2. Soft Robotics

2.1. Soft robots

A conventional robot is defined as a machine made of rigid materials which can be designed and programmed to carry out a complex series of actions automatically.  Traditional rigid robots are characterised by their incredibly powerful and precise performance abilities.  However, being made of rigid materials with rigid joints, traditional robots and machines lack the ability to elastically deform and adapt their shape to external constraints which can also be defined as lack of multi-functionality (Majidi 2014). In addition, Rolf Pfeifer, director of the Artificial Intelligence Laboratory at the University of Zürich, has stated that rigid body systems have limited sensitivity, motility, self-awareness and the capability to adapt themselves as well to their environments (Wihart 2015). The lack of compliance and limited adaptability of rigid robots lead to imperfect human-machine interaction, thus in factories human and robotic workspaces are normally separate to address safety concerns.

As the development of robotics continues to expand beyond manufacturing and industrial automation and into the domains of cooperative human assistance and healthcare, robots are required to be more sensitive, compliant and adaptive towards the human body and its environment (Majidi 2014). Therefore, robots must become less rigid in their materiality to approach the ideals of mechanical compliance and multi-functionality, as materials and organisms do in nature. Put simply, the next generation of robots must be elastically soft and show higher capabilities in interacting with humans or navigating through environmental constraints.

Softness and body compliance are significant features normally apparent in biological systems which contribute to a more natural interaction with their environment. By applying biological principles to the design and control of robots, the new generation of robots referred to as soft robots, is expected to exhibit better adaptability towards unpredictable environments and to exhibit a wide range of behaviours with which humans can establish safe and smooth interaction (Pfeifer et al. 2013).

A robot is classified as a hard robot or a soft robot based on its underlying material (Rus & Tolley 2015). Distinct from conventional robots, soft robots contain little or no rigid material and are instead, mainly composed of soft materials (Majidi 2014 ). As a result, the main strategies in soft robot design are based on the characteristics of soft materials and components, such as compliance and passive dynamics (Wihart 2015). A machine constructed from soft materials with their inherent compliance is less likely to injure a human being. Passive dynamics refers to the ability of the machine to take an under-actuated part in movement instead of resisting movement tendencies.

2.2. Soft materials

In the realm of physics, rigid materials and soft materials are classified by their material rigidity which is measured using the modulus of elasticity also known as Young’s modulus. Young’s modulus is named after the British scientist Thomas Young and measures the stiffness of a solid material.  It defines the relationship between stress and the proportional deformation in a material.  Young’s modulus is an appropriate scale for comparing the rigidity of the materials distributed within a robot.

Most traditional robots are composed of materials with moduli of 10⁹—10¹² Pa such as metal and hard plastics.  However, most of the material in natural organisms such as muscle tissue and human skin normally has a modulus of 10⁴—10⁹ Pa (Rus & Tolley 2015). Thus, the essential reason for biological incompatibility when rigid robots and humans interact is caused by mismatched mechanical compliance between rigid materials and natural organisms (Majidi 2014). Hence, Rus and Tolley (2015) defined soft robots behaving autonomously, as systems that are mainly composed of soft materials with similar moduli to materials present in natural organisms, in the range of  10⁴—10⁹ Pa.

In this case, soft materials refer to silicone elastomers with moduli of 10⁴—10⁵ Pa, rubber with a modulus of about 10⁷—10⁸ Pa and low-density polyethene with a modulus of about 10⁸—10⁹ Pa. (Figure 1). It has become apparent that using these soft materials with a similar Young’s Modulus to materials found in natural organisms is beneficial in improving safety coefficients in the interaction between human and robots.

Figure 1: Young’s Modulus (Pa)

Majidi (2014) claimed that soft robots are mechanically biocompatible and capable of lifelike functionalities since they are composed of materials that match the compliance of biological matter. The advantages of soft robots are briefly summarised as follows, by Rus and Tolley (2015): soft robots offer an opportunity to bridge the gap between machines and humans and have a relatively higher level of multi-functionality compared to rigid body robots owing to their continuously deformable structure with bio-inspired muscle-like actuation; soft robots possess the potential to exhibit advanced adaptability, agility and sensitivity.

2.3. Control

Depending on soft material characteristics such as primary mechanical behaviour, high stress tolerance, elasticity and adaptability, controllable behaviours present a wide spectrum of design opportunities (Wihart 2015). The movements of conventional rigid body robots are described using six degrees of freedom which can be characterised by three rotations on the x, y and z axes and three translations on the x, y and z axes.  Soft robots, however, which are composed of soft materials with the potential to inflate, bend, twist and stretch, provide an unpredictable number of degrees of freedom. As a consequence, the movements of soft body robots are more of a challenge to define and control and require a series of new approaches to design, fabrication and control.

A soft robot consists mainly of an actuator, a sensor and controllers. The actuator is responsible for moving the robot. It requires a control signal and a source of energy. The sensor is a device that can detect and respond to input information and physical environments. Controllers are the devices installed or activated to guide or regulate the movements of a machine. As regards energy sources, soft robotic actuators are normally fluid powered by gas or liquid.

Basically, soft robot behaviours such as inflating, bending and twisting are achieved by pneumatic deformation.  Normally, pneumatic deformation is caused by pressure change inside air chambers. An air chamber is an enclosed space in which pneumatic deformation can occur. Soft robotic actuator behaviours depend on the air chamber design. For instance, a single geometrically shaped air chamber can inflate, while a series of chambers arranged in a row can bend. An example would be the PneuNet actuator (Figure 2 on the left). However, inflation is always the primary event in these behaviours. The PneuNet actuator is able to achieve two different bending behaviours based on inflation at different rates. Low rate inflation triggers the actuator to cause all air chambers to bend into a circular shape, all inflating to the same level (Figure 2 on the right), while a higher rate of inflation triggered by the actuator causes the air chambers to curl up, starting with the first inflated air chamber (Mosadegh et al. 2014).

Figure 2: The PneuNet actuator

2.4. Development and applications

As discussed above, the primary advantage of soft robot systems compared to conventional rigid body robots, is their material compatibility with natural organisms. Thus, soft robots have the potential to integrate a degree of compliance into wearable devices, whereas robots composed of rigid materials have more difficulties in interacting with humans, and rarely achieve the motion generated by natural components and joints.   

The key challenges for creating soft machines that achieve their maximum potential is the development of controllable soft bodies by using soft materials that integrate sensors, actuators, and computation thus enabling the robots to deliver the expected behaviours (Wihart 2015).  Therefore, one of the challenges facing designers could be stated to be the development of stretchable, portable, powered actuators. Another challenge can be defined as the design of more controllable soft robots or in other words, to improve the control of soft robots. This paper mainly focuses on discussing how to address the challenge of controlling soft robots by distributing varied property materials in their design.

3. Variable Property Design

A novel material distribution design system named “Variable Property Design” is referred to and applied to optimise soft robotic design and fabrication.

3.1. Variable Property Design (VPD)

Natural structures have always been an inspiration for architectural design due to the high level of seamless integration and precision with which they serve their function.  In nature, shapes and forms result from a matching between material properties and their corresponding environmental limitations. Also, biomaterials are generated by corresponding material properties with those external constraints they are subjected to, such as structural, environmental and performance criteria. Consequently, the geometry in natural forms is determined by the interaction between material and environmental limitations.

How can we apply nature’s ability to gradually distribute material properties to meet varied external requirements and shift design methods from form-driven to material-driven not just in industrial product application but also on an architectural scale?

Variable Property Design (VPD) is, therefore, a design approach, a methodology, and a technical framework, by which to model, simulate and fabricate material assemblies with varying properties designed to correspond to multiple and continuously varied functional constraints (Oxman 2010). The purpose of Variable Property Design is to optimise design by distributing varied material properties to corresponding performance constraints, as well as to improve the integration of modelling, analysis and fabrication in the work process. Therefore, Variable Property Design provides the potential for distributing material with varying properties in industrial products and in architectural designs which are determined by structural and environmental performance criteria. Besides, certain design methods provide significant potential for approaching the task from a performance-driven perspective, rather than form-driven which can help in achieving a nature-inspired design.

Variable Property Design (VPD) includes the three processes of Variable Property Modelling (VPM); Variable Property Analysis (VPA) and Variable Property Fabrication (VPF).

3.2. Variable Property Rapid Prototyping (VPRP)

Variable Property Rapid Prototyping (VPRP), as a novel method for Variable Property Fabrication (VPF), is designed to gradually vary the intensity of properties in a material. The purpose of VPRP technology is to achieve physical prototyping by graduating properties distributed in a material to correspond with performance constraints.

The typical existing prototyping approach is to assemble a structure using predetermined materials with constant physical properties.  The design and fabrication process occurs using homogeneous materials with predetermined constant properties. Influenced by traditional approaches used in industry, existing fabrication strategies assign various homogeneous materials to particular regions in order to meet specific structural and performance requirements. In traditional architectural design, for instance, in buildings constructed of steel and glass, steel only serves to meet structural requirements while glass only has the function of a building envelope.  In this case, steel and glass are considered as two different homogeneous materials only contributing to specific performance constraints with their unique material properties.

The inspiration by biological forms provides significant potential for the design and fabrication of structural components using materials with varying properties, thus helping to integrate form with function. MIT media lab presents a novel approach entitled Variable Property Rapid Prototyping, which allows designers to create building components with gradient properties designed to correspond with multiple and continuously varied functional constraints (Oxman 2011). She defined the purposes of Variable Property Rapid Prototyping as ‘ to enhance component optimisation of material properties relative to their structural performance as well as to enhance the relation between modelling, analysis and fabrication protocols across media by providing a more efficient and integrated workflow.’

3.3. Variable Property Design Application

3.3.1. Monocoque (architectural structural skins)

A monocoque is an experimental self-supporting architectural skin as described by Neri Oxman (2007). A monocoque embodies a novel design process which enables a matching of material properties with environmental constraints based on a variable property design method (Figure 3 on the left).

In conventional construction, architectural skins are considered to be a combination of inner boundary and external layer with no load-bearing capability.  The design of a monocoque, on the other hand, considers it an integral load-bearing component, also functioning as an architectural skin, thus distinguishing it from traditional design methods. As a result, the monocoque integrates both internal structural framework requirements and external performance requirements into a continuous structural surface.

The demands on a monocoque as a novel architectural skin are to provide self-supporting, load-bearing without any functional, regional distinction between structural and non-structural areas.  Designers achieve this requirement by distributing materials with varying properties such as stiffness and thermal expansion in a continuous load-bearing form, generated using a Voronoi algorithm, the material density corresponding with the assumed weight load (Figure 3 on the right). Thus, the thicker areas with vein-like elements, are designed to bear the majority of the force, while the thinner areas contribute lower load-bearing properties.  Moreover, the distribution of varied materials with fluctuating properties is visualised on a grey scale, thus featuring brighter areas representing material with more stiffness and a higher thermal expansion value, while the darker parts represent softer material with a lower thermal expansion factor.

Figure 3: Monocoque

3.3.2. Beast (Performative Chaise)

The ‘Beast’ is a continuous surface chaise longue created as both structure and skin that meet load-bearing and physiological comfort requirements. ‘The chaise combines structural, environmental and corporeal performance by adapting its thickness, pattern density, stiffness, flexibility and translucency to load, curvature and skin-pressured areas respectively.’ (Oxman 2010)

Similar to a monocoque, the structural requirements of the Beast are first to achieve self-support without structural, material distinctions and secondly to be able to bear human body pressure. The environmental constraint is to be flexible enough to comfortably fit a human body in both a sitting and lying position.

In the initial concept, the form was designed to function as a chaise which could fit body curvature. The cellular pattern was applied to correspond to the form of the human body. Therefore, the smaller cells were arranged in higher curvature regions requiring a higher density, while the larger cells were located more sparsely in shallow curvature regions.  The varying material properties were distributed to correspond to structural requirements. The stiffer material was positioned in the vertical surface regions in which the structure is primarily under compression and softer material assigned to horizontal surface regions which are mainly under tension.  To provide for these constraints, skin pressure was mapped and data included in the initial, structural load analysis (Figure 4). The Beast needs to be more flexible and softer in regions where more body pressure is applied. The designer, therefore, adjusted the relative thickness of each cell to match the pressure level, thus, the thicker and softer material was located in higher pressure regions and stiffer material with less thickness assigned to lower pressure areas.   

Consequently, five different materials with varying densities, stiffness, flexibility and thickness were distributed onto one continuous surface according to curvature, load pattern and skin pressure analysis (Figure 5).  Moreover, in the Beast, the varied material properties were visualised using colour coding as well as using monocoque technology. The brighter colour represents softer material, while the darker colour indicates stiffer material.  The fabrication process was achieved using customised 3D printing technology.

Figure 4: Skin pressure map (left) & Figure 5: the Beast (right)

3.3.3. Carpal Skin (Prototype of Carpel Tunnel Syndrome Splint)

‘Carpal Skin’ is a prototype splint designed for carpal tunnel syndrome patients to treat their illness and better assist wrist movement.  Carpal tunnel syndrome is a condition in which excess pressure acting on the median nerve in the wrist leads to numbness, muscle atrophy and weakness in the hand. The Carpal Skin splint needs to be customised according to individual pain-profiles as in most muscular and nerve-related syndromes, instead of being mass produced.

The main requirement of Carpal Skin is to meet the patient’s anatomical and physiological needs while also considering possible activities, by customising a carpal glove.

To meet requirements, the designer first 3D mapped the pain-profile of a particular patient by which the intensity and duration of pain was illustrated. Depending on the pain profile analysis, a 2D distribution map of different material properties with varying stiffness was generated using a local, nonlinear interaction simulation-based reaction-diffusion algorithm (Figure 6). Finally, the 2D distribution map was projected onto the 3D scanned outer structure using 3D printing technology.

Figure 6: Reaction-diffusion algorithm

As a result, stiff materials were generally applied to assist lateral bending, while soft materials contribute to wrist support and comfort. The local changes in thickness corresponding to the functional requirements of protecting and safeguarding the wrist from hard surfaces (Figure 7 on the left). The material thickness also contributes towards meeting the requirement that the thicker bumps act as soft tissue to increase the flexibility and circulation in certain regions (Figure 7 on the right).

Figure 7: Carpal Skin_mateirial distribution

4. Variable material properties

Soft material properties should be analysed in the first step in order to optimise soft robotic design with material distribution.            

4.1. Material requirements of soft robots

As mentioned in Chapter 2, in the definition of soft robotics, soft material refers to materials with a similar Young’s Modulus to natural tissue, such as muscle and human skin which normally have moduli between 10⁴—10⁹ Pa. Under these conditions, the soft materials of choice are silicone elastomer with a modulus between 10⁴—10⁵ Pa, latex with a modulus between 10⁵—10⁶ Pa and higher strain rubber with a modulus between 10⁷—10⁸ Pa.

4.2. Soft material properties

4.2.1. Initial material selection

Silicone rubber (Young’s Modulus: 10⁴—10⁵ Pa)

Silicone rubber as a synthetic material is an elastomer mainly composed of silicon together with carbon, hydrogen, and oxygen. Typically, silicones are present in liquid form and able to be cured to a flexible solid material by mixing with a catalyst.  Moreover, additives can be applied to change material properties such as viscosity, colour and curing time. Furthermore, by adjusting the mixture ratios, silicone rubber can display variable material properties such as hardness, elasticity and tensile strength.  However, silicone rubber is very sensitive to the mixing ratios which have to be precisely controlled. Another character of silicone is that it is self-releasing, so silicone cannot be cast with other materials or adhere to other objects once fully cured. In addition, compared to other soft materials, especially natural materials, the obvious advantage of silicone rubber is the higher chemical stability, resulting in water resistance, low toxicity, higher durability and UV stability (Wihart 2015). As a synthetic material with adjustable material properties such as hardness, translucency and colour and higher chemical stability, silicone is widely used for prototyping.

Latex (Young’s Modulus: 10⁵—10⁶ Pa)

Latex is a stable dispersion (emulsion) of polymer microparticles in an aqueous medium (Agrawal & Konno 2009).  In nature, about 10% of flowering plant species exude latex after tissue damage. Plants emit natural latex as a milky suspension or emulsion of particles in an aqueous fluid, which when exposed to air, coagulates.  However, synthetic latex can be made by polymerising a monomer such as a styrene that has been emulsified with surfactants. Latex is widely used in soft robot prototyping due to its relatively low cost.

Polyurethane rubber (Young’s Modulus: 10⁷—10⁸ Pa)

Natural rubber, also called India rubber or caoutchouc, is an elastomer (an elastic hydrocarbon polymer) that was originally derived from latex, a milky colloid produced by the Pará rubber tree (Wihart 2015). Natural rubber, defined as a type of soft material, possesses the material properties of stretchability and flexibility.  However, similar to latex, natural rubber can easily turn dark and brittle by exposing it to UV light.

Synthetic rubber, also known as polyurethane rubber, is made of various chemical monomers under controlled conditions, as are other kinds of polymers. In a comparison with natural rubber, polyurethane rubber has the advantage of possessing adjustable material properties by controlling monomer proportions.  Wihart (2015) stated that polyurethane rubber as a synthetic material has elasticity ranging from very soft to stiff and available colour grades in a range from clear to opaque to amber.  Polyurethane rubber, in comparison to silicone rubber, displays higher adhesive and lower self-releasing properties. On a 1:1 mixing ratio, polyurethane rubber is much more controllable than silicone rubber on its own. Due to this fact and the larger range of hard and softness, synthetic rubber as a soft material is widely used in the soft robot prototyping process.

4.2.2. Criteria for the measurement of material properties

To improve the design and control of soft robots made of soft materials, it is essential to analyse soft material properties.  They have variable and more dynamic material properties, in comparison with conventional rigid materials. In addition, analysing material properties is also fundamental to the Variable Property Design method. There are basically three material property measurements for soft material: Shore hardness, behaviour under an external force and the ability to change state from liquid to solid.

4.2.2.1.  Shore hardness

The primary mechanical property to evaluate in soft materials is hardness, measured on the Shore hardness scale according to the US metallurgist Albert Ferdinand Shore in the 1920s who invented the Shore durometer. There are three scales to illustrate Shore hardness: Shore D hardness, Shore A hardness and Shore 00 hardness. Shore D hardness is employed to measure hard rubber and hard plastics. Shore A hardness is used to measure soft and flexible materials ranging from hard rubber to softer rubber. Shore 00 hardness is employed to measure softer materials such as silicone rubber. These three hardness scales overlap with each other (Figure 8).

Basically, Shore A hardness is quantified on a scale from 0 to 100. The highest Shore A hardness at 100, equates to the hardness of the rubber used in shopping cart wheels and hard hats.  At the other end of the scale, the lowest Shore A hardness is 0 which is equivalent to human skin hardness.  The Shore 00 hardness scale measures the hardness of very soft materials. The lowest Shore A hardness point at 0 is roughly equivalents to 45 on the Shore 00 hardness scale. Chewing gum has a Shore 00 hardness of 20.

The Shore hardness scale is rarely used in conventional rigid body robot design since the materials are generally assumed to be hard.  The Shore durometer is only applied to define and communicate the softness of materials (Wihart 2015).  Shore hardness is, therefore, the defining property when analysing materials in soft robotic design.

Figure 8: Shore hardness scales

4.2.2.2.  Behaviour under external force

To distinguish them from rigid materials, the mechanical properties of soft materials are normally described by their material behaviour under external stress. This is measured using the parameters of tensile strength, elongation at break and the 100% modulus (Wihart 2015).

Tensile strength, measured in psi refers to the force needed to stretch the material until it breaks. In other words, tensile strength represents material stretchability. Higher tensile strength indicates lower stretchability. Elongation at break, measured as a percentage of a material’s original dimensions refers to the extent to which the material stretches before it breaks.  Hence, elongation at break data expresses material tenacity, in which the higher value represents higher tenacity. The 100% modulus, measured in psi refers to the force needed to stretch the material to twice its original dimensions. As a result, the 100% modulus represents material flexibility in which a higher force required, relates to lower flexibility.   

Soft materials, in comparison to rigid materials, have a higher adaptability to external environmental changes.  Thus, the comparison and analysis of material behaviour under external force are essential to the design of soft robots and their control.

4.2.2.3. Property changing ability from liquid to solid   

Soft materials also reveal specific behaviour during the morphologic change from liquid to solid such as cure time and shrinkage. Cure time is the period of time needed for a material to fully cure. Shrinkage represents the relative change in dimension between the length measured on the mould and the length of the moulded object 24 hours after it has been taken out of the mould.

The property changing ability of materials and its analysis is a main contribution when deciding on a soft robotic fabrication method.

4.2.3. Material property analysation

To compare and analyse nine selected materials, relevant material property data are listed in Chart 1, including Shore hardness (Shore A hardness and Shore 00 hardness), tensile strength, elongation at break, 100% modulus, cure time and shrinkage.

Chart 1: Soft material properties

Compared to polyurethane rubber and silicone rubber, Trylon latex liquid rubber has the highest shrinkage at 10% of the original length when fully cured after 24 hours. Furthermore, as the result of poor resistance to UV light and oxygen, latex tends to turn dark when exposed to an outside environment. Therefore, to retain design details and a sustainable performance ability, we decided to use a material with higher stability such as polyurethane rubber and silicone rubber, instead of latex.

As seen from the material property data sheet, polyurethane rubber Smooth-On Econ® 60 has the highest Shore A hardness compared to all of the selected silicone rubbers. This is equivalent to a car tyre tread.

Seven selected silicone rubbers are listed on the basis of their Shore hardness levels.

In seven selected silicone rubbers, Ecoflex® 00-30 has the lowest Shore hardness 00 at 30, which indicates a very low hardness level equivalent to a racket ball or gel shoe insole while SORTA Clear® 40 has the highest Shore hardness A at 40, which is similar to an inner tube or pencil eraser. With the lowest tensile strength of 200 psi and smallest 100% modulus at 10 psi, Ecoflex®00-30 is the most stretchable and flexible material.   Displaying the same properties as Ecoflex®00-30 in tensile strength, elongation at break and 100% modulus, Ecoflex® 00-35 Fast has the remarkable property of the shortest cure time. As a result, Ecoflex® 00-35 Fast is more efficient in rapid casting and displays better performance when used as an adhesive while Ecoflex®00-30 works better in casting with more detail.

With the highest Shore hardness A at 40, SORTA Clear® 40 has the highest tensile strength at 800 psi, the lowest elongation at break percentage at 400% and the highest 100% modulus at 90 psi which represent the lowest stretchability and flexibility.  SORTA Clear® 37 has a much shorter cure time compared with SORTA Clear® 40, which indicates higher efficiency in the fabrication process. In addition, Dragon Skin® 10 Fast has a tensile strength of 475 psi with the highest elongation at break factor at 1000%, which reflects a better tenacity with high flexibility.

4.3. Initial experiments

Through initial material property analysation, we gained an elementary knowledge of each selected material with their hardness, stretchability, flexibility and stability. However, these physical material properties were measured under non-functional conditions. Material performance should, therefore, be further examined under functional conditions with a view towards certain performance requirements.

As mentioned in Chapter 2, the primary performance requirement of a soft robotic actuator is inflation. Soft robots can achieve performance requirements such as inflating, bending and twisting based on pneumatic deformation. Inflation is always the primary stage in these behaviours since a bending motion is triggered by the inflation of air chambers as well as a twisting motion.

Therefore, to research soft robotic actuator behaviours using different material properties, it is essential to evaluate materials with varied properties in the light of inflation performance. Consequently, we designed a series of experiments to evaluate the performance of materials with varied properties under inflation.

4.3.1. Single side inflation

The aim of this series of experiments was to evaluate the inflating behaviour of a soft robotic actuator using materials with different properties.  In this case, we simplified a soft robotic actuator to a single unit, cylinder-shaped air chamber.

Since inflating behaviour is more dependent on the flexibility of a material, we applied two of the most flexible materials in order to receive more apparent results. They are Ecoflex® 00-30 with a tensile strength of 200 psi and a 100% Modulus of 10 psi, and Ecoflex® 00-50 with a tensile strength of 312 psi and a 100% Modulus of 12 psi.  We also limited the thickness of the top layer of air chambers to 2mm.Additionally, to compare soft robotic actuator inflation behaviours better, using these two different materials, we employed a series of cylinder-shaped air chambers with graduated diameters of 10mm, 20mm, 30mm, 40mm, 50mm and 60mm.In particular, we aimed to measure inflation behaviour under conditions with no additional external pressure so we decided to measure single side inflation. In soft robotic actuator design, to achieve single side

In particular, we aimed to measure inflation behaviour under conditions with no additional external pressure so we decided to measure single side inflation. In soft robotic actuator design, to achieve single side inflation, the base layer should be non-inflatable.

Basically, there are two methods to make a non-inflatable layer (Figure 9). The first method is to position a rigid layer such as paper or sheet metal inside a soft material layer. The second method is to employ the stiffer material. The limitation of the first method is seen in three-dimensional fabrication as the rigid layer of material has to be pre-shaped in order to be positioned inside the soft material layer.  Therefore, we decided to achieve single side inflation by using a stiffer material in the base layer. We, therefore, employed SORTA Clear® 37 silicone rubber for the base layer.

We measured inflating behaviour by the change in height of the air chambers (Figure 10).

Figure 9: Double sides inflation, Single side inflation with rigid sheet, Single side inflation with stiffer layer.

Figure 10: Single side inflation experiments

Chart 2: Single side inflation data

As shown in Chart 2, the difference in height change data between the two materials indicates that material with lower tensile strength and a lower 100% Modulus has the better inflation behaviour. In another word, the material with higher flexibility can achieve greater deformation.  Moreover, comparing the changes in height with different air chamber sizes, the bigger air chambers contributed to a higher inflation result.

4.3.2. Multiple layer inflation

To further explore inflation behaviour using materials with varied properties, we experimented on multiple layers inflating behaviour.  In this series of experiments, we discovered multiple inflation behaviours by using materials with varying properties inspired by the variable properties design method.  We first analysed the performance requirements of each layer and distributed material with different stiffness to meet the requirements.

We aimed for multiple layer inflation by stacking air chamber above air chamber and to achieve step by step inflation that starts from the lowest layer and gradually expands to the upper layers. To achieve a certain behaviour, we employed stiffer material in the lowest layer, as air pressure increases here first where the air is pumped in, and more flexible material on the upper layers. The required air pressure of the lowest layer should be lower than the required pressure of the upper layers. The air must be able to reach the upper layer which triggers their inflation.

We first experimented on two-layer inflation. For the lower air chamber, we employed Dragon Skin® 10 Fast as the top layer and Mold Star® 20T as the base layer.  For the upper air chamber, we employed Ecoflex® 00-30 as the top layer while its base layer is shared with the top layer of the lower air chamber. As the result, the lower air chamber reached an amplitude of 20mm while the upper air chamber inflated for another 40mm reaching 60mm in total.

We experimented further on four layer inflation. For the first (lowest) air chamber, we employed Dragon Skin® 10 Fast as the top layer and Mold Star® 20T as the base layer. For the second air chamber, we employed Ecoflex® 00-50 as the top layer while the base layer is shared with the top layer of the lowest air chamber. We also employed Ecoflex® 00-50 as the top layer of the third air chamber sharing the base layer with the top layer of the second air chamber. For the fourth air chamber, we employed Ecoflex® 00-30 as the top layer and shared Ecoflex® 00-50 as the base layer with the top layer of the third air chamber.  As a result, the first (lowest) air chamber reached an amplitude of 42mm and the fourth (top) air chamber added another 23mm resulting in a total inflation amplitude of 65mm. The two air chambers in the middle did not inflate which indicated that Ecoflex® 00-50 was not flexible enough (Figure 11).

Besides, we found that Mold Star® 20T is not stiff enough to be a base to support multiple inflated layers since the inflation of the lowest air chamber was double sided in the experiment.

Figure 11: Multiple layers inflation experiments

5. Prototypes

At the design stage, we came up with several prototypes to try to achieve higher performance by distributing variable material properties. In this chapter, materials with varying properties are distributed in both horizontal and vertical directions influenced by the variable material design method.

As mentioned in Chapter 3, generally Neri Oxman distributes materials on a continuous surface depending on performance requirements. Therefore, the material distribution method could be described as vertical distribution where materials are all employed in the same layer.  On the other hand,  materials distributed in different layers could be described as horizontal distribution (Figure 12).

Figure 12: Horizontal distribution (left) and Vertical distribution(right)

5.1.  Horizontal distribution

In horizontal distribution, materials are classified in three hardness levels, H-level 1 (the hardest level), H-level 2 and H-level 3 (the softest level).

5.1.1.  Scales

Animal scales are small and stiff plates connected with inner skin which functioned as protective outer skins. Pangolins, mammals whose scales are made of keratin, the same material as human fingernails, are fully covered in scales to protect themselves from danger. The movement of scales is normally triggered by the underlying skin expanding and bending.

Inspired by pangolin scales, we designed a prototype “Scales”, in which soft robotic actuators made of single unit air chambers function as underlying skin to trigger the movement of the outer layer of scales (Appendix 1). The performance requirement in this prototype was to push out surface scales by the inflation of air chambers underneath. Thus, the scales needed to be stiff enough to avoid bending and the air chambers underneath needed to display good inflating behaviour. We employed thick Mold Star® 20T (H-level 1) as the outer layer of scales and Ecoflex® 00-30  (H-level 3) as the top layer of the air chamber, while the base layer of the air chamber was made of Mold Star® 20T  (H-level 1) to achieve single side inflation in this case (Figure 13).

Figure 13: Scales prototype_material distribution

5.1.2. Valves

Inspired directly from the design concept of concealing and revealing we designed a prototype named “Valves”.  “Valves” aimed to conceal the air chambers at first and then reveal them after being powered.  We designed an outer layer complete with cracks on top of the air chamber layer. In the first stage, the air chambers were covered by the outer layer, but they pop out through the cracks after inflation (Appendix 2).

In this case, we needed to use the stiffer material for the outer layer to keep the original shape and more flexible material for the air chamber to inflate and pop out.  Consequently, we employed Mold Star® 20T (H-level 1) as the outer layer, Ecoflex® 00-30  (H-level 3) as the top layer of the air chamber and Mold Star® 20T (H-level 1) as the base layer of the air chamber to guarantee single side inflation (Figure 14).

Figure 14: Valves prototype_material distribution

5.2.  Vertical distribution

In vertical distribution, materials are classified on three flexibility levels, F-level 1 (the most flexible level), F-level 2 and F-level 3 (the least flexible level).

5.2.1. Grid matrix

Developed from initial single side inflation experiments, “grid matrix” aimed to achieve controllable inflation behaviours within a group of connected air chambers.

In the initial single side experiments, the inflation behaviour of a single unit air chamber depended on the material properties and size of the air chamber. Inflation behaviour is more diverse in connected air chambers. We designed a series of prototypes that we named “grid matrix” with connected air chambers to meet different performance requirements. We applied the variable material design method in designing this series of prototypes.

In grid matrix prototype series one (Appendix 3), we aimed to achieve three different inflation levels with three connected air chambers, all the same size, by distributing varying materials.  We first classified inflation behaviours on three levels, I-level 1, I-level 2 and I-level 3.  In this case, I-level 1 represented the most significant inflation behaviour, in another word, I- level 1 required the greatest deformation. Our experience gained in our initial single side experiment led us to decide to employ the same thickness of material with differing flexibilities, to achieve each level of inflation behaviour.  Thus, we employed the most flexible material Ecoflex® 00-30 (F-level 1) to achieve I- level 1 inflation, less flexible material Ecoflex® 00-50 (F-level 2) to achieve I- level 2 inflation and stiffer material Dragon Skin® 10 Fast (F-level 3) to achieve I- level 3 inflation (Figure 15).

Based on grid matrix prototype series one, we were able to further develop strategies for meeting more diverse performance requirements (Figure 16).

Figure 15: Grid matrix series one _ material distribution

Figure 16: Grid matrix series one further development _ material distribution

In grid matrix prototype series two (Appendix 4), we aimed to achieve the same inflation level with three connected air chambers in different sizes by distributing varying materials.  We set the three different air chamber sizes at diameters of 20mm, 30mm and 40mm.  Experience gained from the initial single side experiments led us to use the most flexible material for the smallest air chamber and stiffer material for the largest air chamber. Therefore, we employed Ecoflex® 00-30 (F-level 1) for the 20mm diameter air chamber, Ecoflex® 00-50  (F-level 2) for the 30mm diameter air chamber and Dragon Skin® 10 Fast  (F-level 3) for the 40mm diameter air chamber (Figure 17).

Based on the grid matrix prototype series two, we further developed strategies aimed at achieving the same level of inflation within multiple connected air chambers (Figure 18).

Figure 17: Grid matrix series two_matretial distribution

Figure 18: Grid matrix series two further development _ material distribution

6. Proposal for an Aposema Mask

6.1. Concept of an Aposema Mask

We aimed to design a mask which could allow wearers to control their exposure to the outside environment, using bubbles which can conceal and reveal.  We employed soft robotic technology to achieve dynamic movement and the variable materials design method to meet more varied preference requirements.

6.2. Requirements 

According to the Aposema Mask design concept, the primary performance requirement is to achieve concealing and revealing movements in the mask while the advanced performance requirement is to control movement to be uniform.

6.3. Design and fabrication 

We referred back to our “Valves” prototype to achieve concealing and revealing views from behind the mask, by popping out inflated air chambers.  Additionally, we designed the air chambers to inflate equally, by employing materials with variable properties based on the “grid matrix prototype series two” prototype.

We designed the size of the bubbles to conceal or reveal information, based on human sensory organs. The anatomical transverse dimension of the human eye is around 24.2mm and the sagittal dimension around 23.7 mm. Oladipo et al. (2009) have indicated that normal nasal width and height in males were 38.3mm and 42.6mm respectively, while those of females were 37.4mm and 44.7mm respectively. Thus, air chambers with similar sizes to human sensory organs with diameters in the range of 20mm, 30mm, 40mm and 50mm served as the concealing or revealing bubbles. To control the air chambers in different regions of the mask, we divided them into six. The first group, located on the forehead, consisted of one 50mm diameter air chamber, one 40mm diameter air chamber and one 30mm diameter air chamber. The second group located around the eye and nose area consisted of one 30mm diameter air chamber and two 20 mm diameter air chambers (figure 19).   

Firstly, we distributed materials with different properties horizontally, to achieve initial structural requirements. We employed the stiffest material to allow the base layer to keep its three-dimensional shape, the most flexible materials in the air chamber layer and a stiffer material in the outer layer. Consequently, we employed SORTA Clear® 37 in the base layer, Mold Star® 20T in the outer layer and the most flexible materials Ecoflex® 00-30 and Ecoflex® 00-50 in the air chamber layer.

Secondly, we distributed two materials with different properties vertically, to achieve the higher performance requirement of same level inflation. For the first group, located on the forehead, we employed Ecoflex® 00-50 (F-level 2) in the 50mm diameter air chamber and Ecoflex® 00-30  (F-level 1) in the 40mm and 30mm diameter air chambers.  For the second group located around the eyes and nose, we employed Ecoflex® 00-50  (F-level 2) in the 30mm diameter air chamber and Ecoflex® 00-30  (F-level 1)in the 20mm diameter air chambers (figure 20).

Figure 19: Aposema Mask diagram

Figure 20: Aposema Mask

7. Conclusion

This paper discussed strategies of distributing materials with varying properties to meet wearable, soft robotic design requirements based on the Variable Property Design method and finally being realised in the Aposema Mask.

Soft robotics, described as a new generation of robot design, is expected to exhibit higher compliance in human-machine interaction and a wider range of behaviours towards unpredictable environments. The technique is however challenged, to further develop appropriate designs and manipulation systems with stretchable, portable actuators and more accurate control based on material distribution. Therefore, a novel material distribution method named “Variable Property Design” was introduced. The developer, Neri Oxman, has described Variable Property Design (VPD) as a design approach, a methodology, and a technical framework that could simulate and fabricate material assemblies with varying property, material distribution to correspond to varied functional constraints. The technique was illustrated by analysing three design cases, “Monocoque”, the “Beast” and “Carpal Skin” in which materials with varying properties were distributed on a continuous surface to meet specific structural and performance requirements.

In the initial design stage, soft materials with varying properties were investigated and tested. To sum up, soft materials such as polyurethane rubber and silicone rubber are distinct from one another due to their different levels of hardness, their flexibility and stretchability which contribute to variable performance behaviours. Materials with variable properties were employed in soft robotic air chambers to evaluate inflation behaviours which are essential to the design of soft robotic actuators. The material experiments indicated that silicone rubber with varying hardness and flexibility would trigger different inflation behaviours.

In the prototype design stage, the Variable Property Design method was applied in three prototypes to meet specific functional and performance requirements. Materials were horizontally distributed in the Scales prototype and the Valves prototype. For the Grid matrix prototype, materials were distributed vertically.

Finally, the Aposema Mask, illustrating an optimised, wearable soft robot was realised by applying variable material distribution. The essential functional requirement was achieved by distributing materials with varying properties, horizontally. The stiffest material was employed in the base layer to keep the three-dimensional form. The more flexible materials were distributed in the performance layer and the least flexible material, complete with cracks, was employed in the outer layer to enable the more flexible performance layer to display popping out behaviour. The performance layer required connected air chambers in different sizes to inflate to the same level. Thus, two of the most flexible materials were distributed according to the varying chamber sizes to enable them to reach the same level of inflation.

As for further developments, material experiments on soft robotic actuator performance could be extended beyond inflation behaviour to include bending and twisting which will lead to more variable soft robotic possibilities.  Additionally, soft materials with varying properties could be colour-coded in designs which would help visualisation of material and performance constraints.

8. Bibliography

1. Agrawal, A.A. and Konno, K., 2009. Latex: a model for understanding mechanisms, ecology, and evolution of plant defense against herbivory. Annu. Rev. Ecol. Evol. Syst., 40, pp.311-331.

2.  Bader, C., Kolb, D., Weaver, J.C. and Oxman, N., 2016. Data-driven material modeling with functional advection for 3D printing of materially heterogeneous objects. 3D Printing and Additive Manufacturing, 3(2), pp.71-79.

3. Iida, F. and Laschi, C., 2011. Soft robotics: challenges and perspectives. Procedia Computer Science, 7, pp.99-102.

4.  Kao, H.L.C., Holz, C., Roseway, A., Calvo, A. and Schmandt, C., 2016, September. DuoSkin: rapidly prototyping on-skin user interfaces using skin-friendly materials. In Proceedings of the 2016 ACM International Symposium on Wearable Computers (pp. 16-23). ACM.

5.  Majidi, C., 2014. Soft robotics: a perspective–current trends and prospects for the future. Soft Robotics, 1(1), pp.5-11.

6.   Mosadegh, B., Polygerinos, P., Keplinger, C., Wennstedt, S., Shepherd, R.F., Gupta, U., Shim, J., Bertoldi, K., Walsh, C.J. and Whitesides, G.M., 2014. Pneumatic networks for soft robotics that actuate rapidly. Advanced Functional Materials, 24(15), pp.2163-2170.

7.  Oladipo, G.S., Fawehinmi, H.B. and Suleiman, Y.A., 2009. The study of nasal parameters (nasal height, nasal width, nasal index), amongst the Yorubas of Nigeria. The Internet Journal of Biological Anthropology, 3(2), pp.1-11.

8.  Oxman, N. and Rosenberg, J., 2007. Material computation. International Journal of Architectural Computing, 1(5), pp.21-44.

9. Oxman, N., 2010. Material-based design computation (Doctoral dissertation, Massachusetts Institute of Technology).

10. Oxman, N., 2010. Structuring materiality: design fabrication of heterogeneous materials. Architectural Design, 80(4), pp.78-85.

11. Oxman, N., 2011. Variable property rapid prototyping: inspired by nature, where form is characterized by heterogeneous compositions, the paper presents a novel approach to layered manufacturing entitled variable property rapid prototyping. Virtual and Physical Prototyping, 6(1), pp.3-31.

12.  Pfeifer, R., Marques, H.G. and Iida, F., 2013, August. Soft robotics: the next generation of intelligent machines. In Proceedings of the Twenty-Third international joint conference on Artificial Intelligence (pp. 5-11). AAAI Press.

13.  Rus, D. and Tolley, M.T., 2015. Design, fabrication and control of soft robots. Nature, 521(7553), pp.467-475.

14. Stokes, A.A., Shepherd, R.F., Morin, S.A., Ilievski, F. and Whitesides, G.M., 2014. A hybrid combining hard and soft robots. Soft Robotics, 1(1), pp.70-74.

15.  Vogel, S., 2013. Comparative biomechanics: life’s physical world. Princeton University Press.

16.  Wihart, M., 2015. The Architecture of Soft Machines (Doctoral dissertation, UCL (University College London)).

17.  Wehner, M., Truby, R.L., Fitzgerald, D.J., Mosadegh, B., Whitesides, G.M., Lewis, J.A. and Wood, R.J., 2016. An integrated design and fabrication strategy for entirely soft, autonomous robots. Nature, 536(7617), pp.451-455.

Image References:

Figure 1: Rus, D. and Tolley, M.T. (2015). Figure 2: Approximate tensile modulus (Young’s modulus) of selected engineering and biological materials. Available at: http://www.nature.com/nature/journal/v521/n7553/fig_tab/nature14543_F2.html [Accessed 10 July. 2017]

Figure 2: Mosadegh et al. (2014). Figure 5. High-Speed Actuation. Available at: http://onlinelibrary.wiley.com/doi/10.1002/adfm.201303288/epdf [Accessed 10 July. 2017]

Figure 3: Oxman (2010). Monocoque. Available at: http://web.media.mit.edu/~neri/site/projects/monocoque1/monocoque1.html [Accessed 10 June. 2017]

Figure 4: Oxman (2011). Figure 12. Chaise design informed by material properties assigned to pressure map registration, body form and body weight. Available at: http://www.tandfonline.com/doi/pdf/10.1080/17452759.2011.558588?needAccess=true [Accessed 10 June. 2017]

Figure 5: Oxman (2011). Figure 16. Chaise performative. Prototype for a Chaise Lounge. Available at: http://www.tandfonline.com/doi/pdf/10.1080/17452759.2011.558588?needAccess=true [Accessed 10 June. 2017]

Figure 6: Oxman (2011). Figure 17. Material distribution charts. Available at: http://www.tandfonline.com/doi/pdf/10.1080/17452759.2011.558588?needAccess=true [Accessed 10 June. 2017]

Figure 7: Oxman (2011). Carpal Skin. Available at: http://web.media.mit.edu/~neri/site/projects/carpalskin/carpalskin.html  [Accessed 10 June. 2017]

Figure 8: Smooth-On (2017). Durometer Shore Hardness Scale. Available at: https://www.smooth-on.com/page/durometer-shore-hardness-scale/ [Accessed 10 July. 2017]

Figure 9: Peng, S. (2017). Double sides inflation, Single side inflation with rigid sheet, Single side inflation with stiffer layer.

Figure 10: Peng, S. (2017). Single side inflation experiments.

Figure 11: Peng, S. (2017). Multiple layers inflation experiments.

Figure 12: Peng, S. (2017). Horizontal distribution (left) and  Vertical distribution(right).

Figure 13: Peng, S. (2017). Scales prototype _ material distribution.

Figure 14: Peng, S. (2017). Valves prototype _ material distribution.

Figure 15: Peng, S. (2017). Grid matrix prototype series one _ material distribution.

Figure 16: Peng, S. (2017). Grid matrix series one further development _ material distribution.

Figure 17: Peng, S. (2017). Grid matrix series two _ material distribution.

Figure 18: Peng, S. (2017). Grid matrix series two further development _ material distribution.

Figure 19: Meyer, A., Peng, S. and Rueda, S. (2017). Aposema Mask diagram.

Figure 20: Meyer, A., Peng, S. and Rueda, S. (2017). Aposema Mask.

Chart 1: Peng, S. (2017). Soft material properties.

Chart 2: Peng, S. (2017). Single side inflation data.