a small bicycle hand pump (8mm action length and 1.5mm diameter) as its piston. In accordance with the design guidelines, it is cheap, lightweight, and reliable.

Figure 21. The slider-crank mechanism in prototype

Figure 21. The slider-crank mechanism in prototype


Figure 22 . The pumping precess of prototype I

Figure 22 . The pumping precess of prototype I

The hand pump completes one cycle of air compression for each turn of the crank. As the crank’s rotational speed directly affects the reciprocating movement rates, which determine the air flow rates, the gear speed is an essential part in this design. It uses a DC geared motor (maximum no load speed 84rpm under 12V) with additional gears box(gear ratio 3:1) to regulate the speed. Theoretically this should result in the hand pump completing around 28 cycles per minute under 12V, less when the power supply is lower than that (e.g. when there isn’t sufficient sunlight).

The prototype was tested using different voltages and power supplies (i.e. battery and solar panel). Using batteries with 5V and 12V the maximum pressure of 20 psi was reached within 5 and 12 minutes respectively. Using a 20W solar panel, and leaving it in the sunlight for about half an hour, the hand pump managed to work twice, for several minutes each.

Figure 23. Prototype I test with different voltages 5V

Figure 23. Prototype I test with different voltages 5V

These experiments proved the feasibility of the Golem compressor. It is worth mentioning that the support structure for the compressor is made out of plywood to reduce the weight. The resulting system weighs less than 1kg, much less than conventional air compressor, therefore making it a lot more portable. All the parts used in this design are cheap and can easily be purchased. The noise is also kept low.

The first prototype meets the initial requirements. However, there are several issues that remain. Due to the use of only one hand pump the resulting stress on the crank is not symmetrical. During the compression phase of the system, the entire load is concentrated on half of the circle. In other words, the reciprocating motion is only half as efficient as possible. The second concern is the air flow rate. Although the rotary speed is reduced, it is still too fast and energy-intensive, especially for the solar panel which generates power very slowly. Lastly the 3d printed components are not strong enough to bear the torque from the crank, which leads to the instability issues.

3.3.2 Autonomous pump prototype II

Figure 24. Golem compressor prototype II

Figure 24. Golem compressor prototype II


These issues were addressed in the development of the second prototype for the Golem compressor. One of the most obvious changes, was the addition of 3 more pistons. Half of them run on a half cycle delay of the other ones. When one set of the pistons is compressing air, the others take air into their chambers. This results in the stress on the crank and the gears to be evenly distributed. Additionally, the size of the hand pumps (4mm action length and 1mm diameter) is much smaller than the one used in prototype 1, which results in smoother pumping motions. The gear ratio was also changed to slow down the rotation speed and increase the torque. Prototype 1 managed to reach the required pressure for the PAMs (20psi), but the overall system requires pressures between 20 and 50 psi, as stated in section 3.2. Increasing the systems torque will result in higher pressure. Lastly, prototype 2 includes a feedback system that allows the compressor to run autonomously, and an air output flow regulator on the end of the tank to control the expansion velocity of the PAM.

The parameters and structure of the design are optimised. Usually, for a reciprocating piston pump, the ratio of the crank and the connecting rod is less than 1/4 (Qu and Zhang, 2011). Based on this standard, a Grasshopper simulation was built to calculate the best result for the length of the crank and the rods.

Figure 25. Grasshopper simulation

Figure 25. Grasshopper simulation

The second prototype uses the same DC motor as the one used before. However, the gear ratio was changed by a factor of roughly 10. This results in an output speed of around 2.5rpm (compared to 28rpm for prototype 1). Therefore, the cranks torque was also amplified. As the “horsepower required to compress air” diagram mentioned in section 2.5, there is negative correlation between speed and torque of the motor used in the Golem compressor. Unlike the first one, the prototype 2 has gears on both sides of the structure to get equilibrium force. The material used for the 3d printed joints has been changed from ABS to nylon, which is much stronger and more durable.

Figure 26. Servo air flow controller

Figure 26. Servo air flow controller

As stated above, the new system includes a feedback system that measures the pressure in the air tank. The microcontroller turns the DC motor on when the pressure drops below the lower pressure threshold (20psi), and turns it off when the upper threshold (50psi) is reached. Because of the low speed of the compressor, it usually takes a few hours for the pressure to reach the predefined value. The new system also added an air speed controller to regulate the flow rates. The controller is composed of a manual pneumatic valve and a small servo that receives a signal form the feedback system, and is able to control the inflating of PAMs at different speeds.

Figure 27 shows how the final compressor works. A practical test of the second prototype showed improved performance compared to the first one. It managed to compress air up to 30 psi. Also, compared with the prototype 1, it has a smoother pumping motion, and the design’s construction is more stable

Figure 27. Photographs of pumping process

Figure 27. Photographs of pumping process

3.3.3 Air reservoir

Another important component of the compressor is the air reservoir. The choice of an air tank depends on its weight, volume and safety. Industrial air tanks are usually capable of withstanding pressures of up to 1000psi. This results in very heavy containers. However, the pressure required in the project Golem is lower than 50psi. Therefore, the air reservoir can be a thin- walled bottle which is much lighter, and able to keep its shape under 50psi.

To choose the right tank, firstly, the required volume is estimated. The average volume of the PAMs in Golem is around 24ml when it is fully inflated (10mm diameter and 300mm length). PAMs work at pressure of 20 to 50psi. According to the Pascal’s law (F = p A), it needs about 9ml of compressed air at 50psi to be actuated. The volume of a Coke can is 330ml. Theoretically, if the air in a Coke can is compressed to 90psi, it can provide compressed air for a PAM for about 40 inflations assuming the pressure is kept at 30psi. Because of the energy consumption during the air transmitting, the practical experiment showed that a bottle of 650ml compressed air at a pressure of 50psi can actuate the PAM approximately 25 times. Therefore, the second prototype uses a 650ml aluminium tank with an automatic safety release valve, which is pre-set to 90psi (weights around 240g).

3.4 Evaluation

The second prototype shows definite improvements over the first one, however it still does not achieve the goal of 50psi. Due to the energy loss, the conversion efficiency of the home-made compressor is relatively low. Various potential improvements, such as fitting bearings to the shaft, or applying oil between the gears, have been tried, but the problems remain. The second shortcoming is the energy supply. Solar panel would be an ideal energy source for project Golem, however, in practical terms, a battery is still required to ensure a stable energy supply, which adds additional weight to the portable structure. Therefore, further research and development is required, to improving the conversion efficiency and optimising the energy strategy.

4. Applications

Both projects covered in this report, Golem- kit and project Golem, require a pneumatic system, however, the requirements for this system are different for each project. Golem-kit, which is an air-driven invention toolkit with modular joints, requires a control system, an air compressor (non-portable), an air reservoir, and pneumatic actuators (PAMs). The main function of this kit is for prototyping pneumatic structures which includes building and testing them. The pneumatic system presented in this report meets these requirements.

Figure 29. Golem-kit inventions

Figure 29. Golem-kit inventions

The focus of the thesis is a portable pneumatic system attached on the movable structure, which contributes to the project Golem, a self-sufficient walking creature.

The project starts from a double-layered tetrahedral structure that converts the single linear movement of a PAM into multiple mechanical movements. The basic unit has 3 PAMs that can be actuated separately and control the upper tetrahedral points to 8 positions. As the whole unit is built by Golem-kit, the structure is extremely lightweight. It therefore requires very low pressure to generate movements, such as swinging and sliding.

Figure 30. Basic module of Golem Three PAMs control the form of the upper tetrahedron

Figure 30. Basic module of Golem
Three PAMs control the form of the upper tetrahedron

In the first stage of the project, the pneumatic system was a separate part from the structure, connected only via pipes that transported pressurised air from the compressor to the structure’s PAMs. This resulted in limited range of movement for the robot. To create a self-contained autonomous walking system, a rolling cube that consists of four modular units was created. These four modules are connected to form a cube, with the tetrahedron in the centre, which leaves enough space for the pneumatic system to be installed. When the PMAs are actuated, the cube changes its form to shift its centre of gravity, which results in a rolling motion.

The pneumatic system inside includes the Golem compressor, air tanks, and the control board. All sides of the tetrahedron are covered with solar panels to ensure that at least two of them are always exposed to sunlight. Air is compressed by the pump inside and distributed to the PAMs via the control board. As long as the structure is exposed to sunlight, it will constantly compress air and store it in the tank. When the pre-set upper threshold of the pneumatic system is reached the compressor will stop. Every times the cube rolls, it consumes air, thereby lowering the pressure in the system, which will activate the compressor. Currently the cube is designed to takes a few hours for every movement.

 Figure 31. Walking cube prototype

Figure 31. Walking cube prototype

The pneumatic system functions as the kernel of the structure, distributing air and energy to the different limbs. This allows the cube to achieve the ability to walk and be self-sufficient. It has the merit that it can adapt depending on its surroundings. For example, the structure can walk on sand and potentially work in the desert. Also, because of its lightweight design, the cube is easy to carry, compared to conventional large scale structures. Lastly, as the cube moves slowly and due to the soft properties of the PAMs it doesn’t pose danger to humans. The drawbacks of the design are the dependency on light and the requirement for regular maintenance. Additionally, it does not have any sensors to communicate with its surroundings. All these issues need to be taken into consideration during the further development of the next prototype.

5. Conclusion

The use of pneumatics in the field of kinetic structures is not particularly well developed, compared to other locomotion systems. The research covered in this report proves, that there is great potential in pneumatics. One of the obvious merits of air driven actuators, especially the PAMs, is their lightweight construction. These combined with the solenoid valves weigh a lot less than regular linear actuators, and it does not need to support any loads on structural joints. Also, as a soft driver, the pneumatics system provides compliance and is human- friendly.

This report and the research it covers addresses the potential applications of pneumatics and produced a portable pneumatic system which is designed for portable structures. Two prototypes were developed in the process and the experiments done prove the feasibility of the autonomous system’s design. However, these also revealed problems of this system, e.g. the relatively inefficient energy conversion. These issues need to be addressed further and solved in order to produce a useful system.

The walking cube developed in project Golem is an example for the potential application of the Golem pneumatic system. The system provides a wide variety of possibilities for students and designers to perform research in the field of kinetic structures.the autonomous pump can also be used in separate architectural applications. For example, it is suitable to drive a kinetic facade that is exposed to sunlight, and doesn’t require fast movement. Additionally, Golem-kit has great potential in aiding in the design process of small structures and robots.

The pneumatic technology today, focuses more on industrial products. The Golem pneumatic system, on the other hands, can benefit research of pneumatics in the field of kinetic structures. It provides a way to overcome limitations in this area. However, the system is not perfect and still has problems, e.g. the energy supply. Hence, more experiments and design prototypes are required for further improvements.


Bengtson, H. (2011). Differences Between the Centrifugal Pump and Positive Displacement Pump. [online] Brighthub Engineering. Available at: http:// positive-displacement-and-centrifugal- pump-and- how-they-work/#imgn_7 [Accessed 9 Apr. 2016].

Blowtac, R. (2016). Single Cylinder Mini Air Compressor (TC-20) – High quality Single Cylinder Mini Air Compressor manufacturer from Taiwan | Sun Mines Electrics Co., Ltd.. [online] Available at: http:// Mini-Air-Compressor/TC-20.html [Accessed 15 Jul. 2016].

Bondin, W. (2013). Morphs 1.0. [online] Interactive Architecture Lab. Available at: http:// morphs-1-0 [Accessed 20 Jul. 2016].

Chawla, R. (2016). How to Size an Air Compressor?. [online] Available at: http:// compressor.html [Accessed 08 Jul. 2016]. (2016). Horsepower required to Compress Air. [online] Available at: http:// compressed-air-d_1363.html [Accessed 14 Jul. 2016].

Gabriel, M. (2011). DYNAMIC SYSTEMS responsive, adaptive, kinetic. (2016). Golem Kit | invent with air- power. [online] Available at: [Accessed 22 Jul. 2016].

Heatherwick Studio, (2004). Rolling Bridge | Heatherwick Studio. [online] Available at: bridge/ [Accessed 30 Mar. 2016].

Hooper, J. (2013). Basic pneumatics. Durham, N.C.: Carolina Academic Press.

Jansen, T. (2008). Once a physicist: Theo Jansen. Physics World, 21(10), pp.58-58.

Matthews, C. (2014). The Differences Between Centrifugal Pumps Vs. Positive Displacement Pumps – Empowering Pumps. [online] Empowering Pumps. Available at: differences-centrifugal-pumps-vs-positive- displacement- pumps/ [Accessed 9 Apr. 2016].

Minh, T., Kamers, B., Ramon, H. and Van Brussel, H. (2012). Modeling and control of a pneumatic artificial muscle manipulator joint — Part I: Modeling of a pneumatic artificial muscle manipulator joint with accounting for creep effect. Mechatronics, 22(7), pp. 923-933.

Image Reference

Figure 1: HRG, (2004).Research Projects. [online] Delft University of Technology. Available at: http:// architectural-engineering-and-technology/ organisation/hyperbody/research/research-projects/ [Accessed 20 Mar. 2016].

Figure 2(a): Juncheng, C. (2016). Soft walking pyramid.

Figure 2(b): Juncheng, C. (2016). Three legged robot.

Figure 3: Juncheng, C. (2016). Golem-kit: A pneumatic multipurpose modular toolkit.

Figure 4: SMC, (1997). BASIC PNEUMATICS: a manual for fluid power components and practical applications. Franklin: Indianapolis.

Figure 5(a): Shadow, (2016). Shadow Air Muscles. [online] Available at: products/air-muscles/ [Accessed 22 Mar. 2016].

Figure 5(b): Festo, (2016). Round cylinders | Festo Great Britain. [online] Available at: https:// [Accessed 20 Jul. 2016].

Figure 6: Veneva, I., Vanderborght, B., Lefeber, D. and Cherelle, P. (2013). PROPULSION SYSTEM WITH PNEUMATvIC ARTIFICIAL MUSCLES FOR POWERING ANKLE-FOOT ORTHOSIS. Journal of Theoretical and Applied Mechanics, 43(4).

Figure 7: Juncheng, C. (2016). The homemade PMA in Golem.

Figure 8: Juncheng, C. (2016). Contraction ratio test.

Figure 9: Hooper, J. (2013). Basic pneumatics. Durham, N.C.: Carolina Academic Press.

Figure 10: Bengtson, H. (2011). Differences Between the Centrifugal Pump and Positive Displacement Pump. [online] Brighthub Engineering. Available at: http:// hvac/62815-the-positive-displacement-and- centrifugal- pump-and-how-they-work/#imgn_7 [Accessed 9 Apr. 2016].

Figure 11(a): rs-online, (2016). RS Pro Diaphragm Electric Operated Positive Displacement Pump. [online] Available at: positive-displacement-pumps/7026898/.

Figure 11(b): Flojet, (2016). Xylem Flojet Diaphragm Electric Operated Positive Displacement Pump. [online] Available at: positive-displacement-pumps/7059327/ .

Figure 12: Softroboticstoolkit, (2016). Fluidic Control Board. [online] Available at: http://

Figure 13(a): Juncheng, C. (2016). Pneumatic control board A.

Figure 13(b): Juncheng, C. (2016). Pneumatic control board B.

Figure 14: Anker, (2016). Anker Foldable 21W Solar Charger.

Figure 15: (2016). Horsepower required to Compress Air. [online] Available at: horsepower-compressed-air-d_1363.html [Accessed 14 Jul. 2016].

Figure 16: Juncheng, C. (2016). Tension force experiments.

Figure 17(a): Juncheng, C. (2016). Experiments result.

Figure 17(b): Juncheng, C. (2016). Linear growth diagram.

Figure 18: SMC, (1997). BASIC PNEUMATICS: a manual for fluid power components and practical applications. Franklin: Indianapolis.

Figure 19: Juncheng, C. prototype I.

Figure 20(a): Juncheng, compressor diagram.

Figure 20(b): Juncheng, process.

Figure 21: Juncheng, C. mechanism in prototype.

Figure 22: Juncheng, C. precess of prototype I.

Figure 23: Juncheng, C.(2016).Prototype I test with different voltages 5V/12V.

Figure 24: Juncheng, C. (2016).Golem compressor prototype II.

Figure 25: Juncheng, C. (2016). Grasshopper simulation.

Figure 26: Juncheng, C. (2016).Servo air flow controller.

Figure 27: Juncheng, C. (2016).Photographs of pumping process.

Figure 28: Draper. (2016). Draper 28820 Expert Aluminium Refillable Pressure Sprayer, 650cc.

Figure 29(a): Juncheng, C. (2016).Golem-kit inventions, Tensigrity robot .

Figure 29(b): Juncheng, C. (2016).Golem-kit inventions, Three legged robot.

Figure 30: Juncheng, C. (2016). Basic module of Golem.

Submit a Comment