Synergizing microfluidics with soft robotics: A perspective on miniaturization and future directions

Abstract

Soft robotics has gone through a decade of tremendous progress in advancing both fundamentals and technologies. It has also seen a wide range of applications such as surgery assistance, handling of delicate foods, and wearable assistive systems driven by its soft nature that is more human friendly than traditional hard robotics. The rapid growth of soft robotics introduces many challenges, which vary with applications. Common challenges include the availability of soft materials for realizing different functions and the precision and speed of control required for actuation. In the context of wearable systems, miniaturization appears to be an additional hurdle to be overcome in order to develop truly impactful systems with a high user acceptance. Microfluidics as a field of research has gone through more than two decades of intense and focused research resulting in many fundamental theories and practical tools that have the potentials to be applied synergistically to soft robotics toward miniaturization. This perspective aims to introduce the potential synergy between microfluidics and soft robotics as a research topic and suggest future directions that could leverage the advantages of the two fields.

INTRODUCTION

Microfluidics can be combined with soft robotics, but should it be? This question could spark as much interest as criticism. The answer may be yes; after all, both rely on fluids to do work, and both use similar materials such as polydimethylsiloxane (PDMS)^1,2 as the main building component. As a side anecdote, both fields can relate their rise to Professor George M. Whitesides, who first introduced a wide usage of soft lithography to microfluidics in 1998¹ and then introduced a soft robotic gripper² and a multigait soft robot³ in 2011 that kick-started the decade of increasingly rapid development in soft robotics. However, the fact that these two fields share certain similarities does not automatically justify that they should be combined synergistically or produce impactful fundamental and applied research developments for the benefit of human society.

The original intention is to review recent progresses in the interdisciplinary field involving both microfluidics and soft robotics; however, there have not been a sufficient number of published articles to warrant such a review article. Therefore, we will instead share our perspective on the current trends and prospects for the future of this interdisciplinary field. For a holistic view of the field of soft robotics, readers are referred to the following review and perspective articles on applications of soft robots,⁴ soft rehabilitation and nursing-care robots,⁵ soft grippers,⁶ manufacturing of soft robots,⁷ pneumatic energy sources for soft robots,⁸ perceptive soft robots,⁹ controlling and simulating soft robots from a thermodynamics perspective,¹⁰ human-in-the-loop development of soft wearable robots,¹¹ elastic inflatable actuators for soft robotic applications,¹² and grand perspectives.^13,14 We would also like to give special recognition to the work of Suzumori et al. on flexible microactuator and its applications to robotic mechanisms from 1991¹⁵ and the work of Grosjean et al. on micro-balloon actuators for aerodynamic control from 1998,¹⁶ as their pioneering work inspired the present perspective.

Microfluidics has made significant contributions to a wide range of areas such as life science research,¹⁷ personalized medicine,¹⁸ and air and water quality control;¹⁹ the list goes on and on. Microfluidics continues to impart tremendous impact and growth toward fundamental biology, which is evident through its involvement in recent genetics research.^20,21 The reason that microfluidics can bestow significance in so many areas is perhaps because of its versatility as a tool, a highly complex and useful mechanism that sprung out of fundamental fluid mechanics and micro-electromechanical system (MEMS) research. Behind microfluidics is a buildup of scientific knowledge of fluids behavior,²² control and manipulation,^23,24 of fluids in microsystems, and the philosophy regarding miniaturization of macroscale systems down to a small scale while imparting improvements and retaining the original intentions. Therefore, just maybe, the science and philosophy of microfluidics could be useful for soft robotics.

Up to this point, the soft robotics community and the microfluidics community have not seen much collaboration, which is evident by the low number of publications involving interdisciplinary research of the two fields. This is not surprising, as soft robotics is still a relatively new discipline, and microfluidics is commonly regarded as fundamental research revolved around life sciences.¹⁷ Also, the link and usage cases for microfluidics in soft robots and vice versa are unclear to researchers. Therefore, to help researchers on both sides of the aisle to fully appreciate and understand the potential impact and contributions of the interdisciplinary research of microfluidics and soft robotics, a brief overview, which is by no means exhaustive, of recent separate developments of microfluidics and soft robotics is examined below.

SOFT ROBOTICS OF THE 2010s

To understand the benefits of soft robots, it is better to discern the limitations of traditional hard robots first. In the words of Professor George M. Whitesides, “the most important limitation is that they, [hard robots], are what is called ‘non-collaborative.’”¹⁴ Indeed, hard robots are often best used in very predictable and programmable manners²⁵ requiring impeccable precision and accuracy, such as industrial settings. When it comes to interacting near humans, collaborations with humans, as well as handling soft and delicate objects, soft robots are advantageous over hard robots due to their inherent compliance. This is not to say that hard robots cannot interact with humans, rather a call for improvements upon the current robotic systems, which perhaps involves a hybrid makeup of both hard and soft robotics systems working together. However, as far as this perspective is concerned, the inherently non-threatening and compliant nature of soft matter allows soft robots to work with and help human synergistically.

Independent functioning of soft robots

Soft robots can be roughly categorized into two types: independent functioning soft robots and soft wearable robots. The former is the starting point of the field of soft robotics, which began by studying and mimicking animals lacking internal skeletons, such as squid, worm, and starfish. Examples (Fig. 1) of this category include soft robotic grippers² for handling soft agricultural and commercial goods as well as biological sampling on deep reefs,²⁶ autonomous soft robotic fish for documenting marine life as non-disruptively as possible,²⁹ an untethered soft robot³⁰ for traversing in harsh conditions, a “growing” soft robot²⁷ for potential applications in search and rescue in unstable rubble, and a modular soft manipulator for minimally invasive surgery.²⁸

FIG. 1.

(a) Soft robotic grippers are reproduced with permission from Ilievski et al., Angew. Chem. Int. Ed. 50(8), 1890–1895 (2011). Copyright 2011 John Wiley & Sons. (b) Bellows-type gripper collecting soft coral. Adapted from Galloway et al., Soft Robot. 3(1), 23–33 (2016). Copyright 2016 Author(s), licensed under a Creative Commons Attribution (CC BY) license. (c) A soft robot lengthens through various challenging constrained environments without active control. Alternatively, the robot passively deforms to navigate the obstacles. Reproduced with permission from Hawkes et al., Sci. Robot. 2(8), 1–8 (2017). Copyright 2017 The American Association for the Advancement of Science. (d) Stiffness controllable flexible and learnable manipulator for surgical operations performing bending and elongation. Reproduced with permission from Ranzani et al., IEEE Trans. Robot. 32(1), (2016). Copyright 2016 IEEE. (e) Details of a soft-bodied robotic fish. Marchese et al., Soft Robot. 1(1), 75–87 (2014). Copyright 2014 Author(s), licensed under a Creative Commons Attribution (CC BY) license. (f) Image showing the resilience of the untethered soft robot to snow conditions. Reproduced with permission from Karpelson et al., Soft Robot. 1(3), 213–223 (2014). Copyright 2014 Mary Ann Liebert, Inc.

Soft wearable robots

Soft wearable robots are becoming increasingly popular for applications in the biomedical field especially for making assistive technology and rehabilitation. Some examples (Fig. 2) include a soft wearable glove for at-home rehabilitation,³¹ a soft exosuit for improving walking in post-stroke patients,³² a soft wearable robot for the shoulder,³³ a soft wearable biomedical device for lymphedema treatment³⁵ as well as mechanotherapy³⁴ and a variety of applications for the lower body, such as for ankle–foot rehabilitation³⁶ and knee motion.^37,38 The reason for its popularity lies in its inherent compliance and collaborative characteristics to safely interact with humans.¹³ It is the consensus in the field of soft robotics that soft wearable robots do not replace traditional exoskeletons but instead add a new dimension to advance wearable assistive devices, as commented by Connor Walsh in a perspective article: “Soft wearable robots will not replace traditional rigid exoskeletons that are emerging as a valuable clinical tool for patients with severe impairments, but will instead offer complementary capabilities for applications which require soft systems.”¹¹

FIG. 2.

Examples of soft wearable robots: (a) a soft wearable glove for post-stroke rehabilitation. Reproduced with permission from Polygerinos et al., Rob. Auton. Syst. 73, 135–143 (2015). Copyright 2015 Elsevier. (b) Soft exosuit for post-stroke rehabilitation. Reproduced with permission from Awad et al., Sci. Transl. Med. 9(400) (2017). Copyright 2017 The American Association for the Advancement of Science. (c) Soft wearable robot for the shoulder. Reproduced with permission from O’Neill et al., “A soft wearable robot for the shoulder: Design, characterization, and preliminary testing,” in 2017 International Conference on Rehabilitation Robotics (ICORR) (IEEE, 2017), pp. 1672–1678. Copyright 2017 IEEE. (d) A soft wearable device for mechanotherapy. Reproduced with permission from Payne et al., “Force control of textile-based soft wearable robots for mechanotherapy,” in 2018 IEEE Proceedings of International Conference on Robotics and Automation (IEEE, 2018), pp. 5459–5465. Copyright 2018 IEEE. (e) Soft modular knee sleeve for assisting walking. Reproduced with permission from Park et al., “Active modular elastomer sleeve for soft wearable assistance robots,” in IEEE Proceedings of International Conference on Intelligent Robots and Systems (IEEE, 2012), pp. 1595–1602. Copyright 2012 IEEE.

Soft actuators

The mainstay of a soft robot is its soft actuators, which have indeed been the research focus even before soft robotics became a real research interest.⁶ Many of the soft actuators can trace their origins to robotic artificial muscles driven by the motivation to mimic certain aspects of muscle function. Zhang et al.³⁹ conducted a systematic review of various types of robotic artificial muscles. No single technology can tackle all the problems; certain soft actuators allow for high dynamic torque applications, for instance, the cable system present in the Exosuit⁴⁰ developed by Harvard's Biodesign Lab, whereas soft fluidic actuators may allow for better conformity to the body, especially across joints and muscles evident by soft robotic gloves⁴¹ and soft grippers⁴² for grasping soft and delicate objects and for applications involving haptics, such as those developed by HaptX.⁴³ As for smart material actuators such as shape memory alloys,⁴⁴ dielectric elastomers/electroactive polymers,^45,46 and electro/magneto-rheological fluid,^47,48 they still require more research and development and technical breakthroughs regarding safety, reliability, response time, force output, efficiency, and fabrications before being practical for soft robotic applications as suggested by Whitesides's¹⁴ call for better materials for soft robotics applications. The potential of smart material actuators is immense, as it can directly convert electrical energy into useful mechanical forces and torques. On the other hand, soft fluidic actuators and cable-driven actuators require a medium to convert electrical energy into mechanical energy, which is less efficient than direct energy conversion in theory.

When it comes to practical implementations of soft robots, soft fluidic actuators employing hydraulics (with extension to any liquid) and pneumatics (with extension to any gas) are by far the most common methods.¹⁴ Fluidic pressurization also represents the biggest bottleneck issue that must be solved for soft wearable robots to become genuinely relevant in helping people in everyday lives. Before diving into the challenges facing soft fluidic actuators, its advantages are briefed below. First, the fabrication of soft fluidic actuators is practical and reasonably straightforward. There are two primary types of fabrication methods: molding^30,49,50 and heat bonding⁵¹ (Fig. 3). Molding typically uses a two- or three-part 3-D printed or laser-cut mold, where silicone, commonly Elastosil by Wacker or Ecoflex by Smooth-On, is poured into the mold, cured, and bonded by adhesive to at least one other silicone part to create soft fluidic actuators. With the improvements in 3D printing technology, particularly stereolithography (SLA), creating high-quality molds has been streamlined and democratized. Heat bonding uses a heat press to bond two pieces of thermoplastics films, such as polyurethane, polyester, and nylon polyamide from 3M or Bemis, to form an elastic balloon. The same thermoplastic films can also be used to create the strain-limiting fabrics,⁵¹ which are used in conjunction with the soft fluidic actuators to produce motions, torques, and forces toward the body. Xurographic cutter plotters and laser cutters have automated the process of cutting complex shapes out of thin films allowing the creation of complex strain-limiting fabrics with micrometer precision and accuracy. This opens the door to complex Kirigami skin designs⁵² that, when combined with soft fluidic actuators, can create locomotion.

FIG. 3.

(a) 3D modeling representation of the molds for casting a soft robotic gripper. Reproduced with permission from Bilodeau et al., “Monolithic fabrication of sensors and actuators in a soft robotic gripper,” in IEEE International Conference on Intelligent Robots and Systems (IEEE, 2015). Copyright 2015 IEEE. (b) The heat press lamination fabrication approach to making a soft robotic glove. Reproduced with permission from Connolly et al., Extrem. Mech. Lett. 27, 52–58 (2019). Copyright 2019 Elsevier.

Second, soft fluidic actuators are highly versatile. They can be designed with or without strain-limiting layers to achieve a variety of motions, torques, and forces, and their size can be scaled up or down with accommodating tubing sizes to pair with. The variety of motion is created by inflation of multiple internal chambers, pneumatic bladders, or balloons in a preset sequence, which expands low modulus structures, while the higher modulus locations resist internal chamber expansion causing a desirable motion. The higher modulus locations can be made by only increasing the thickness of the material or by adding an additional strain-liming layer or material.

Third, soft fluidic actuators are economical. For instance, 1 l of Ecoflex silicone costs around US$40 and 0.84 m² of a Bemis elastomer membrane costs around US$20 from Bemis. If the production of soft fluidic actuators is tuned up to industrial scale, the per-unit cost is further decreased. The major cost of soft fluidic actuators comes from the control systems, which comprises solenoid valves, pumps, pressure sensors, microcontrollers, batteries, etc. The cost of the control system also depends on the application, the number of components, and the size of the soft robot; by rough estimation, the cost of the control system is 70%–95% of the total cost of making a soft robot.

Last, soft fluidic actuators can be miniaturized easily. Albeit, not all soft robots benefit from miniaturization, but it would be particularly useful for some, namely, soft wearable robots such as the soft robotic glove for neuromuscular rehabilitation.⁵³ Nevertheless, even in the realm of soft wearable robots, soft fluidic actuators suit certain applications and fall short for others. The limitations of soft fluidic actuators have been demonstrated by the Soft Exosuit project led by Harvard Biodesign Lab,⁵⁴ where Bowden cables used in the exosuit replaced McKibben artificial muscles, a type of soft fluidic actuators used in earlier prototypes.

A McKibben artificial muscle⁵⁵ is perhaps one of the most well-known soft fluidic actuators that were first introduced in the 1950s. It is a hollow soft braided tubular bladder that shortens when pneumatic or hydraulic energy is applied onto the inner surface of the tube. The advantage of McKibben artificial muscles is that it can generate large forces and has a high power density; hence, it has an excellent power to weight ratio, and it is simple and inexpensive to manufacture. A feedback PI or PID controller is sufficient to control McKibben artificial muscles.⁵⁶ The disadvantage and perhaps the bottleneck of a McKibben artificial muscle is that it requires a large compressor or a fluidic energy source, which limits its applications as untethered mobile systems.⁸ Here lies the first challenge: current soft fluidic actuators still lack robustness for applications requiring high dynamic torque, as the pressure requirement would be high, demanding a large pump, which would not be practical to be worn.

MOTIVATION TO BRING IN MICROFLUIDICS

Miniaturization

Miniaturization of soft fluidic actuators is perhaps a practical means to produce useful and aesthetically pleasing products for users to adopt and embrace—the first step of many to initiate the adoption of soft wearable robots as assistive devices. By all means, aesthetics is a highly subjective topic; depending on the person and the application of the soft wearable robot, the aesthetic tolerance is different. Minimizing the disruption to the human form factor is perhaps a universal aspiration when designing wearable systems; hence, it is a valuable proposition to focus on miniaturizing soft wearable robotic systems into a low-profile form factor that conforms to the human body without impeding motion.

Aesthetics comes with a cost in functionality for certain applications as miniaturization may reduce the amount of pressure, torque, and toughness of fluidic actuators. Retaining both functionality and aesthetics calls for fundamental and applied research, which must be application and end-user driven, meaning that the motivation of miniaturizing soft wearable robots should be problem-focused and only undertaken if it is achievable and satisfies end-user requirements. For instance, if 20 psi pressure or above is required to actuate soft fluidic actuators for a specific application, then miniaturizing the soft fluidic actuators is still desirable to reduce the bulk of the soft robot, but the pump required is too large and that would be the bottleneck, and perhaps an alternative solution exists.

Case study of the soft robotic glove

The soft robotic glove⁵⁹ made by Harvard Biodesign Lab provides insight into the progression of soft wearable robots. It is visually seen that the aesthetics of the glove improves while retaining its functionality from the early pneumatic network bending actuators⁵⁷ to the latest sew-free anisotropic textile composites⁵¹ (Fig. 4).

FIG. 4.

(a) Soft robotic glove prototype based on a pneumatic network soft fluidic actuator. Reproduced with permission from “Towards a soft pneumatic glove for hand rehabilitation,” in IEEE International Conference on Intelligent Robots and Systems (IEEE, 2013). Copyright 2013 IEEE. (b) Soft robotic glove based on fiber-reinforced and sleeved soft fluidic actuator. Adapted from Cappello et al., J. Neuroeng. Rehabil. 15(1), 1–10 (2018). Copyright 2018, licensed under a Creative Commons Attribution (CC BY) license.⁵⁸ (c) Soft robotic glove based on sew-free anisotropic textile composites. Reproduced with permission from Connolly et al., Extrem. Mech. Lett. 27, 52–58 (2019). Copyright 2019 Elsevier.

Low-profile, high resolution, and digitization

As previously mentioned, aesthetics is a highly subjective issue, but it can be commonly regarded that low-profile systems are more preferable than bulky systems if the same functionality can be maintained. Hence, the idea is to use multiple smaller low-profile soft fluidic actuators to mimic the functionality of a larger soft fluidic actuator, using a simple balloon as an example (Fig. 5).

FIG. 5.

Concept image showing the difference in the profile between a single large pneumatic bladder constrained by a high stretch strain-limiting fabric (top left), a single large pneumatic bladder constrained by a low stretch strain-limiting fabric (bottom left), miniaturized pneumatic bladders constrained by a low stretch strain-limiting fabric (top right), and miniaturized pneumatic bladders constrained by a low stretch strain-limiting fabric on a curved substrate (bottom right).

Miniaturizing soft fluidic actuators is most persuasive in applications where compliance, compression, and conformity are highly valued, and it also opens the door for potentially better control systems as each actuator can be considered a single digital point. In such a strategy, instead of inflating a single large balloon by reading analog pressure data, a miniaturized balloon can just be considered as an on or off switch, inflated or deflated. When combined in a matrix or an array with strain-limiting layers synergistically, they potentially allow for more efficient motion or means of providing torques and forces, achieving the same functionality as a single soft fluidic actuator, which will be elaborated on in the next section.

It should be noted that not all soft fluidic actuators can be miniaturized, such as those requiring a large bending torque or substantial volume changes and deformation. There would be an argument for using a less stretchy strain-limiting layer with a single not fully inflated soft fluidic actuator to achieve the same low-profile aesthetics, but that is not ideal as the force is concentrated at a single point and difficult to conform to a curved substrate especially for wearable systems. Exaggerated examples are shown in Fig. 5 for illustrative purposes.

The increase in the number (i.e., arrays or matrices) of mini/micro soft fluidic actuators offers more design possibilities especially in the creation of the next-generation prosthetics, orthotics, haptics, and surgical tools; however, it may complicate the fabrication process and increase the amount of tubing required as each mini/micro soft fluidic actuator is independently or semi-independently actuated. This would require extensive automation going forward. A new strategy for control could also be envisioned where instead of controlling soft fluidic actuators by analog sensing of pressure, bending, and forces, soft robots could be controlled by merely counting which actuator is inflated and which ones are not. This philosophy could be considered the digitization of the analog soft fluidic actuators via miniaturization.

The feasibility to incorporate an array of mini/micro soft fluidic actuators that are individually controlled is best evaluated by examining the possibilities to miniaturize the backend (control box), which involves mainly a DC motor driving a peristaltic pump or a piston-diaphragm pump and a solenoid valve or equivalent to turn the flow on or off. Linearly increasing the number of pumps and valves to match the number of actuators will mitigate the benefits of miniaturizing the actuators as off-the-shelf fluid pumps and valves, regardless of pneumatics or hydraulics, are bulky and inefficient. More recently, solid state stretchy pumps based on the theories of electrohydrodynamics emerge;⁶⁰ for this type of pumps, miniaturization is not a problem as they are inherently small and flexible. However, safety issues of using high voltages must be addressed for practical implementations. Nevertheless, electrohydrodynamic pumps have the potential to become an essential component for liquid-based soft robots requiring relatively low flow rate and pressure. Alternatively, if the number of pumps and valves can be kept very low by the adoption of microfluidics, the resulted soft robotic system would be truly wearable as the backend (control box) may be pocketed or in a fashion that it does not stick out too much, located more proximally, centrally on the human body, as close to the center of the gravity as possible.

All in all, the motivation of incorporating microfluidics into the creation of a soft robotic system are twofold: (1) allowing miniaturization and digitization of soft fluidic actuators and (2) miniaturizing the size and improving the functionality of the control system for soft robots, with the ultimate goal of creating a soft robot that the end-user wants. This section is better ended by directly quoting the statement from Wyss Institute on the development soft robotic glove for neuromuscular rehabilitation: “This holistic approach, [involvement of potential end-users in every step of the soft robotic glove's testing and development], ensures that technology development goes beyond achieving functionality to incorporate social and psychological design elements that promote translation and seamless adoption by its intended end-users.”⁵³

CONCERNING ACTUATION AND SENSING

The conception of an idea, whether completely novel or a better implementation of an existing solution, should involve both technology developers upstream and stakeholders downstream from the very beginning, perhaps before or after the creation of the first proof of concept. When conceptualizing soft robotics, the first thing that comes to mind for technological developments is undoubtedly actuation for which a series of questions needs to be answered, which technique to choose, how to control it, at what cost? On top of these, how much miniaturization is appropriate and how to achieve it are valid concerns as well.

As established above, miniaturization is a compromise between functionality and aesthetics if done purely to make the soft fluidic actuators and the control system hardware smaller or an afterthought for further development. However, when miniaturization is considered a part of the strategy for social acceptance of soft wearable robotics, it morphs from the act of miniaturizing existing soft fluidic actuators and its control system into the development of formfitting a soft fluidic actuator system for socially acceptable soft wearable robots. We think these two are inherently different and worth a ponder; it changes the perspective on the motive, and we will elaborate further in the following sections. For now, actuation and technical implementation is the topic of discussion.

Power source for actuation

Soft wearable robots are applications driven; therefore, the actuation system is unique for different applications, although they do share certain similarities. From a schematic viewpoint, soft wearable robots can be divided into two major components, which can be coined as the frontend and the backend, respectively (Fig. 6). The frontend contains soft fluidic actuators and feedback sensors as mentioned in the previous section, and the backend contains a control system, which itself is made up of hardware and software including but not limited to valves, pumps, control sensors, microcontrollers, and batteries. The frontend is connected to the backend and indeed communicates via fluidic and signal pathways. Ideally, miniaturizing the frontend or the backend or both should not compromise the versatility of soft wearable robots, which might seem counterintuitive but can be done. In the next section, we will introduce some example applications that benefit from the strategies presented here.

FIG. 6.

Schematic representation of the frontend and backend of a generic soft wearable robot. Please note that the images of the components are not to scale and only serve as a generic visual representation of each component.

As a starting point, the fluidic actuation power source must be optimized for both size and efficiency, and the optimization is application specific as there are various pneumatic/hydraulic energy sources. Wehner et al.⁸ explored and quantified the pros and cons of various pneumatic energy sources for soft wearable robots. They also identified key metrics for measuring soft robotic power systems, which are energy density (joules/gram), flow capacity (liters/gram) normalized by the mass of the entire fuel system. Please refer to Table I for a fuel system-level summary. Compressed air cylinders can provide high pressures and flow rates but deplete rather quickly resulting in short operating times. Combustion using methane and butane provides too much pressure for most applications, not to mention the safety aspect of using combustible fuel. Onal et al.⁶¹ explored the idea of generating pneumatic pressure via hydrogen peroxide (H₂O₂) decomposition driven by the advantages such as silent operation, bypassing electric energy conversions, and the possibility of embedding the chemical pump inside elastomer substrates. However, the limitations of this system such as the temperature increase in the chemical pump and an exhaustive system-level development make it impractical. In the end, battery-powered microcompressors along with a good fluid recirculation design as well as a reservoir holding pressure to compensate for the relatively small flow rate produced by the microcompressors might offer an excellent solution for the pneumatic power system used onboard soft wearable robots.

TABLE I.

Fuel system-level summary. Reproduced with permission from Wehner et al., Soft Robot. 1(4), 263–274 (2014). Copyright 2014 Mary Ann Liebert, Inc.

	Fuel system mass as evaluated (g)	Maximum pressure (MPa)	Maximum flow rate (SLM)	Fuel system energy density (J/g)	Efficiency as evaluated (%)	Fuel system flow capacity (L/g)
Microcompressor small	130–171	0.2–0.34	1.7–3.5	72.6–99.3	36.7–39.9	0.78–1.38
Microcompressor large	422–756	0.26–0.55	5.7–10.7	14.8–33.6	31.2–40.3	0.17–0.37
Liquid CO2 small	49.9–662	5.6	18–127	30.4–44.1	<49	0.074–0.11
Liquid CO2 (refillable)	1573	5.6	300	58.1	<49	0.14
Air cylinder (refillable)	909–1975	20 and 30	350–650	27.1–83.6	<39	0.05–0.14
Butane combustion	62.7	n.a.	n.a.	2.37	0.5	0.094
Hydrogen peroxide decomposition	357	0.0507	n.a.	10–28	3.3–9, 45	0.04–0.05

Flow control for actuation

Microvalves have seen a tremendous increase in research and development over the first decade of the 21st century as summarized by a few excellent review articles,^62,63 which is mainly spurred on by the growth of microfluidics and lab-on-chip research. A very recent review article on microvalves⁶⁴ indicates that many of the same technical and fundamental challenges still remain. The choice of microvalves for soft robots is the question to be answered. Passive microvalves are merely miniaturized versions of large pneumatic/hydraulic check valves that function like an electrical diode allowing flow in one single direction. Active microvalves can be mechanical or non-mechanical. Most active mechanical microvalves are actuated by a flexible membrane via magnetic, electric, piezoelectric, and thermal actuation, with the most common and commercially available being miniature solenoid valves⁶⁵ and miniature piezoelectric valves.^66,67 Active non-mechanical microvalves are typically actuated via an electrochemical reaction, a phase change, and the use of rheological materials or ferrofluids. Another type of microvalve is called an in-line microvalve using external pneumatic pressure actuation,^24,68 which has robustness and practicality. In-line microvalves are inherently soft and can be directly integrated within soft materials providing outstanding seal and response time. Their downside is that their actuation requires external macroscale mechanical solenoid valves, which must be miniaturized and reduced in number to be used in portable devices such as soft wearable robots.

One of the most widely used in-line microvalves (Fig. 7) is so called Quake's valve.^24,68 The way it functions is straightforward. A pneumatic control line intersects a fluidic line separated by a monolithic membrane, meaning that all the materials for the entire structure of microvalves are the same. This alleviates the potential of interlayer delamination due to adhesion and thermal failures. The monolithic membrane is typically made out of polydimethylsiloxane (PDMS), which is a soft material with Young's modulus ∼750 kPa, hence allowing large deflection with low pneumatic pressure. When the pneumatic control line is pressurized, the monolithic membrane deflects and closes the fluidic line via pinching with a millisecond response time. The pressure required to close a Quake's valve is simply the minimum pressure needed to close the monolithic membrane plus the back pressure inside the fluidic channel, which could be below 50 mbar.²⁴ By crossing a single control line over multiple fluidic channels and varying the width of the monolithic membrane, it is possible for a single control line to independently control each fluidic line based on the level of pneumatic pressure inside the control line, allowing the potential for multiplexing, which reduces the number of external solenoid valves.

FIG. 7.

A 3D scale diagram of the pneumatic in-line microvalve developed by Unger et al., often referred to as Quake's valve. Reproduced with permission from Unger et al., Science 288(5463), 113–116 (2000). Copyright 2000 The American Association for the Advancement of Science.

The challenge for Quake's valve to be used for soft wearable robots is its need for an external solenoid valve for each in-line valve and the associated bulky control system.⁶⁹ However, if one macro-sized solenoid valve or an equivalent can control multiple Quake-style valves, a concept of multiplexing, then it would be a game-changer regarding the number of fluidic elastomer actuators that can be integrated into fabric or reducing the number and size of large fluidic elastomer actuators using a large number of much smaller independently controlled fluidic elastomer actuators. Earlier work on multiplexing via Quake's valves has been demonstrated by his group⁷⁰ where 3574 microvalves were independently actuated by using 22 external control solenoid valves, forming 1000 independent compartments as a microfluidic memory storage device. More work on multiplexing via Quake's valves followed later.⁷¹ Although these systems can help the development of soft wearable robots to achieve better and more efficient control and shrink the size of the control block, the complexity hinders their practicality.

Now, we arrive at the latest development of combinational microfluidic multiplexers. Lee et al.⁷² developed a microfluidic multiplexer that can control 19 independent fluidic channels using only 4 control lines and 8953 channels using only 10 control lines! The previous system can only address 16 independent fluidic channels using 4 control lines. Also, as the number of control lines increases, the capability of a microfluidic multiplexer developed by Lee et al. increases exponentially and outpaces any other microfluidic multiplexer. Therefore, the multiplexors can decrease the number of external proportional solenoid valves required, but in real applications, demultiplexers and additional proportional solenoid valves are also needed, which halves the efficiency, meaning that to truly control 19 independent fluidic elastomer actuators, a minimum of 8 external proportional solenoid valves are required. Also, all of the previously mentioned fluidic multiplexers are explicitly made with microfluidic applications in mind and fabricated via multilayer soft lithography,²⁴ which might not be appropriate for soft wearable robots. The control channel pressure must also be further reduced from existing microfluidic multiplexers to increase the pneumatic/hydraulic power source efficiency. The elastic monolithic membrane might also need to be upgraded for a higher number of cycles before failure. Most importantly, an alternative must be found to replace proportional solenoid valves to control the pressure within the control channels to be more efficient and less bulky.

As mentioned before, no single microvalve technology is suitable for all applications. Although the so-called Quake's valve offers much versatility in the form of independent control of a large number of digital on–off soft fluidic actuators, it still requires albeit small but rigid proportional solenoid valves. For certain soft robots requiring mostly dynamic actuation, valves might not be required, which is not to say that no valves exist in the backend, but rather, no valves are required to individually activate each soft fluidic actuator.

Flow regimes and significance for soft robots

Given that Hagen–Poiseuille law and Ohm's law are analogous, the equivalent fluidic resistors and capacitors can be identified within a fluidic system. Also, various electric circuit theories can be applied to microfluidic networks that help simplify the design phase of microfluidic chips.²² In a microfluidic network, the channels with hydraulic resistance become fluidic resistors, the compliant channels or soft chambers become fluidic capacitors, flow rate becomes current, pumps that deliver current and pressure become batteries, and pressure drops due to hydraulic resistances become voltage drops. Kandlikar and Grande⁷³ defined microchannels as having a characteristic length from 1 μm to 200 μm and minichannels from 200 μm to 3 mm; this definition was based on the rarefaction effect observed in gas as described by the dimensionless Knudsen number, where above 200 μm, rarefaction effect is negligible. However, for simplicity, we will consider any channel below 1 mm and above 999 nm as microchannels.

Usage of hydraulic has certain disadvantages, such as the fact that it requires a reservoir, which increases the bulkiness and weight of the actuation system. Therefore, pneumatics is preferable. It is well known that air is a compressible fluid; in the strictest sense, compressible fluid means that the density of the said fluid changes with changing volume given the same mass of the said fluid. Incompressible fluid in contrast has a constant density under all conditions. However, there is no true incompressible fluid, and even a liquid is slightly compressible under high pressures. Hence, an assumption can be made that any fluid that does not change too much in density during laminar flow with Mach number less than 0.3 can be considered an incompressible fluid. Mach number compares the speed of the fluid to the speed of sound in the same media. The flow and pressure change for air within microchannels, which will be coined as air microfluidics, precisely satisfies the above criteria to justify the assumption of air as an incompressible fluid. The errors associated with the above assumption are tolerable, and the equivalent hydraulic resistance is used as a design principle and engineering estimation instead of the fundamental theories. Hence, air microfluidic systems can be modeled using equivalent hydraulic resistance and electrical circuit analogy. The reason that it is named equivalent hydraulic resistance is because it is not a physical property of air, but a design parameter for air microfluidics.

The additive laws of fluidic resistors in series and parallel apply to microfluidic networks; hence, passive microfluidic elements can be very powerful in controlling the sequential pressurization of soft fluidic actuators, reducing energy-consuming active valves, and simplifying control of soft wearable robots. It must be noted that the additive laws of fluidic resistors in series and parallel are only valid for the applications where Reynold's number approaches to zero and for long narrow channels far apart.²² However, the additive laws can be used as a useful tool for the design of most air microfluidic chips. More accurate pressure drops and flow rates can be determined through semi-analytical solutions, numerical simulations, and experimental measurements.

Bruus²² consolidated various hydraulics resistances for straight channels with different cross-sectional area shapes. Equations (1)–(3) show the hydraulic resistances of three most common channel cross-sectional shapes used in microfluidics, where $a(\mathrm{m})$ is the radius of the circle for Eq. (1); $h(\mathrm{m})$ and $w(\mathrm{m})$ are, respectively, the height and width of the rectangle for Eq. (2); $w(\mathrm{m})$ is the side length of the square for Eq. (3); $\mu (\mathrm{kg}/\mathrm{m}\mathrm{s})$ and $L(\mathrm{m})$ are the fluid dynamic viscosity and length of the fluidic channel for Eqs. (1)–(3),

$$\mathrm{Circle}:{\displaystyle \frac{8}{\pi }\cdot \mu \cdot L\cdot {\displaystyle \frac{1}{a^4},}}$$ (1)

$$\mathrm{Rectangle}:{\displaystyle \frac{12\cdot \mu \cdot L}{1-0.63\cdot \left({\displaystyle \frac{h}{w}}\right)}\cdot {\displaystyle \frac{1}{h^3\cdot w},}}$$ (2)

$$\mathrm{Square}:28.4\cdot \mu \cdot L\cdot {\displaystyle \frac{1}{w^4}.}$$ (3)

If a more accurate pressure drop or flow rate is desired, Hagen–Poiseuille's law can be modified for compressible fluids such as air by timing it with an extra correction factor that expresses the average pressure relative to the outlet pressure.⁷⁴ Equation (4) shows Hagen–Poiseuille's law for gas flow in a round pipe assuming constant viscosity and no-slip boundary condition, where $Q({\mathrm{m}}^3/\mathrm{s})$ is the volumetric flow rate; $R(\mathrm{m})$ and $L(\mathrm{m})$ are the radius and length of the channel, respectively; $\mu (\mathrm{kg}/\mathrm{m}\mathrm{s})$ is dynamic viscosity; and $P_i(\mathrm{MPa})$ and $P_o(\mathrm{MPa})$ are inlet pressure and outlet pressure, respectively,

$$Q={\displaystyle \frac{\pi \cdot R^4}{16\cdot \mu \cdot L}\cdot \left({\displaystyle \frac{P_i^2-P_o^2}{P_o}}\right).}$$ (4)

Most of the previous studies on gas flow in microchannels have focused on the slip-flow regime studying rarefaction effects and developing models. For instance, Roy et al.⁷⁵ developed a two-dimensional (2D) finite element based model for predicting microscale flow up to the transition regime for reasonably high Knudsen number flow inside microchannels and nanopores. At a Knudsen number of 0.058, a 4%–8% higher-streaming velocity is detected from the slip-flow results that there is a corresponding no-slip solution for the same pressure ratio; the increased velocity is due to lower-shear stress. Figure 8(a) shows the channel centerline pressure distribution, and Fig. 8(b) shows the centerline velocity solution. It can be seen that the pressure distribution between the slip solution condition and the no-slip solution differs very little (∼4% maximum difference).

FIG. 8.

(a) Centerline pressure distribution normalized with outlet pressure $P_{out}$. Solutions are compared for the slip and no-slip conditions for five different pressure ratios $(P_r=P_{in}/P_{out})$. (b) Centerline velocity distribution normalized with the speed of sound at the inlet $C_{in}$. Solutions are compared for the slip and the no-slip conditions for five different pressure ratios $(P_r=P_{in}/P_{out})$. Both (a) and (b) are reproduced with permission from Roy et al., J. Appl. Phys. 93(8), 4870–4879 (2003). Copyright 2003 AIP Publishing LLC.

Arkilic et al.⁷⁶ studied gas flow in a rigid and straight microchannel and developed a 2D approximate analytical solution based on 2D perturbation of a continuous flow model with first-order slip boundary conditions as shown by Eq. (5), where $\dot{m}(\mathrm{mg}/\mathrm{s})$ is the mass flow rate; H, W, and L are height, width, and length of the channel, respectively, all with a unit $(\mu \mathrm{m})$; $\mu (\mathrm{N}\mathrm{s}/{\mathrm{m}}^2)$, $R(\mathrm{J}/\mathrm{mg}\mathrm{K})$, and $T(\mathrm{K})$ are dynamic viscosity, gas constant, and temperature, respectively; $P_i(\mathrm{MPa})$ and $P_o(\mathrm{MPa})$ are inlet pressure and outlet pressure, respectively; $\sigma$ is the complete momentum accommodation coefficient, which is almost always 1 for engineering calculations; and $K_o$ is the outlet Knudsen number. The Knudsen number is 0.155, which means that the gas flow is within the transition flow regime. The gas flow occurred in long rigid silicon microchannels with nominal heights of 1.33 μm, a nominal width of 52 μm, and a length of 7500 μm. The result shows a substantial difference in the pressure distribution and the mass flow between the slip-flow solution and the no-slip solution and that the slip-flow solution matches experimental data almost perfectly [Fig. 9(a)],

$$\dot{m}={\displaystyle \frac{H^3\cdot W\cdot P_o^2}{24\cdot \mu \cdot L\cdot R\cdot T}\cdot \left[{\left({\displaystyle \frac{P_i}{P_o}}\right)}^2-1+12\cdot \sigma \cdot K_o\cdot \left({\displaystyle \frac{P_i}{P_o}-1}\right)\right].}$$ (5)

FIG. 9.

(a) Helium mass flow for a 1.33 μm channel (95% confidence intervals indicated). The solid curve is the solution to Eq. (5), assuming full tangential momentum accommodation, and the dashed curve is the solution to Eq. (5) setting Knudsen number to zero (no-slip solution). Reproduced with permission from Arkilic et al., J. Microelectromech. Syst. 6(2), 167–178 (1997). Copyright 1997 IEEE. (b) The comparison of pressure drop with experimental results and 2D analytical solutions [Eq. (5)]. Reproduced with permission from Hsieh et al., Int. J. Heat Mass Transf. 47(17–18), 3877–3887 (2004). Copyright 2004 Elsevier.

Hsieh et al.⁷⁷ verified the 2D analytical solution by Arkilic et al. through looking at the slip regime, where the Knudsen number ranged from 0.001 to 0.02. The channel is 24 mm long with a width of 200 μm and a height of 50 μm. Figure 9(b) shows the pressure drop as determined experimentally and by 2D analytical solutions. Although this 2D analytical solution [Eq. (5)] ignores the entrance and exit losses, they are the basis for any gas flow in long microchannels where the entrance and exit losses are small compared to shear stress and gas acceleration; hence, inlet and outlet effects could be ignored.

There are other analytical and semi-analytical models reported for simple cross sections, and most employ the first-order slip boundary condition, which has shown to be relatively accurate up to a Knudsen number of 0.1. Hence, it would be sufficient to use Eq. (5) for accurate calculations of pressure drops and flow rates for air microfluidic systems.⁸² Other models such as Navier–Stokes equations with a second-order slip boundary condition, Burnett equations with a slip boundary condition, direct simulation Monte Carlo, and lattice Boltzmann are primarily used for a much larger Knudsen number in the transition flow and free molecular flow regimes suitable for nanoscale gas flows and hence are outside the scope of this perspective, which focuses on discussing microscale systems.⁸³

Many practical applications of microchannels also incorporate bends, sudden contraction or expansion, T-junctions, and Y-junctions. Agrawal⁸⁴ reviewed various experimental and numerical investigations. The conclusion of this review suggests that flow separation occurs at the bend and sudden contraction or expansion for laminar gas flow in microchannels resulting in a higher pressure drop than straight microchannels. The exact results of these experiments and numerical simulations are not crucial for synergizing microfluidics in soft robotic applications; instead, these minor losses should at least be accounted for within the designs of the air microfluidic systems.

As for applications for soft robots, Vasios et al.⁷⁸ used viscous flow to simplify the actuation of fluidic soft robots where narrow tubes are used to achieve sequencing in the actuation of soft fluidic actuators [Fig. 10(a)]. They also developed modeling and optimization tools to identify optimal tube characteristics for achieving a complex target response of soft fluidic actuators. Preston et al.⁷⁹ developed a soft ring oscillator, which eliminates the need for hard valves and electronic controls [Fig. 10(b)]. The soft ring oscillator allows periodically oscillating outputs, which can be used in applications including undulating and rolling motions in soft robots, size-based particle separation, pneumatic mechanotherapy, and metering of fluids. Similarly, Gorissen et al.⁸⁰ developed hardware sequencing of inflatable nonlinear actuators for autonomous soft robots [Fig. 10(c)], which uses passive flow restrictions without the need for complex on-board valves or bulky tethers. Ben-Haim et al.⁸¹ exploits bistability of equilibrium states for hyperelastic thin shells, as well as pressure differences induced by viscous effects for reducing the number of fluidic inputs [Fig. 10(d)]. The theme here is using the physics of microfluidics combined with elastic or hyperelastic materials to simplify the control algorithm and control hardware. These are significant first steps and proof of concept that show the potential of synergizing microfluidics with soft robots. Still, much more work needs to be done, especially interdisciplinary collaborations, to apply these fundamental research studies to real-world applications.

FIG. 10.

(a) Viscous flow in the tubes interconnecting the constituent actuators to design soft robots capable of achieving a variety of responses when inflated with a single pressure input. Reproduced with permission from Vasios et al., Soft Robot. 7(1), 1–9 (2020). Copyright 2020 Mary Ann Liebert, Inc. (b) A soft ring oscillator composed of three inverters connected in a loop, resulting in a systematic instability (because no stable state exists). Reproduced with permission from Preston et al., Sci. Robot. 4(31), 1–9 (2019). Copyright 2019 The American Association for the Advancement of Science. (c) By tuning the nonlinear properties of the actuators with intermediate flow restrictions, a single fluidic supply tube can be used without the need for internal valves to actuate a soft robot. Reproduced with permission from Gorissen et al., Adv. Mater. 31(3), 1–7 (2019). Copyright 2019 John Wiley & Sons. (d) Schematic picture of a serial chain of elastic chambers with a single inlet. Reproduced with permission from Ben-Haim et al., Soft Robot. 7(2), 259–265 (2020). Copyright 2020 Mary Ann Liebert, Inc.

Microfluidics and interfacing with soft robots

When thinking about microfluidics, perhaps the first thing that comes to mind is a microfluidic chip, which for many lab-on-chip applications serves as a platform for passive⁸⁵ or active^23,24 control and manipulation of fluid carrying reagents and biological samples in the microscale and even the nanoscale. Also, a microfluidic chip is the physical manifestation of the equivalent hydraulic resistance and capacitance as well as the active control strategies such as Quake's valves. A microfluidic chip that can be used for manipulating fluid to inflate and deflate soft fluidic actuators may be the missing piece for soft robotic applications that require a significantly higher number of soft fluidic actuators (i.e., digitization) and a miniaturized control box to achieve low-profile aesthetics. Although the focus is on using air as the pressurization fluid, liquid may also be substituted for various soft robotic applications.

The compressible characteristic of air and the equivalent hydraulic resistance can be used as tools to design dynamic sequencing for soft fluidic actuators. For example, parallel microchannels below 500 μm may be used as resistors to enable sequential inflation of a series of mini soft fluidic actuators based on the concept of the paths of least resistance (Fig. 11). Elastic inflatable chambers between microchannels may be used as capacitors to store pressure energy. From a fundamental theory perspective, these tools are simple, but practically useful control strategies for a particular soft robot, which requires good microchannel network design that can be proposed and optimized through both simulation and experimental approaches. At the present day, there are no guidelines or simple semi-analytical models to design the microchannel networks to achieve this; albeit, computational fluid mechanics and finite element simulations can provide useful predictions at the expense of cost and time. Very complex pneumatic digital logic can also be built using these tools, as demonstrated by Ref. 86. However, the usefulness of pneumatic digital logic is still unclear for soft robotics. If dynamic sequencing is not desirable, microfluidic chip may be used as a wearable manifold providing a fluidic pathway from the pump to the soft fluidic actuators to minimize the usage of tubing and fittings as seen in the soft robotic knee brace for managing knee osteoarthritis by Gao et al.⁸⁷ Of course, all of this is predicated on having numerous mini/micro soft fluidic actuators.

FIG. 11.

(a) Simulation demonstrating the concept of the path of least resistance with parallel microchannels. (b) The exit velocities of the microchannels shown in (a).

To convert this idea into a useful and practical device is complex and currently faces a few challenges. The first of which is how mini/micro soft fluidic actuators and strain-limiting layers can be produced in a reliable, scalable, and efficient manner. As mentioned before, larger-scale soft fluidic actuators are commonly produced via molding, pouring silicone into a 3D printed mold, and then gluing two pieces of silicone together using uncured silicone. This process may have difficulties at a smaller physical scale as the uncured silicone may clog the inlet channels, and the walls may be too thick or unevenly distributed. Skylar-Scott et al.⁸⁸ recently introduced a fused deposition modeling (FDM) 3D printing technique whereby polymers with different modulus may be printed at the same time from a specially designed nozzle [Fig. 12(a)]. When a higher printing resolution is achieved, this technology has the potential to integrate microfluidic chips within soft robots as one piece. Wirekoh and Park⁸⁹ have developed artificial muscles that are flat when uninflated and produce uniaxial tensile stress when inflated as a replacement for bulky McKibben muscles incorporating modulus modifying materials (notably Kevlar strings) into its walls as the strain-limiting mechanism. Another technique is to use soft lithography or one of its variations, such as those introduced by Taylor et al.⁹⁰ where they created milliscale soft robots via ultraviolet ozone bonding of PDMS and Mylar transparencies that can withstand internal pressure exceeding 1000 kPa [Fig. 12(c)].

FIG. 12.

(a) Multimaterial multinozzle 3D printheads. Reproduced with permission from Skylar-Scott et al., Nature 575(7782), 330–335 (2019). Copyright 2019 Springer Nature. (b) Fabrication of milliscale soft grippers via ultraviolet ozone bonding of PDMS and Mylar sheets. Reproduced with permission from Taylor et al., Adv. Mater. 30(7), 1–8 (2018). Copyright 2018 John Wiley & Sons.

Connolly et al.⁵¹ introduced a method to produce soft fluidic actuators, in particular, bending actuators via heat press and taking advantage of the anisotropic property of textiles, which when combined with an elastic or rigid plastic film produces a strain-limiting layer [Fig. 3(b)]. This approach presents a few advantages. First off, the strain-limiting layer can be tailored to produce different desired motions, forces, or torques. Second, the elastic and rigid plastic film and the fabric are very thin, enabling low-profile soft wearable robots. On the other hand, the process can be tedious and labor intensive, especially during the washing of the water-soluble polyvinyl alcohol (PVA) film that forms the soft fluidic actuator chamber and gluing the tubing into the inlet of the soft fluidic actuator. This is perhaps acceptable for making larger soft fluidic actuators as presented in Ref. 51 and one-off prototypes, but it is not reliable for making a large quantity of mini or micro soft fluidic actuators. The correct fabrication process depends on the application, but the future of producing mini and micro soft fluidic actuators and strain-limiting layers should involve digital manufacturing and lots of automation to ensure quality, scalability, quick design iterations, and low cost.

The second challenge is related to how a microfluidic chip should be made. Traditionally, microfluidic chips are fabricated out of glass- or silicon-based materials using photolithography technology or PDMS using soft lithography technology. The latter can be directly translated into making microfluidic chips for soft robotic applications as the softness of PDMS can be tuned readily. With the improvement in 3D printing technology, instead of making the molds via photolithography, molds can be 3D printed to achieve microchannels with a cross-sectional characteristic length of 750 μm or smaller. One lingering issue is concerning how to reliably bond two or more pieces of PDMS extracted from the 3D printed mold because the resulted surfaces are less smooth than those from the molds made via photolithography. Also, methods to integrate microfluidic chips with soft robots are worth investigating. Although they can adhere to the soft robot mechanically or chemically, the ideal way is perhaps to incorporate microfluidic systems directly into the material of soft robots considering the goal of miniaturizing the actuation system as much as possible. This is still a practical challenge as commercial 3D printers are yet to achieve the resolution, material, and process required for reliable and cost-effective printing of internal microchannel networks, but increased commercial and research efforts in the 3D printing field make this very promising.⁹¹

Third, tubing for soft wearable robots that potentially involve the use of a large number of mini/micro soft fluidic actuators can become messy and unmanageable very quickly. The tubing should be integrated onto fabric as an extra layer or replaced altogether by fluidic channels as part of the composite fabric. In addition, simple ways of tailoring the modulus of different parts of a piece of fabric are needed, and this is perhaps in the form of elastic or plastic membranes that can be heat pressed onto the fabric selectively or built into the fabric manufacturing process.

Fourth, the control box or the backend should be easily pocketable or integrated as part of the clothing. The control box should be as flat and formfitting as possible, well incorporating all the necessary hardware. The control box ideally has a similar size to that of modern-day smartphones, and the thinness would be difficult to emulate but could be achievable where only the pump is slightly thicker while keeping other areas thin [Fig. 13(a)]. The pump needs to be energy efficient. As mentioned prior, battery-powered microcompressors offer the best energy density (J/g) as shown in Table I. Among the microcompressors, the piston-diaphragm pump is currently the most widely available off-the-shelf option, often sold as aquarium pumps or fluid dispensing units for medical devices. These off-the-shelf pumps often have two sections: the motor portion and the piston-diaphragm portion. To make the control box more compact and low-profile, the pump can be integrated with the frame of the control box, where instead of having a separate piston-diaphragm portion [Fig. 13(a)], it can be integrated directly with the frame or the walls of the control box as compact as possible [Fig. 13(b)]. The frame of the control box also serves to dampen the noise and vibration from the motor. The piston-diaphragm assembly can be custom made via 3D printing and molding. Also, the motor that drives the piston-diaphragm could be reduced in size while retaining an appropriate torque and speed for a particular application by purchasing high-end coreless DC motors such as those offered by Faulhaber Micromo. In addition, multiple pumps can be connected in series and parallel to predominantly increase head pressure or flow rate, respectively, for the reason that a larger pump would increase the thickness of the control box, which is undesirable. Therefore, the focus is to reduce the thickness as much as possible. The empty spaces within the control boxes, as seen in Fig. 13, are reserved for other electronic components such as microcontrollers, pressure sensors, and batteries.

FIG. 13.

(a) Off-the-shelf piston-diaphragm pump integrated with a control box/backend. (b) Custom low-profile control box/backend by using smaller coreless DC motors with integrated piston-diaphragms in parallel or series.

Last, the sensors that measure pressure and strain as feedback for controlling actuators need to meet certain criteria based on the applications. In the context of synergizing microfluidics with soft robots, it can be broadly categorized into two sections: soft microfluidic sensors for soft robots and microfluidics-enabled soft robotic sensors. Soft microfluidic sensors are perhaps the first and to date, the most recognized usage for microfluidics with soft robotics. Researchers have been actively investigating microfluidic systems integrated with liquid metals such as gallium and Indium (EGaln)^92–94 for sensing purposes. Subsequently, a class of entirely soft sensors that can measure pressure and strain was developed for soft wearable robots. The purpose of these entirely soft strain sensors is to avoid rigid and hard components in soft wearable robots altogether. Park et al.⁹⁵ developed a type of hyperelastic pressure transducer by embedding microchannels of conductive liquid eutectic gallium–indium inside silicone rubber. Park et al.⁹⁶ subsequently improved the designs of the entirely soft sensor by switching from the soft lithography fabrication technique to a 3D-printing based multi-layer molding technique to create the microchannels that carry the conductive liquid. Also, the capability of the entirely soft sensor is expanded from just sensing pressure to also sensing strain. Figure 14(a) shows this entirely soft sensor. Further improvements to this class of entirely soft sensors were made to increase its robustness, eventually reaching results of up to 396% of strain and less than 2% change in electromechanical specifications through 1500 cycles of loading–unloading.⁹⁷ All of these improvements enable this type of liquid microfluidics-based strain sensor to monitor hip, knee, and ankle sagittal plane joint angles as a wearable sensor [Fig. 14(b)]. One major disadvantage regarding this type of sensor is the utilization of toxic conductive liquids, which can be hazardous when used as a wearable sensor.

FIG. 14.

(a) Fully soft strain sensor. Reproduced with permission from Park et al., IEEE Sens. J. 12(8), 2711–2718 (2012). Copyright 2012 IEEE. (b) Soft strain sensors placed at each lower limb joint to capture motion in the sagittal plane. Reproduced with permission from Mengüç et al., Int. J. Rob. Res. 33(14), 1748–1764 (2014). Copyright 2014 SAGE Publications. (c) Photos of the sensor illustrating stretching at approximately 0% and 100% strain. Reproduced with permission from Atalay et al., Adv. Mater. Technol. 2(9), 1–8 (2017). Copyright 2017 John Wiley & Sons.

The latest technical improvement of the all soft strain sensor resolved the safety issue regarding the potential leakage of the toxic conductive fluid inside microchannels by replacing it with a conductive knit fabric as electrode and a silicone elastomer as a dielectric.⁹⁸ The fabrication method has also seen improvements by introducing a batch manufacturing procedure involving laser cutting [Fig. 14(c)]. Entirely soft sensors are preferable for soft wearable robots; however, they are not always necessary as most wearable sensors are miniaturized. For instance, the 9-axis IMU (ICM-20948) by TDK has a package size of only 3 × 3 × 1 mm³, which can be easily encapsulated within soft silicone and integrated as part of the soft wearable robot. Also, EMG sensors have seen a tremendous reduction in size regarding the peripheral equipment required to process the signals, and the EMG electrodes have been developed to be reusable, washable, and integrated with fabric,⁹⁹ which is spearheaded by Athos, an athletic wearable technology company.

As for microfluidics-enabled soft robotic sensors, it derives from the idea of digitizing soft fluidic actuators via microfluidics. Basically, soft fluidic actuators that can be arranged in a tight array or stacked measure the force, torque, or compression acted upon them instead of acting as a force, torque, or compression provider. The advantage here is that these sensors are compliant and elastic, which could fit over a contour or irregular shapes and create a 3D topographical map of the measured parameters [Figs. 15(a) and 15(b)]. Soft organic sensors or small piezoresistive sensors could be placed or float within the digitized soft fluidic actuators or directly integrated with its material to act as pressure or strain sensors [Fig. 15(c)]. The microfluidic channels and tubing connected to these sensors act as a pressurization pathway as well as a signal pathway. For instance, an ultra-thin electrical wire could be embedded within the microchannels to carry analog electrical signals from the piezoresistive sensor within the digitized soft fluidic actuators. It must be noted that the examples shown in Fig. 15 are for illustrative purposes only and by no means cover any specific application.

FIG. 15.

Concept images showing digitized soft fluid actuator/microfluidics-enabled soft robotic sensors arranged in a tight array (a) for sensing compression and normal forces and (b) torque. (c) Concept image showing incorporating soft organic sensors or small piezoresistive sensors within a digitized soft fluidic actuator/soft microfluidic sensor.

All in all, this section presents a perspective on how the actuation of soft robots, especially soft wearable robots, is intertwined with miniaturization. Also, we explored how the same concept used on actuators can be used for sensors, which we coined as microfluidics-enabled soft robotic sensors. Microfluidics provides a convenient and direct avenue to allow the inclusion of numerous mini/micro soft fluidic actuators, independent control of said mini/micro soft actuators by different versions of Quake's valve, fast millisecond actuation of numerous mini soft fluidic actuators via microfluidic manifolds and integrated pathways, and sequential inflation patterns via passive microfluidic channel networks via the concept of equivalent hydraulic resistance and capacitance. Although not directly related to microfluidics in a technical sense, reducing the size of the piston-diaphragm pump while retaining an appropriate level of performance is vital for the miniaturization of the control box as the pump is probably the single largest component regarding thickness. The effort spent on research and development of actuation and sensing ultimately allows for the exploration of entirely new applications and improving existing systems through soft robots. This may lead to a new paradigm regarding wearable assistive technology in terms of making them more dynamic, functionally better, more compliant, and more comfortable. This also opens doors to wearable technologies, which have been mostly limited to sensing, such as haptic wearables that can be used for leisure activities for the elderly via augmented/virtual reality. Below is a compilation of potentially applicable research directions that can have a real impact on the world.

CONCERNING APPLICATIONS

The goal of this section is to provide a glimpse into different types of real-world applications that could benefit from the existing research and development, as mentioned in the previous two sections. The applications described below are intentionally presented in a broad and open-ended fashion as it serves as a guide, a vision, and a springboard for other novel ideas. Experts in each field hold their own opinion on whether microfluidics-enabled soft robots are genuinely impactful for a particular application within the said field; nonetheless, the potential is there.

Daily quality of life

It is a universally shared aspiration to lead a good life and age gracefully. However, life can be unpredictable, a health-related complication, diagnosis of a debilitating disease, slow progression of musculoskeletal disorders, or traumatic injuries often cause a decrease in quality of life. In this fight of self-preservation, there exist many devices that help us heal from acute injuries and manage chronic diseases; we envision this is where soft robots, especially the wearable kind when synergistically combined with microfluidics, can provide innovation and new design paradigm to elevate existing solutions or produce completely new categories of useful products. Although microfluidics-enabled soft wearable robots do not ultimately eradicate the root causes of diseases, they can, however, improve an individual's quality of life and that alone is worth pursuing.

Chronic pain symptoms, impairment, and disability^100,101 are some of the main reasons for a lower quality of life. Chronic musculoskeletal conditions are injuries and pain involving joints, ligaments, muscles, nerves, tendons, and structures that support limbs, neck, and back. They are one of the leading contributors to disability worldwide¹⁰² and are often intertwined with depression¹⁰³ in the form of co-morbidity. Therefore, assistive devices that can manage the chronic pain symptoms and effectively improve the physical function and mobility of their users could improve the users' quality of life tremendously as they also help improve mental health. For chronic conditions, an assistive device should ideally be worn daily; thus, ease of use and aesthetics become a focus just as much as functionality. Depending on applications, the primary functionality changes, but the secondary goal should always be to reduce pain symptoms.

The digitized mini/micro soft fluidic actuators along with strain-limiting layers can be perhaps used for targeted zonal compression and damping, which can be useful as external biomechanical torques for better range of motion, bracing for stability, enabling dynamic proprioception and massages, and improving various physiological/neuromuscular responses. When the digitized micro/mini soft fluidic actuators are deflated, they are flat and just part of clothing and become functional when inflated. Ideally, tight-fitting apparel should be used as the base for soft wearable robots as they hug the body contour enabling the soft fluidic actuators to become functional with less volume expansion, therefore enhancing the efficiency of the soft fluidic actuators. For more in-depth details on the biomedical applications of soft robotics, readers are directed to a comment on the human-in-the-loop development of soft wearable robots¹¹ by Walsh and an excellent review by Cianchetti et al.⁴

The current commercially available assistive devices for managing musculoskeletal conditions are braces, sleeves, kinesiology tapes, and orthotics; these devices can range from $20 off-the-shelf sleeves to customized functional braces in the range of thousands of dollars. Microfluidics-enabled soft wearable robots could improve these devices by adding dynamic functionality without introducing any bulk to the limbs as digitized soft fluidic actuators are miniaturized. Only a tolerable amount of bulk in the form of a low-profile control box for fluidic control and transportation that can be pocketed or integrated with fabric is introduced. As mentioned earlier, most of the cost comes down to the control box with the major expenses on the pump, microcontroller, and battery as the expenses on the electronic valves and pressure sensors are negligible. It is unlikely to compete on cost alone with a $20 sleeve but instead offers a significant amount of performance boost via dynamic functional improvements that need to be thoroughly and independently verified, which would most likely make it a $200 solution competing with custom functional bracing and orthotic solutions. Although microfluidics-enabled soft wearable robots present tremendous potential as active assistive devices, they are not meant to replace existing sleeves and braces, but rather complementing the existing lineup of assistive devices for the betterment of patients.

Prosthetic applications

As mentioned before, one major advantage of using numerous mini/micro soft fluidic actuators is that when they are arranged in a close matrix, they can mimic an evenly distributed area compression force when combined with strain-limiting layers without the need of a large volume that fits over an uneven contour. In other words, by decreasing the size and increasing the number of soft fluidic actuators, the resolution of the treatment area and the ability to control pressure at specific locations dramatically improve. This meant that a 3D entity entirely can be made out of these mini/micro soft fluidic actuators controlled by one single microfluidic chip. These actuators could change in shape, volume, and stiffness that could fit on top of the body contours or fill up spaces within an irregular cavity. For example, prosthetic socket fit and comfort has always been an issue,¹⁰⁸ especially for diabetic patients or users experiencing co-morbidities; this is perhaps where the technology presented herein could provide an additional solution or an add-on to existing prosthetic socket fit strategies.¹⁰⁹ Instead of replacing current prosthetic socket suspension methods, namely, vacuum suction, the two can be combined to provide better socket fit and comfort.

Pneumatic or even hydraulic actuated prosthetic socket (Fig. 16) is not new, as seen in the review by Paterno et al.,¹¹⁰ but the question of its low adoption remains valid. There are many potential answers here, but one interesting answer is perhaps lacking tailored, varying and evenly distributed area compression forces, which make it difficult to fit an uneven contour requiring different compression levels. Regardless, thorough clinical studies would be needed to determine the usefulness of future pneumatic prosthetic sockets and benchmark them against the current best strategies. In a similar vein, cushions, in general, could be modified to produce dynamic fit, comfort, and massaging using microfluidics-enabled soft fluidic actuators. Whether this technology is appropriate for the application comes down to the cost, reliability, user acceptance, and competitiveness to other emerging solutions.

FIG. 16.

(a) Pneumatic inflatable inserts used in limb prosthetics, including the Air Contact System, PNEU-FIT, Pump It Up!, and Pneumatic Volume Control System. Reproduced with permission from Sanders and Cassisi, J. Rehabil. Res. Dev. 38(4), 365–374 (2001). Copyright 2001 Department of Veterans Affairs.¹⁰⁴ (b) Pneumatic actuator inserts for interface pressure mapping and fit improvement in lower extremity prosthetics. Reproduced with permission from Carrigan et al., “Pneumatic actuator inserts for interface pressure mapping and fit improvement in lower extremity prosthetics,” in Proceedings of the IEEE/RAS-EMBS International Conference on Biomedical Robotics and Biomechatronics (IEEE, 2016), pp. 574–579. Copyright 2016 IEEE.¹⁰⁵ (c) Rendering of smart variable geometry socket (SVGS) technology applied to a transfemoral suction socket. Reproduced with permission from Greenwald et al., J. Prosthet. Orthot. 15(3), 107–112 (2003). Copyright 2003 Wolters Kluwer Health, Inc. (d) Socket insert having a bladder system. Adapted from Phillips, U.S. patent 7,655,049 B2 WO 2004/071337 A2 (August 26, 2004).¹⁰⁷

Sports and safety apparels

Sports and safety apparels (including shoes) present a big commercial market. Miniaturization is perhaps the most important here as the bulk of the control box must be minimized so that it does not interfere with the movement of the person during sports, active leisure activities such as hiking, or working in warehouses or construction sites. This application mostly involves prophylactically minimizing the risks of injuries to muscles and joints, improving protection from helmets and pads, and postural-alignment haptic feedback. Again, the question comes down to whether this technology of using microfluidics-enabled mini/micro soft fluidic actuators would be over-engineering existing sports and safety apparels and whether it is appropriate for specific applications.

Haptics

Haptics is a significant sector where microfluidics-enabled mini/micro soft fluidic actuators could become very useful. HaptX,⁴³ a company working on virtual reality exoskeletons, was the first to adopt this technology; they have coined their technology as “microfluidic skin,” which incorporates numerous silicone-based micro pneumatic bladders on a panel connected to microfluidic channels for fluidic transport [Fig. 17(a)]. In theory, when the micro pneumatic bladders are made sufficiently small and close in an array pattern, users' perception no longer treats them as individual tactile forces but rather a smooth area force. One major challenge with this type of haptic technology is how to create a wearable control box. Since each micro pneumatic bladder requires independent control, a microfluidic valve or equivalent could be used to miniaturize the control box drastically. Alternatively, electromechanical microvalves built into a microfluidic chip could also replace larger solenoid valves. Perhaps the biggest technological challenge here is achieving fast response time of this type of a wearable control box for haptic applications requiring simultaneous independent control of multiple soft fluidic actuators.

FIG. 17.

(a) HaptX microfluidic skin. Adapted from HaptX, “Industrial-grade haptic technology” (2020). Copyright 2020 HaptX Inc. (b) The feedback control loop of the soft haptic system. The SPA-skin provides a highly conformal interface. A strain sensor of stretchable metallization can be used as the input to a feedback loop used to control the actuator inflation and exerted force. High-speed data acquisition allows strain sensing and actuation over a range of actuation frequencies from 0 to 100 Hz. Reproduced with permission from Sonar et al., Soft Robot. 7(1), 22–29. Copyright 2020 Mary Ann Liebert, Inc.

It must be mentioned that HaptX, as a company, naturally holds patents¹¹¹ and patents pending on this technology. Academia's role would be to tackle the fundamental grand challenges regarding better materials, control strategies, and miniaturization. For instance, Sonar et al.¹¹² presented a soft pneumatic actuator skin [Fig. 17(b)] incorporating self-sensing that is capable of high-frequency vibrations (∼100 Hz). Last, perhaps the mini/micro soft fluidic actuators could be further miniaturized and placed together or even overlap with each other to improve the sensation further when touching a smooth surface; perhaps the micro-balloon actuators could be so small that they are considered a digital on/off point instead of having to control pressure. Alternatively, magnetorheological or electrorheological fluids could replace the pneumatic system altogether if safety and response time can at least be on par with pneumatic systems.

Pediatrics

Certain versions of microfluidics-enabled mini/micro soft fluidic actuators could also be used for pediatric applications and nursing applications for children and toddlers. The functionality of mini/micro soft fluidic actuators is applications dependent, but generally speaking, they are meant for comfort, assurance, a warm embrace, and many usages already discussed above.

All of the applications mentioned above share a commonality where the actuation timing is essential, whether using a passive microfluidic chip, a microfluidic wearable manifold, Quake's valve system, or indeed, any type of microfluidic control system, specific actuation response time must be met. For instance, a system meant to sync with human gait, the actuation must be fast on the millisecond scale. Similarly, a system that is unpredictable, such as a haptic system, must minimize the delay between when the signal is received by a valve and the mini/micro soft fluidic actuator or a massage system requiring predefined sequential inflation and deflation to mimic percussive massage. All in all, activities of daily living such as those described above present some high impact applications for microfluidics-enabled soft robots where miniaturization is perhaps key for user adoption, but for more specialized purposes, what microfluidics brings regarding functionality plays a considerable role.

Medical instrumentation (minimum invasive surgery)

When it comes to medical equipment, one of the first things is surgical instruments, in particular, open surgery instruments, due to the conformable nature of soft robotics. Spurred on by recent innovation in automation and robotics, minimum invasive surgery (MIS) is becoming a widespread reality in terms of the number of hospitals capable of delivering MIS but also the type of surgery possible. One of the initial goals of soft robots in MIS is to improve endoscopes by creating soft and compliant systems that mimic the motion of a worm, snake, or octopus's tentacles. For a more in-depth review of current soft robots in the research phase for MIS and future directions, readers are referred to an excellent review by Runciman et al.¹¹³

Microfluidics allows numerous mini or micro soft fluidic actuators to be fitted onto a soft robot to control locomotion, such as forward propulsion or bending. The idea of digitizing mini or soft fluidic actuators could enable tethered soft robots in the form of an earthworm that could navigate unstructured and arduous paths to perform MIS. In other words, a microfluidics and soft robotics enabled next-generation endoscope that is smaller in size and more functional could be envisioned and created. Ranzani et al.¹¹⁴ demonstrated the capability of microfluidics with micro-balloon actuators for bending by creating a 12-layer soft peacock spider via multilayer soft lithography (Fig. 18). There is plenty of space in a surgical operation room; therefore, there is no need to miniaturize the control box. A precision fluid control system such as those from Fluigent, Elveflow, or open-source modular system such as the μPump¹¹⁵ can be used to control the microfluidics-enabled soft robot for MIS applications. The mini/micro soft fluidic actuators could produce locomotion in the form of peristaltic crawling and bending as shown by the peacock spider, at the same time bringing uninflated bladders/catheter to produce room for surgery and diagnosis via an onboard camera. Similar to existing endoscopes, a channel for surgical tools could be incorporated. Pneumatic actuation would most likely need to be replaced with hydraulic actuation in order to ensure that air bubbles do not get trapped inside the human body due to leakage, and hopefully, low pressure actuation could be developed to ensure that an accidental burst of a mini/micro soft fluidic actuator does not cause damage to the patient.

FIG. 18.

Generation of the 3D structure through the injection-induced self-folding. Reproduced with permission from Ranzani et al., Adv. Mater. 30(38), 1–6 (2018). Copyright 2018 John Wiley & Sons.

Educational tools

Microfluidics-enabled soft robots can be used as biomedical education tools as well, where they can be used to mimic both the functionality and aesthetics of organs, tendons, ligaments, and muscles, which when used in conjunction to cadavers, textbooks, and images could improve the student's learning experience. For instance, multiple soft fluidic actuators, when combined in series, could mimic the motion of tendons and muscles. In addition, the cardiovascular system could be mimicked by microfluidic channels using dyed water as the flow medium to mimic blood or other bodily fluids. This could be particularly useful in the field of biomechanics, kinesiology, and physiology.

Pressure mapping

Also, microfluidics-enabled mini/micro soft fluidic actuators could be used as a high resolution (spatially and magnitude-wise) elastic pressure mapping system when assembled in a matrix fashion. The idea here is to embed small piezoresistive sensors or micro-sized soft pressure sensors made from organic materials within the mini/micro soft fluidic actuators and microchannels as a conduit for signal traces and a pathway to inflate the mini/micro soft fluidic actuators to a predefined initial pressure level and volume. The existing state-of-the-art commercial systems are pressure-sensitive resistors,¹¹⁶ which are thin and flexible at just a few mils, but inelastic due to the dielectric layer used in the laminate. The microfluidics-enabled mini/micro soft fluidic actuator arrays are elastic and can be used in a situation requiring said elasticity, such as conforming to a contoured surface and testing the pressure distribution of cushions and apparels.

Sensory testing

Furthermore, microfluidics-enabled mini and micro-sized soft fluidic actuators could be used as a quantitative and objective medical assessment for sensory testing of patients with diabetes or sensory neuropathy in general. For instance, a sock or sleeve embedded with mini and micro-sized soft fluidic actuators along with strain-limiting layers could be worn on the foot and lower leg to test for the extent and severity of loss of protective sensation due to diabetes. The current gold standard is the Semmes–Weinstein monofilament test,¹¹⁷ which can be challenging to implement clinically and introduce inconsistencies in results. For in-depth discussions on the pros and cons of different types of sensory testing, readers are referred to an excellent review article by Craig et al.¹¹⁸

Human–machine interaction

The applications above show some improvements or novel new designs in some very practical, albeit specialized applications that are of interest to people of different and perhaps unique expertise. They could be achieved in not too long, given the application is worthwhile, and various engineering challenges are addressed. However, it would be beneficial to end this section on a grander vision, one that perhaps is far from pragmatism, but nonetheless worthwhile. In many science fiction movies, robots live and work among humans not just in the form of specialized robots that are preprogrammed for precision factory or warehouse work, but humanoid robots that can conduct various human labor and activity with better physical attributes, an advantage of mechanical systems instead of biological structures. For these humanoid robots to work safely alongside humans as nursing robots, caretaker robots, chef robots, and many more, there must exist multiple redundant safety and compliance features in the form of software and hardware. As for microfluidics-enabled soft robots, they could be used as part of these futurist humanoid robots in the form of an outer skin layer that covers parts of the robot that physically contacts human. This skin layer has two primary functions: one is for sensation, in particular, pressure feedback for the humanoid robot's fine motor control for delicate movements and another is the precise actuation of the fingers, head, and facial expressions. Of course, a futurist humanoid robot will be a combination of different technologies with one of them being microfluidics-enabled soft robots.

This section is by no means exhaustive. Therefore, the focus is to provide some insights into what microfluidics can do in addition to life science and a glimpse into the many hopefully high impact and some wild applications or directions that one can take by synergizing microfluidics and soft robots. In the next section, an often neglected subject during the research and development of novel systems and devices is discussed, that being aesthetics and ease of use.

CONCERNING AESTHETICS AND EASE OF USE

Anyone who has worked with soft wearable robotics or wearables, in general, knows the importance of a low-profile sleek form factor, the 21st-century athleisure look; that elusive material or system fueling our imagination; that one day our clothing without changing its form factor can be of physical assistance to us. Regardless of applications, what ultimately frustrates or stall many from implementing the ideal system imaginable is the technology available to them at the time, the fundamental bottlenecks that stand in the way for real-world adoption, the genuine issue of time and resource, and the way of conceptualizing.

Since synergizing microfluidics with soft robots is revolved around and has its utility most magnified in the creation of soft wearable robots. One of the key requirements for wearables is its aesthetics and ease of use, which plays the key role in user acceptance and thus justifies the need of a section. As mentioned briefly in the previous sections, aesthetics is a highly subjective topic and can be challenging to quantify. Engineers and designers strive to create goods that appeal to the masses, which is particularly evident in the consumer electronics sector.¹¹⁹ Therefore, soft wearable robots, a consumer product that is worn on the human body, could be aesthetically scrutinized in the same vein as consumer electronics.

User acceptance and device abandonment have always been issues in assistive technology and wearable technology in general. One-third of wearable assistive devices have been abandoned due to various reasons.¹²⁰ Assistive devices come in many different sizes, form factors, and purposes ranging from hardware to software. For instance, aesthetics for a wheelchair is entirely different to wearable insulin pumps or an iPad app for cognitive training or a knee brace for managing osteoarthritis. One can see those above-mentioned devices fall into different categories regarding the inherent form factor and necessity of assistive technology and, therefore, cannot be judged according to the same standard. For instance, aesthetics is more forgiven for a device if it only needs to be worn for a short amount of time, is crucial to the user's everyday wellbeing, has historically been bulky and existed for a long time, or has been accustomed by users for its look, users tolerate. Figure 19 is a simplified yet, in our opinion, unequivocal attempt to try to illustrate the importance of aesthetics and ease of use. As I mentioned before, aesthetics is subjective, and Fig. 19 is subjective and does not highlight the multifactorial nature of aesthetics and, by extension, the art of design; still, the message is clear. There are many unknowns regarding the role that soft wearable robots will play, but one thing certain is that soft wearable robots bring a new paradigm regarding functions and designs of wearable devices; therefore, as much as function trumps all, aesthetics and ease of use are perhaps paramount for user acceptance and adoption.

FIG. 19.

Relative importance of aesthetics and ease of use for soft wearable robotics to achieve user acceptance and adoption.

Factors for user adoption

Soft wearable robots are inherently compliant in the material sense, which is inherently advantageous regarding aesthetics and ease of use as they can conform to the body while potentially donned in ways similar to that of apparel. This is echoed in a review by Veale and Xie in 2016,¹²¹ where they looked at the current and emerging compliant and wearable robotic orthoses and documented advantages and disadvantages of various actuation systems. Their review analysis reveals that compliance, high specific power and force, natural motion characteristics, infinitely variable back-drivability, ease of control, and efficiency are the key performance requirements. At the same time, low mass, slim form, low cost, modularity, environmental compatibility, and quietness are key physical requirements. Last, as a must, sanitary cleanliness, safe exposed parts, and limited range of motion, speed, and force are safety requirements.

Not all metrics benefit from miniaturization; therefore, a closer inspection of the metrics is worthwhile. Infinitely variable back-drivability and ease of control perhaps do not benefit much from miniaturization, as they are more related to the design and implementation of the soft wearable robot. As mentioned earlier, high specific power and force metrics may be negatively affected by miniaturization, although this could be somewhat alleviated by well-designed strain-limiting layers and placement of mini/micro soft fluidic actuators. Compliance and natural motion characteristics should see a considerable boost from miniaturization and especially digitization as multiple mini/micro soft fluidic actuators mimicking that of a large soft fluidic actuator contour to the human body better potentially allowing for less restriction to the movement of the user. Pressurization efficiency should be increased as miniaturization of the soft fluidic actuators reduces the volume of inflation; however, the efficiency of the whole system is applications dependent. Low mass is a direct consequence of miniaturizing the control box, and the slim form factor benefits from miniaturization in two ways; first, the volume of the fluidic actuators is decreased; thus, they stick out from the body surface less, and second, the tubing diameter is reduced and the control box becomes flatter. As mentioned before, most of the cost of the materials and parts associated with soft robots is from the control system; therefore, the use of less electromechanical valves, smaller pumps, and batteries reduces the cost of the device. Smaller pumps do not always translate to quieter performance; however, it does mean that thicker insulation could be used to damp the vibration and lessen the noise produced from the pump. Perhaps an added benefit of miniaturization and digitization of soft fluidic actuators is to improve better ventilation via the gaps between each soft fluidic actuator, thus improving sanitary cleanliness. The above is a general perspective on the pros and cons of miniaturization and digitization. Metrics are often intertwined; a change in one aspect has both positive and negative impacts. As a side note, Veale and Xie also indicated that to wear an orthosis comfortably all day, the weight of an orthosis located on the torso, each foot, or each hand should be less than 15%, 1.25%, or 3.75% of the user's body weight, respectively, this correlation based on a subjective conclusion from two earlier papers.^122,123 Also, the thickness of distally located orthosis should ideally have a thickness of 10 mm, which is based on various passive orthosis dimensions.

Social acceptance and user perception

We dive deeper into the social acceptance side of soft wearable robots. Although no direct studies have been reported on soft wearable robots regarding social perception and adoption, studies on wearable sensors and wearable electronics give a holistic understanding of the social challenges of soft wearable robotics, albeit indirectly. Davis proposed¹²⁵ that the perceived usefulness and perceived ease of use are the two factors that contribute to the end-user's acceptance of a new information system. Davis proposed a model that was coined as the technology acceptance model (TAM) with early computer-based information systems in mind and did not directly consider the proliferation of portable and eventually wearable technologies. Regardless, TAM sets the foundation for technology innovators and device makers to recognize the user acceptance side of technological advancements. Malhotra and Galletta¹²⁴ later extended the TAM model to include psychological attachment (Fig. 20), which encompasses various social influences that make or break people's commitment to new technologies. In Malhotra and Galletta's own words: “when social influences generate a feeling of compliance, they seem to have a negative influence on the users' attitude toward use of the new information system. However, when social influences generate a feeling of internalization and identification on the part of the user, they have a positive influence on the attitude toward the acceptance and use of the new system.” In other words, perhaps the user is more likely to adopt new technology when the said person has personal investments and appreciation for the capabilities and the overall trend of a particular technology, not just the intended use, and that piece of technology has been promoted via means of which the users have a high degree of respect and liking. Similar to Davis, Malhotra and Galletta mainly focused on user acceptance of information technology in a broad business or company setting. However, more specific studies for user acceptance of wearable technology and assistive devices began immerging around the turn of the 21st century.

FIG. 20.

The technology acceptance model extended to account for psychological attachment. Reproduced with permission from Malhotra and Galletta, “Extending the technology acceptance model to account for social influence: Theoretical bases and empirical validation,” in Proceedings of the 32nd Annual Hawaii International Conference on Systems Sciences (IEEE, 1999), p. 14. Copyright 1999 IEEE.

Wearability

In 1998, Gemperle et al.¹²⁶ explored design for wearability. They proposed 13 guidelines for wearability (Fig. 21) and examined 6 in detail, which are placement, form language, human movement, proxemics, sizing, and attachment. The researchers found that the most unobtrusive placement for wearable systems on the human body is the collar area, rear of the upper arm, forearm, back, side, and front ribcage, waist and hips, thigh, shin, and top of the foot. Softening the edges of a wearable device and keeping it within the intimate space of the wearer are important design factors to consider. Last, due to the variation in the human body, any wearable device must be designed so that it can be worn comfortably for the intended users. Gemperle et al. mentioned accessibility for different activities, weight distribution, thermal concerns, interaction issues, material preferences, and long-term effects on the body are future research directions for design for wearability.

FIG. 21.

Guidelines for wearability. Reproduced with permission from Gemperle et al., “Design for wearability,” in International Symposium on Wearable Computers Digest of Papers (IEEE, 1998), pp. 116–122. Copyright 1998 IEEE.

A few last comments on user adoption

Kintsch and DePaula¹²⁷ developed a framework for the successful adoption of assistive technology, which many soft wearable robots are. In the words of Kintsch and DePaula: “Assistive devices are often purchased and tried, but true success fails because users and their caregivers are unable to integrate the device into their daily lives.” Indeed, as mentioned earlier, roughly one in three assistive technology devices is abandoned.¹²⁰ Kintsch and DePaula mentioned that perhaps the high percentage of assistive technology abandonment is due to not involving stakeholders, especially the users in the design and implementation phases, as well as not fully incorporating the user's goals and preferences in the design. To further elaborate, the adoption process should involve the users, the caretakers, the assistive technology specialists or physical therapists, and the developers of the assistive technology device. Assistive technology has to be not only useful and usable, but also aesthetically pleasing, age-appropriate, fashionable, and culturally and socially acceptable. To this end, the developers of assistive technology should try to design devices in a way that they do not unnecessarily single out the user in his/her social environment; in other words, if a device looks “handicapped,” the user will most likely not adopt it. This idea is reinforced by Toney et al.¹²⁸ where they mentioned that due to social weight, the user of wearable technology should interact with wearable technology in a streamlined fashioned that masquerade the technology. Of course, for certain assistive technologies, it is difficult, if not impossible, to achieve what is said above. Still, the overall message is undoubtedly helpful in the design of assistive technologies, and soft wearable robotics must follow the same design philosophy as well.

Soft wearable robotics share many design philosophies similar to wearable sensors as both systems have applications in the realm of assistive technology and are worn on the human body. Bergmann and McGregor¹²⁹ reviewed wearable sensor design in a biomedical setting from the viewpoint of what patients and clinicians want. The authors pointed out that the key user preferences for wearable sensors are that they are compact, embedded, and simple to operate and maintain, further citing that wearable sensors should blend in with clothing and be comfortable. Regardless of age and health, aesthetics and social stigma are universal concerns with wearable sensors. Also, the authors mentioned that ideally, the users would forget that they are wearing sensors throughout the day. In addition to the user preferences above, users are also concerned with forgetting to wear the sensors and the cost of purchasing wearable sensors. The authors noted that the users prefer to have the wearables sensors embedded into clothing instead of wearing them on top of or underneath clothing. Last but not least, Bergmann and McGregor suggested that “any device developed explores user preferences with the same rigor as they would for engineering related requirements.”

It should be to no one's surprise that the clear-cut winner to achieve the most desirable aesthetics, as highlighted by this section, is to use smart materials that transfer electrical energy directly into mechanical energy. That been said, a simultaneous parallel path where mature fabrication techniques, as well as the tools already developed in microfluidics, should be utilized to the advantage of the designers of soft robots to achieve products that make an immediate impact in people's lives in the near future. Microfluidics could provide some of the technical answers as highlighted in the previous sections, and the philosophy that comes with microfluidics regarding miniaturization, digitization, and the intertwined nature of aesthetics and functionality is a welcoming one. Perhaps if designers consider miniaturization or the potential for miniaturization to achieve the desired aesthetics as one of the design criteria, the adoption of soft wearable robots could be profound. In this case, the embodiment of microfluidics, miniaturization, and multiplexing could be illuminating.

CONCLUSION

In conclusion, to answer the question posed at the very beginning: “microfluidics can be combined with soft robotics, but should it be?” The answer is a yes, whereby steadfast in its philosophy, but also malleable and creative in its execution. The philosophy is forthright: miniaturizing both the control system and the fluidic actuators while maintaining proper functionality whenever and wherever possible. The execution is much less straightforward; first, it must fit the application; second, it must be feasible in both fundamentals and practice, and third, it must be advantageous to alternatives. This is challenging, which is the reason driving the need of an interdisciplinary approach, with many fundamental and applied research questions to be answered, in particular, microfluidic control strategies, fabrications, and how to use these tools to benefit our society. We are only scratching the surface of what is possible from synergizing microfluidics with soft robotics; the potential is immense.

DATA AVAILABILITY

The data that support the findings of this study are available from the corresponding author upon reasonable request.