Origami-Inspired Haptics: A Literature Review

Origami, the art of paper folding, addresses traditionally conflicting requirements for designing both powerful and low-cost haptic interfaces. As of today, there are no comprehensive reviews that cover origami-inspired haptic actuation, sensing, design processes, and fabrication. To fill this void, this paper summarizes existing origami-inspired haptic technologies in terms of workspace, degrees of freedom (DoF), power consumption, control methods, sensing, and actuation. The paper also presents an iterative design process for pattern generation and origami-inspired haptic characterization. Using this design process, we demonstrate a case study that involves the design, development, and evaluation of a novel syringe haptic interface for dental simulation. Our main findings suggest that origami-inspired haptics need more mathematically optimized and versatile design, as well as novel materials that offer a room for embedded sensing and actuation.

Origami (paper folding) and kirigami (paper cutting) art are utilized to create two-dimensional and three-dimensional objects [1]. 1 Their origin is associated with the discovery of paper [1], and is believed to have evolved independently in various cultures, such as Japanese, Chinese, and European [1], [3].
Even though the original purpose of origami was for decoration and recreational activities (e.g., decorating letters and gifts or making toys), there has been a renewed and increased interest in understanding the behavior of foldable structures, both for artistic and scientific objectives.Developments in computational mathematics, box-pleating, The associate editor coordinating the review of this manuscript and approving it for publication was M. Shamim Kaiser . 1 While there is a debate on whether to include kirigami as a subset of origami or as separate entity [2], for the convenience of this paper, they will be referred together as origami, unless specified otherwise.tessellations, and wet-folding have inspired a wide spectrum of cross-disciplinary applications, such as packaging, art and entertainment, education, science and engineering.Since 1989, such developments have been presented and discussed in the quadrennial international meeting on Origami in Science, Mathematics, and Education (OSME) [4], [5], [6], [7].

B. FOLDING/CUTTING PATTERNS AND MATERIAL
The seven Huzita-Justine axioms (also known as the Huzita-Hatori axioms) define how combinations of basic folds can construct different origami patterns [8].Many of these are tessellations, which are mosaics covering an infinite geometric plan with repetitive geometric shapes, without gaps or overlaps [9].Origami tessellations involve taking a piece of paper and collapsing it into a series of repeated patterns in order to create a particular shape or to modify the mechanical properties of the structure.Tessellations are typically created in two phases: (1) pre-creasing which involves making a series of reference folds that will be needed to create the final model and (2) collapsing where the created folds are manipulated to produce the desired tessellation.
The wide spectrum of applications has inspired the development of a plethora of tessellation patterns.The most common of these include the Miura, Waterbomb, Yoshimura, and Kresling patterns [10].For kirigamis, tessellations come in forms of cuts, namely parallel and cross-cuts [3], [11].The pattern selection is of paramount importance to decide the design outcomes.Therefore, specific patterns are commonly used in different application areas.

C. APPLICATIONS IN HAPTICS
Origami principles promise to address conflicting requirements for designing haptic interfaces.''Haptic'' refers to the science of applying tactile, kinesthetic, or both sensation to human computer interactions [12].Haptic interfaces refers to haptic devices, consisting of manipulator with sensors, actuators, or both, and the software-based computer control mechanisms.Typically, optimizing for one of the features comes at the expense of other features [12], [13].Take, for example, the multiple degrees of freedom (DoF) feedback, which is the number of independent dimensions for controlling haptic feedback.While the human hand has 27 DoFs, allowing versatile movements, most equivalent haptic devices have only a few DoFs to avoid complex and expensive electromechanical designs.Traditional haptic interfaces thus have limited realism and/or high costs, especially compared to those of audial-visual interfaces such as television and virtual reality headsets.Consequently, popularization of haptic interfaces has been limited to desktop lab equipmnet and vibration-based tactile devices, such as smartphones and game controllers.
However, by using origami principles, multi DoF mechanisms can be created at relatively low cost.For instance, in the case of cm-scale vibrotactile actuator, Korres et al. estimated that origami-inspired fabrication can lower the unit cost to below $1 [14].This is because origami-inspired actuation and sensing are based on folding planar sheets of materials into dynamic two-dimensional and three-dimensional haptic requirements, such as exertable force, speed, vibration frequency bandwidth, and position resolution.The low cost also allows easy replacement or disposal of the origamiinspired structures.
Another design benefit of origami principles is its reconfigurability, which is based on the device's ability to change its stiffness and inertia.As such, origami's ability to change its shape comes in handy.Yet another benefit is its miniaturizable nature.As long as the planar material is foldable, origami can be scaled down to centimeter/millimeter scale.This gives the potential to miniaturize traditionally bulky haptic interfaces, which is especially useful in wearable haptic applications.Lastly, origami allows scalable manufacturing through adaptable, automated, printable, and rapid design and fabrication.By exploring these benefits more extensively, haptic interfaces could be innovated into more expressive, mobile, and affordable technology.
The past decade has witnessed an increasing interest in implementing origami principles for haptic applications.The rising publication trend in origami-inspired haptic interface development is shown in Figure 1.These publications are found through EBSCO title-abstract-keyword searches under origami/kirigami haptics, or terms related to the communication of haptic information, which were: ''origami/kirigami skin'', ''origami/kirigami display/pixel'', ''wearable origami/kirigami'', ''origami/kirigami rehabilitation actuator''.The data is as of.

D. PAPER ROADMAP
Several topical review articles on origami-inspired design have been published, covering important aspects, such as sensing, actuation, design process, fabrication, and potential applications.Nevertheless, the field of origami-inspired haptics is yet to be comprehensively reviewed.The present paper aims to review and analyze origami-inspired haptic applications focusing on the outcomes in the last decade.Design workflow and tools are also tailored and discussed specifically for haptic applications.The objectives of the this paper are: 1) provide a comprehensive review and analysis of the latest developments and applications in origami-inspired haptics, 2) present a high-level overview on design and fabrication processes for origami-inspired haptics, 3) discuss the challenges and limitations in origami-inspired haptics, and 4) identify future perspectives and potential crossdisciplinary innovations.This paper also presents a case study involving the design, development and evaluation of a syringe haptic interface for dental simulation.This case study serves as a demonstration of how origami principles may be used in haptics.

II. ORIGAMI HAPTICS DESIGN WORKFLOW
Currently, there are design workflows for general origami engineering applications [10] and for traditional haptic interfaces [13].However, there are no design workflows for FIGURE 2. Origami haptics design workflow.The workflow can be divided into an input, four design stages, and an output.The workflow serves as a guide for system verification and task/user validation tailored for applications in wearables, displays, E-skins, and rehabilitation [13].
origami-inspired haptic interfaces, as they involve diverse sensing and actuation mechanisms [15].Thus, this section will bridge this gap by introducing a generalized design workflow in which the existing literature can be framed.This provides insight into the gaps in the design process that can be addressed through current and future design tools, as discussed in Section III.
The proposed origami haptics design workflow is shown in Figure 2. The input is the desired profile, while the output is the completed device, both highlighted in pink.The four design stages are origami pattern generation, pattern analysis, prototyping, and haptic characterization, as highlighted in light blue.Blocks and arrows with thick borders (-) are based on origami modeling and design [10].Blocks and arrows with dashed lines (---.) are additions based on the V-model.The V-model is an iterative development process that has been adapted for the design of haptic systems, among other mechatronic applications [13].In particular, the closed-loop backward arrows between each block and between the origami pattern generation and haptic characterization blocks indicate iterative workflow optimizations that the design process might adapt when a problem occurs in a given block.For example, the pattern generation phase can be redone to improve actuation performances measured in the haptic characterization phase.The pattern analysis phase can be extended to new materials that could be added to prototyped designs.As another example, during the haptic characterization testing phase, there is much need to return to the immediately previous prototyping phase, such as for remaking or tweaking prototyped parts.Iterative approach is advantageous and quite often necessary.The general requirements for each stage, along with their relevant works and challenges are discussed below.

A. INPUT-DESIRED PROFILE
The design requirements are largely dependent on the intended use and application of the haptic interface.These requirements correspond to the haptic features discussed in Section I-C.They are closely related to the geometric nature of origami itself and are thus achieved by adaptation of geometrical transformations.For example, Giraud et al. showed that micro to cm-scale haptic devices worn on fingers are more easily fabricated and minituarized using origami principles.This was done by desigining foldable flexible sheets to act as low-profile slider-crank and cam-follower mechanisms [16], [17].These geometric requirements also tie in with material considerations of the physical origami implementation.For parallel manipulatorbased actuators, the thickness of the origami material can be dynamically adjusted at joints to output various stiffness [18].For kirigamis, material properties can be extended to programmable kinematics, super-stretchability, and dynamics [11].The use of metamaterials can create more realistic haptic experience, such as heating, flexibility, and/or soft sensing and actuation [11].Additionally, origami haptics are valued for their ability to have excellent actuation capabilities at high DoFs, in a low cost and miniature scale [17].As such, the desired input profile is designed to realize these strengths.The main challenge at this stage is in mapping the desired haptic interface to the desired profile, as this is still largely a manual intuition-based process.

B. ORIGAMI PATTERN GENERATION
After formulating the desired input profile with respect to the aforementioned design requirement considerations, the pattern generation phase helps to realize the tangible outcome.Currently, when formulating a haptic origami design, the default approach is to use direct design, which means utilizing existing crease patterns, surface shapes, or designs.The advantage of the direct design approach is that it has a wealth of literature available for general origami applications, and haptics.In terms of engineering applications, examples include the Miura pattern as fold-core and space structures, the Yoshimura pattern as beverage cans, the Waterbomb pattern as medical stents, the Kresling pattern as bio-inspired robots [10], [19] , and kirigami patterns for metamaterials, sensors, and soft robotics [11].
Many origami-inspired haptics applications are based on these and some other generic patterns through experimental analysis and modeling.Determining the right thickness and material of the origami was researched by [19] while origami joint stiffness modeling was analyzed by Zuliani and Paik and Giraud et al. [18], [20].Vibration modeling is introduced by Becker et al. [21] while kinematic modeling of origami is described by Giraud et al. [17].Details of these works are included in Table 1.
The disadvantages of direct design is that existing patterns might not be optimal.The inverse design, on the other hand, addresses this by computing novel patterns based on the desired profile.Several computational origami algorithms have been developed to generate optimal patterns, based on the Huzita-Justine axioms [10].A common approach is to represent the complex three-dimensional structure as a reduced skeleton model comprised of vertices and edges that can then be represented as a tree structure [22].A series of optimizations are then applied to this graph usually starting with the circle-river method [23] with which all the axial paths of the origami structure can be found.These can form what is called the origami ''base'' of the structure.The rest of the pattern which is not related to the axial paths can be generated on top of them without the need of any topological optimization and are mostly folds responsible for providing the necessary angles and scale to represent the structure in the 3D space.
It is worth noting that the inverse design approach should not be used in a vacuum as its capabilities are still limited.Notably, the aforementioned algorithms assume zero thickness and an uncut parallelogram as the starting origami shape [19].This means that even if the existing inverse design software cannot design for some input requirements, it might be able to do so once it can leverage material thickness and different starting shapes.Specifically for engineering applications, origami works go as far as proposing preliminary tools for inverse design of deployable structures [10], [19], [24], [25].
In terms of haptics, there are no inverse design tools based on tactile characteristics, such as vibration sensing and actuation.Closest tools attempt to create intuitive folding patterns, such as the types created by origami-folding artists through their experience.In particular, based on empirical models, Chang et al. developed an origami pattern generating software that takes into account the pattern's profile, feel, and desired user input strength [26], [27].Naturally, the challenge in this stage is to develop comprehensive inverse design tools.

C. PATTERN ANALYSIS
After an origami pattern has been generated through direct or indirect methods, it should be analyzed to test whether it can fulfill the design requirements.Specifically, the generated origami pattern's full folding motion, rigid/non-rigid folding, and structure should be analyzed, while also going through iterate design optimization [10], [11].The kinematic analysis in particular could be combined with that of traditional mechanisms used in haptics as well [13].These are common to both origami engineering and origami haptics.Much like the previous stage, more comprehensive tools need to be built specifically for origami-base pattern analyses.

D. PROTOTYPING
If an origami pattern is found to be suitable for the design, the next step is to implement it as a prototype via one of the fabrication methods.The challenge in this stage is that regardless of the chosen method, there will most likely be many time-consuming reiterations of prototyping after failing to satisfy the desired haptic characterization.The prototyping steps can be ordered as follows: 1) Pre-folding fabrication tools and methods 2) Folding Method 3) Parts Integration 4) Control Methods The prototyping phase consists of first determining which fabrication processes and materials to use in the design, and then determining which folding methods should be implemented to realize the design.Unlike other stages in the design workflow, the decision making process for the prototype stage, particularly the fabrication and folding processes, has not been well discussed in the literature, despite being a common process for any device development.
Steps for prior stages, namely generation and analysis, are well understood for the corresponding software.For parts integration and control methods, there are also existing software and literature reviews [28].For later stages, namely haptic characterization, the choice of measurement and instrumentation are highly dependent on the desired haptic properties.

1) PRE-FOLDING FABRICATION TOOLS AND METHODS
The authors propose in Fig. 3 a detailed procedure solely for the fabrication and folding processes of the prototyping phase, based on existing haptics work.The aim is to categorize and feature some of the options available to origami haptics that have been used in the field.The categorization is inspired from [10] and [28], including some additional categorizations proposed in this paper.Throughout the process, the user makes decisions on fabrication processes, materials, and folding methods, based on the design profile and requirements.However, more often than not, the available materials to the user in a project influence what process to take [16].
The first and foremost point of consideration is whether the design will utilize a mix of materials.This is important as it first determines the ultimate haptic capabilities of the final design output.Some materials have physical characteristics, such as flexibility or conductivity that other materials do not, and it is often of interest in origami haptics projects to have a combination of these characteristics so as to meet haptic design requirements.Hence, to combine physical characteristics of two discrete materials into a single form, the user should select the appropriate fabrication methods.The following is a list of commonly used fabrication methods in the field and some of the motivations for choosing them.
• 3D Printing: 3D printing allows rapid fabrication of complex shapes, whether it be for sensing or actuation.It can also use a deceptively large range of materials that have excellent properties advantageous to origamiinspired haptics.As an extension, 4D printing can be applied to origami-inspired haptics as well, just as it has done for engineering and robotics applications [10], [28].Granted, the size and complexity of the prints are dictated by the workspace and resolution of the printing technology.
FIGURE 3. Details of the origami design workflow prototype stage.The process starts with deciding on the pre-folding fabrication tools and methods, which dictates the folding methods.The folding method, in turn, dictates parts integration and control.The detailed methods used are chosen through a series of questions.The diamond-shaped nodes denote decision making.
• Chemical synthesis: The chemical fabrication process holds several advantages above the electromechanical processes for its ability to directly influence or modify the physical composition of the material.For example, a non-conductive material can be made conductive, and produce voltage on contact with skin as discussed in [29].Meanwhile, other processes, such as chemical vapor deposition is discussed in [30] and parylene-based microelectromechanical system process is discussed in [31].Unlike 3D printers discussed above, the fabrication procedures for these are oftentimes manual, such as mixing materials in multiple steps, and could be more time consuming to develop accurate shape, folds, and cuts.
• Cutting: This process in its essence is the equivalent of the traditional elementary paper-and-scissor origami hand work, except that it is capable of high precision, automation, and larger scale.The process also enables cutting of surfaces for origami not traditionally achievable such as metal [32].A noteworthy mention is that this process is also the primary process used for kirigami haptic projects.The cutting process has seen some innovative results when combined with paper fabricated from the chemically synthesis process.Tabassian et al.
[33] researched a kirigami actuator cut from vanadium oxide nanowire-coated paper that activated on voltage, moisture, and light.In practice, to avoid tears that occur due to the cuts and wears, extra layers of reinforcing materials may be needed, interfering with the folding motion.
• Photolithography/ablation: This process involves the use of a laser to engrave or ablate from a surface composed of conductive nanocomposite material.The result is a paperlike surface that can sense electromagnetic signals and produce heat.This surface can also distort and restore its shape without losing its properties [26], [27], [34].In essence, it allows nearly the same design and implementation as a paper-based implementation would with added strength and capabilities.Thus, this process offers advantages in ease of sensing and actuation when used in origami haptics.Recent work involving this process have chosen materials, such as polyethylene terephthalate (PET) polymers [26], [27], kapton, and carbon fiber [35], [36], [37], which can flexibly fold and withstand high heat [38].The method in general works on single layer at the time, requiring adhesion/binding process for assembling multiple layers.
• Laminate manufacturing: This multi-material fabrication process is also known as smart composite microstructure manufacturing.The process combines a range of rapid, inexpensive, programmable planar fabrication processes, such as laser cutting [26], [27], lithography, and lamination, with or without electrical circuitry [28].Projects made from this process are able to contain multiple advantages of the material properties involved.For example, combining a flexible and durable polymeric layer of material with a fragile metal sensing layer of material would give the final resulting lamination durable and conductive capabilities.Laminate manufacturing was used for PCB production in [17] to create a custom motor that responded to electrical charge using piezoelectric materials.Piezoelectric materials are particularly useful because these materials can work through folded transmissions, especially well for millimeter scale [28].Materials like fiberglass [35], [36], [37] are also good candidates for this process.
Unlike the previous methods, this method would work strictly on sheets of materials.
• Molding: Much like lamination, this process involves making a mold that can accept various materials to create a part that has a mixture of properties.The mold itself can be made from inexpensive processes like 3D printing, and the mold can accept a wide range of materials from silicon to ethylene-vinyl acetate copolymer (EVA) [39], and even cardboard [40].The advantage of this process is that it is cost-effective and easily executable without specialized equipment, while being able to make a material that combines properties from the original sub-materials.The disadvantage is that for a more complicated structure, the material is harder to inject and remove, where workarounds such as segmenting it into multiple molds would alter its mechanical properties.
Each haptic design will have different requirements, and thus it is important to choose the appropriate materials and the corresponding fabrication methods to achieve the desired output.

2) FOLDING METHOD
After fabricating the chosen materials, the manufactured result is ready to be folded into the appropriate origami pattern.The most obvious and direct method is to fold by hand .Granted, this can be tedious and sometimes impossible depending on the material.Therefore, it is necessary to pre-crease the pattern during the fabrication phase.By doing so, the crease is easier to fold, either by hand or by other means, such as by a robotic manipulator [28].
An effective way to approach pre-creasing is to interpret fold lines as hinges.A hinge in its most basic form can be conceptualized as the line at which a plane is divided into two, after which the two newly divided planes can rotate with respect to this line as their common axis.In practice, a plane will have a certain thickness and rigidity or elasticity associated with it.Therefore, a hinge may be made on the plane through some subtraction of material.In general, a hinge will be implemented during the fabrication phase through laser cutting, computer numerical control (CNC) routing, or additive 3D printing.
On the other hand, a hinge can be implemented not by the subtraction of material, but through the direct formulation of shape.If a hinge is created in an interlocking shape (e.g.like a door hinge) during the fabrication phase, then the pattern can be folded without subtracting material or making any cuts.However, for interlocking hinges, their forms should be incorporated to the pattern form as early as during the pattern generation phase.A hinge can also either be made on a rigid or elastic structure, [10], which influences what type of hinge to implement.Refer to [19] for more types of hinges and their proposed fabrication methods.
Sometimes, the pattern may not have any need for folding at all.Some patterns are able to be self-folded solely from embedded actuation, although for complex folds, the order of folding may have to be precalculated [28].Self-folding may be conducted on a large automated scale as well.Becker et al. proposed an industrially scalable process for creating micro-scale origami sensors.The scalability comes from photo-lithographic fabrication of circuit-embedded polymer sheets.This sheet, through a chemical process, can self-fold into a 3D sensing structure [21].

3) PARTS INTEGRATION
Depending on the haptics requirements, the folded origami pattern may require sensors or actuators to fulfill their aim.Especially, practically all moving patterns will require actuators to operate.Sensor implementation depends on the design requirements, and should be used if the haptics interface needs information input from its surroundings.For example, Banerjee et al. developed a surgical arm based on origami patterns that used tactile sensors based on piezoelectric resistive fabric and conductive ink.The sensor was able to provide information on haptics feedback parameters, such as force and stress [10].Unlike sensors, there is the issue of bulkiness in existing motor-based actuation, however, which needs to be addressed to enable better haptic applications.

4) CONTROL METHODS
If the origami design utilizes sensors or actuators, then it should have a control scheme to prevent extreme overshoot behavior and increase its stability as a system.Given that practically all movable origami designs use actuators, then implementing a control system is unavoidable.Like origami robots, origami-based haptics, especially those with kinesthetic force feedback, greatly benefit from sensor-based closed-loop control of their folds.Even without closedloop control, sophisticated motion control may be required, which can be thoroughly developed through both kinematic and inverse-kinematic understanding of the device.While 33314 VOLUME 12, 2024 Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
the materials used in these haptic interfaces can be linear, those that involve elastic or small smart materials must often accommodate its nonlinearities and hysteresis [28].

E. HAPTIC CHARACTERIZATION
Haptic characterization is the evaluation of the previously mentioned haptic characteristics.El Saddik et al. and Kern et al. discuss this in terms of system, tasks, and user-centered evaluation methods, based on software design concepts, such as quality of service (QoS) and quality of experience (QoE) [12], [13].
System-centered evaluation involves mechanically testing the range of capability for each haptic trait.This is in terms of physical, spatial, and temporal attributes [12], or in terms of workspace, output force depending values, output motion depending values, mechanical properties, impedance measurements, etc [13].
For a given sensor and actuator, it is useful to take note of their DoF and also their active working area.Force characteristics of an actuator or sensor are measured through a force sensor attached to the end effector of a device at test.The sensor should be calibrated to the overall haptic system with sufficient resolution to sense the smallest desired input.The parameters of interest differ from one design to another, and are typically divided into static & dynamic characteristics.Static characteristics include peak force/torque, maximum continuous force/torque, minimum output resolution, sensitivity, dynamic range, power consumption, and others, while dynamic characteristics include parameters, such as impulse response and step response [13].Dynamic parameters can be measured or derived by using vibrometers and high-speed cameras.
Through task-centered evaluation, the haptic interface is tested as a controller for a specific task.The task settings are typically simulations through virtual reality and teleoperation.The specific tasks, such as pick and place, tracing, and task primitives, are discussed in [13].
Finally, user-centered evaluation involves usability and/or performance tests.These are most commonly done through self-reporting and/or behavior analysis.Questionnaires to capture the user experience interacting with the device is an example of self-reporting.It is worth noting that usability tests are subjective as they capture the personal experience of the users.On the other hand, behavior-based assessment provides a more objective assessment, such as the quality of performance, task completion time of a particular task, the number of errors committed during the task, etc. Nonetheless, it remains a challenge to develop more objective evaluation methods.
Recently, an emerging technique for evaluating the user experience is to directly tap into the brain and evaluate neural/cognitive processes associated with the haptic interaction [41].With the help of machine learning, brain imaging techniques such as electroencephalography (EEG) can be used to assess the intricate neural/cognitive processes associated with the interaction and thus provide an objective measure of the quality of user experience [42].

F. OUTPUT-COMPLETED DEVICE
To complete the haptic interface design, there should be a decision on suitable application-specific fabrication methods, depending on the complexity and the scale of consumption and training.In many cases, this involves more automated or numerically controlled methods that produce more robust and high quality products.The challenge here is to produce industry-quality methods, as it would require more time and resource.

III. ORIGAMI HAPTICS DESIGN TOOLS
To achieve the origami haptic design process, tool combinations for each of the light gray design stages are proposed in Figure 4.The figure offers two design guides: the availability of each tool and the benefits of each tool.
In terms of availability, the origami pattern generation and analysis stages, as well as the fabrication aspect of prototyping stage, share similarities to general origami engineering applications and can be managed by repurposing existing open source origami or engineering softwares [10].These tools are colored in white as open source tools.They are freely available to the public, paving the way for widespread adoption.While some of these tools are developed on a voluntary/individual basis and does not guarantee reliable simulation and production, some others, especially finite element analysis (FEA) softwares, offer paid options with more industry-ready features.For electromechanical integration aspect of the prototyping stage and the haptic characterization and interfacing stages, as shown in gray, more closed source tools are required.While they can design haptic-specific features and evaluate them through system and task-centered evaluation, they are still under research and development, and may not be readily available or available only upon request.Lastly, tools for selecting control methods and user-centered evaluation methods are yet to be developed for origami haptics, as shown in dark gray.In fact, in many of the existing origami haptics literature, the haptic characterization was performed manually, and falls short of psychophysical evaluation.
In terms of benefits, a software plug-in for Rhinoceros 3D and Grasshopper 3D, known as Kangaroo, has traditionally been the most relevant tool to bridge the input and prototype stages [10].These correspond to the first four stages in Figures 2 and 4. Kangaroo can parametrically design origami structures through a beginner-friendly graphical user interface.Chang et al. successfully used the system to design haptic buttons [26], [27].More recently, Suto et al. developed another software plug-in for Rhinoceros 3D and Grasshopper 3D, known as Crane, based on Kangaroo's design philosophy.In addition, they added fabrication-oriented analysis features, such as origami thickness calculation [19].Both of these tools are open-source.Aside from these, machine learning (ML) is recently utilized as a design tool.The benefit of indicate closed source tools, both of which have been used in origami haptics.The dark gray fill color indicates existing tools that are yet to be used in origami haptics.Some of the blocks were taken from [10].Note that each tool's capability comes with limitations, and many of the FEA tools are commercial.
ML is that it generates optimal patterns by mapping the design's geometric parameters to mechanical properties, such as strain, stress, and movement.For instance, Alderete et al. explored inverse design of kirigami motion through neural network-based mapping of simulated deformation data to its cutting patterns [25].Despite Crane's wide capabilities, it is still short of simulating non-rigid folding, where straight facets of the origami are allowed to bend as in many traditional origami structures.As non-rigid folding is an integral feature of origami that could offer potential benefits, this can be checked through conventional finite element analysis.Ranging from commercial products, such as ABAQUS, ANSYS, and COMSOL Multiphysics to open-source alternatives, they can analyze the deformation and strain associated with the generated structure under various material and geometric settings [43].
Designed to be a full design software, Crane can be used for fabricating prototypes.Namely, Crane considers thickness of materials and availability of laser cutters, CNC routers, and 3D printers for its selection of hinge types.If the desired device needs actuation/sensing but cannot do so through innate materials, however, Interactive Robogami should be used for these features.As its benefits, Interactive Robogami considers actuation mechanisms.For example, it simulates origami movements and stability based on different servomotor placement and payload.Interactive Robogami also supports parts integration, such as generating printed circuit board (PCB) design into its origami structures [10].Nevertheless, based on the most recent work developed by its authors, this tool is not yet open-source [44].For control methods, there are no tools to select and tune existing techniques and algorithms.The closest existing tool would be the use of machine learning.Just as with origami pattern generation, the benefit of ML-based method is in utilizing large mechanical datasets to find optimal methods and parameters.However, this is yet to be attempted for origamiinspired haptics [45].In practice, the prototyping methods are more diverse than what these tools can accommodate.As detailed in the previous section, much of the literature has manual fabrication processes, even skipping software-based generation and analysis at times.
For actuation, only a few studies considered task-centered evaluation.For example, Mintchev et al. evaluate the effect of using an origami-based force feedback device on various activities, such as flight control, grasping perception, and surgical simulation [37].Williams et al. demonstrated simple VR-based surface touch torsion and shear motion [38].Kim et al. demonstrated that kirigami-sensor embedded virtual reality gloves can be used to play virtual musical instruments [48].For sensing, however, task-centered evaluation such as detecting joint movements around wrists, elbows, and shoulders are frequently demonstrated.Since they are combined with system-centered evaluation, human interface is generally required for any testing.Nonetheless, more complex task demonstrations, such as joint-controlled teleoperation [51] and E-skin controlled phone calls, were rare [29].The software development compatibility here can vary widely from haptics-friendly applications, such as Unity, Unreal Engine, and Chai3D, to more generalpurpose microcontroller software such as Arduino.Notably, Williams et al. demonstrated their setup on a Chai3Dbased VR simulation [38].Finally, user-centered evaluation should be achieved through self-reported behavior from psychophysics tests, although this was not found in the current literature.For more quantifiable analysis, users' cognitive reaction can be measured through EEG-based Neurohaptics, such as the one discussed in [41].
Based on the characterized haptic performance, a mass production ready design can be developed.As with the haptic characterization step, most of the literature falls short of mass production step.For those that do, one example of application integration was where Chang et al. arranged multiple copies of origami structures into controllers, such as keyboards and rotational knobs, which can be interfaced with typing and animation applications respectively [26], [27].In another demonstration, Salerno et al. and Mintchev et al. integrated their origami actuators into game controllers, which can be interfaced with their gaming applications [35], [36], [37].These examples oversaw extended circuit and parts design and assembly.
In general, these tools are only intended to be used as a theoretical starting point, as each tool's capabilities are limited.Additionally, the combination assumes that haptics applications benefit from having both base geometry and design requirements as inputs, computer generated origami patterns, and complex mechanical analysis and prototyping.

IV. TECHNOLOGIES FOR ORIGAMI-INSPIRED HAPTICS
Hardware technologies used for origami-inspired haptics can be discussed by categorizing them into those for sensing, actuation, or both.Figure 5 gives a visual and tabulated summary of these technologies in terms of their capabilities.The two plots and tables will be discussed in the subsequent sections.
Furthermore, Table 1 gives details on their mechanism, folding patterns, fabrication method, materials, haptic characteristics, and profile.In particular, the descriptions within haptic characteristics refer respectively to DoF, workspace volume, force (or sensing ability)/ speed, and vibration frequency.The three different types: sensing, actuation, or both, are highlighted in different shades of gray.The table will be discussed along with the plots.

A. ORIGAMI-INSPIRED SENSING TECHNOLOGIES
Some of the technologies sense force/rotation, followed by many technologies that sense strain, EMG signals, temperature, and pH. Figure 5a) shows devices with mechanical and signal sensing capabilities.Devices with both actuation and sensing capabilities are shown in both plots.''0'' values indicate lack of adaptability for the given capability.Many of the technologies have electrical signal sensing capabilities, where the ease of signal measuring is shown through the Z-axis, or the strain it can tolerate.Because many of the values from the sources were not standardized, some of the values were inferred based on their component specifications [35], supplementary video [26], [27], and signal measuring capabilities [21], [54].The table on the right of the plot reports the plotted data with their respective 33318 VOLUME 12, 2024 Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
reference.For visual presentation, entries with similar values are combined on both the plot and its table.
Notably, there can be a great range of size: from cmscale [21] to even nm-scale [56].There was a trend that each work had to make a trade-off between the plotted mechanical capabilities.This shows a lack of technology that can satisfy the three most common sensing capabilities, and also how the entries clustered around the origin of the plot, designed for multi-sensing, are just as novel as the entries on the edge of the plot.

1) FORCE/MOTION SENSING TECHNOLOGIES
Mechanical sensing has been achieved through both origami and kirigami-inspired methods.For origami-inspired sensing, the most common technology involves electromechanical sensors.Some of these used commercially available sensors.This was especially the case for parallel-manipulator based designs mentioned above.Salerno et al. [36], [57] and Mintchev et al. [37] embedded Hall effect sensors into their origami structures to sense 3 DoF.In particular, their devices can measure up to 2 N force, up to 1.2 N/mm stiffness, and ±20 • rotation.Similarly, Giraud et al. used a force/torque sensor to measure 3 DoF, comprising of 678 mN compression (400 and 150 mN for X and Y axes respectively) and 8 Nmm stall torque [17].Chang et al. used a proximity sensor to measure the compression of their kirigami rotational erection system (RES) [26], [27].This corresponded to around 10 N linear force.
Others prepared their sensors in-house for more integrated design.Li et al. folded two strips of electrodes measuring capacitance.Upon compression, their device measured up to 50 Hz of 0.4 N force [55].Yue et al. fabricated a microfiber pressure sensor to measure up to 30 mm linear displacement and 80 • pitch and roll, and 20 mm diagonal displacement, all with up to 20 N force sensing [40].On a more biomimetic dimension, Becker et al. introduced integrated micro-origami magnetic sensor (IMOS) arrays which works like human follicles for detecting tactile motion over a 3 dimensional space [21].Here, each ''follicle'' comprises of origami-mounted magnetic needles coupled with magnetic sensors so that any needle motion is sensed.Additionally, Chang et al. integrated a flat capacitor circuit under their RES for detecting touch, which is triggered at 3 N force [26], [27].
For kirigami-inspired sensing, the most common mechanism was to embed strain sensors into a parallel cut kirigami pattern.Kim et al. developed a piezoelectric strain sensor that can sense V-range output from up to 320.8% elongation [48].Lee et al. made a self-powered electroactive design which can sense up to 80% strain [52].Some other works are based on coaxial variation of parallel kirigami cuts.Evke et al. placed this pattern on shoulders, which measured up to 140 • angular displacement [49].Zhang et al. used a similar pattern which measured up to 80% strain [51].

2) OTHER SENSING TECHNOLOGIES
Reflecting the diverse senses related to haptics, there is a wide range of sensing technologies.These sensors are all fabricated in-house, as they focus on investigating useful properties tailored for wearable applications that are not addressed by conventional electromechanical systems.Except for EMG and 25% strain sensing Miura pattern origami device [54], all of these technologies are based on kirigami-related patterns, owing to its very low, flat profile for wearable applications.For example, Li et al. designed a cross cut kirigami device for sensing EMG at mV range, and can withstand up to 105% strain.The aforementioned parallel kirigami cuts also apply for EMG sensing at mV range, with max 200% and 350% for [31] and [32] respectively.Notably, Morikawa et al. embedded their kirigami device directly onto a mouse muscle for accurate sensing.In addition, Won et al. fabricated an electrophysiology sensor to measure mV-range EMG, and 10 Hz-range EEG signals [34].
In fact, the remaining applications are also based on parallel kirigami cuts.Noushin et al. created a device that can measure sweat levels between 2 and 13 pH and temperature between 22 to 62 • C as well as elongate by up to 220% strain.Similarly, Rojas et al. developed a crowd monitoring system that can elongate up to 60% [53].

B. ORIGAMI-INSPIRED ACTUATION TECHNOLOGIES
As shown in Figure 5b), most of the actuation technologies pertain to force/torque actuation.A value of 0 on an axis indicates lack of the corresponding capability or untested potential.Based on the plot, most actuation technologies focus on outputting force, while some others focus on outputting torque.These are crucial for kinesthetic feedback in haptics.Only one known work outputs a high vibration frequency of up to 300 Hz [17].This is more than the 200 Hz vibration for typical tactors used in haptics, and less than the 1000 Hz force refresh rate, which is the highest a human can sense.The entries [30], [34] also have heating capability, which is useful in simulating warm touch and material feeling for wearable haptics.As in Figure 5a), the table on the right of the plot reports the plotted data with their respective reference, and entries with similar values are combined on both the plot and its table.
Also, it is worth noting that most of these technologies were origami-inspired.This was due to lack of rigidity in kirigami-inspired structures.Similar to the figure for sensing, the works for actuation also had to make a tradeoff between the plotted mechanical capabilities.This shows a lack of technology that can satisfy the three most common actuation capabilities.Also, it should be noted how the entries clustered around the origin of the plot, designed for multi-actuation, is just as novel as the entries on the edge of the plot, as balanced features are considered as a key indicator of rich expression in haptics.

1) FORCE/TORQUE ACTUATION TECHNOLOGIES
One recurring actuation method to actuate origami-inspired haptics devices was through electromechanical motors, which can be paired with parallel manipulators.Williams et al. employed this method to reduce complexity and device footprint [38].Their device, which is the manipulator made of hinged origami folds, can deliver haptic feedback in 4 DoF with a workspace of 0.64 cm 2 with ± 30 • rotation.Its maximum force and torque feedback are 1.6 N and ±5 Nmm respectively, and its bandwidth goes up to 9 Hz.They also implemented closed-loop feedback control of their motors through thorough inverse kinematics.Specifically, their manipulator position is estimated using measured motor current.
Furthermore, Salerno et al. used manipulators with the more complex Waterbomb-evolved parallel manipulator structure [35].This demonstrated the ability of the device to simulate different levels of stiffness.The manipulator worked as a portable kinesthetic haptic interface, with pitch and roll rotation ranges of ± 20 • .It provided a button-like displacement range of 35 mm, rendered stiffness in the range of 0.02 to 1.2 N/mm, and applied force of up to 2 N with an average power consumption of 1 W. Mintchev [37] then developed more compact variations of this design, which was extended into arrays by Salerno et al. for producing discretized 3D contours, shapes, and motion patterns, demonstrating its scalable manufacturability [36].Building on these works, Zuliani and Paik created sliding fiberglass in a double layer Kapton joint to vary the stiffness of the manipulator structure [18].The stiffness is adjustable between 0.002 to 0.02 Nm/ • .Aside from using electromagnetic motors, Giraud et al. developed a low-profile three-DoF fingertip haptic interface driven by piezoelectrics, also based on the Waterbomb pattern [17].The device measured 36 × 25x26 mm 3 , weighed 13 g, and was capable of rendering up to 300 Hz vibrotactile and cutaneous force feedback.As for the force feedback, the device produced 678 mN in compression, 400 mN in shear for the Y direction, and 150 mN shear in the X direction.This was achieved through a PCB piezomotor, which consisted of a stator in the form of a PCB and piezoelectric crystals arranged in a circle on its two faces.Two rotors with flat circular faces were placed above and below the stator, held together through a spring-loaded shaft.A traveling wave along the crystals drove the rotor through friction.The actuator had 20 mm diameter and provided free speed of 96 rotations per second, 8 Nmm stall torque, and a maximum power of 40 mW.
Simple electrostatics creates actuation as well.Li et al. folded two strips of electrodes into a spring structure.Upon 7 kV voltage input, the structure repels itself and outputs up to 0.051 kPa pressure [55].For the particular structure they designed, this corresponded to around 60 mN.
Another recurring actuation method was pneumatics.Yue et al. presented a Waterbomb-inspired structure that actuates up to 30 mm in linear direction [40].It can handle loads up to 20 N. Zhakypov and Okamura presented a fully 3D printed pneumatic 4 DoF fingertip haptic interface also based on the Waterbomb pattern [46].This low profile (40 × 20 mm 2 ) device was capable of producing up to 1.3 N shear and 7 N normal, and 25 Nmm of torque.On a larger scale, Liu et al. developed a pneumatic wrist rehabilitation using the Yoshimura folding pattern [39].This device supports 2 DoF with 7.5 N force.
Acome et al. explored a potentially more miniaturizable method through electrohydraulic actuation [47].Aimed at hinge-like bends, a pouch was filled with liquid dielectric (silicone oil) and was divided into a supplying part and a swelling part.The electrical supplying part was covered with opposite flexible electrodes on both sides.When a high voltage (8 kV) was applied across the electrodes on the supplying part, the electrodes pulled each other (zipping motion) through an electrostatic force.This pushed the dielectric fluid from the supplying part of the swelling part.The shape changed the swelling part, along with the rigidity of the strain-limiting layer, resulting in angular deflection of the actuator (max angle of 80 • ).Given the dimensions of the device (weight 13 g, pouch dimension 2 cm x 4 cm and liquid volume of 3 ml), the force was about 130 mN.
While the works above have relied on traditional electromechanical transmission mechanisms, Tabassian et al. work is based on more novel approaches.While its kirigami structure can be electrically actuated, it can also actuate from light and moisture [33].Although their device's vertical actuation force is in the range of mN scale, it can effectively eliminate the hardware complication resulting from complex circuitry, warranting additional studies.

2) NON-FORCE/TORQUE ACTUATION TECHNOLOGIES
The technologies discussed so far pertain to mechanical actuation.Won et al., on the other hand, explored the rare use of kirigami as a transparent stretchable wearable heater [34].The heater temperature can rise up to 40 • C in a few seconds using a 9 V input.By utilizing the kirigami structure, it can maintain its performance even upon 200% strain.Note that this work is omitted from Figure 5b) since the plot does not have an axis for heat actuation.

C. ORIGAMI-INSPIRED SENSING AND ACTUATION TECHNOLOGIES
Some of the actuation and sensing mechanisms above have been combined to varying degrees.Both Figures 5a) and b) show devices with both actuating and sensing capabilities.For force/torque sensing and actuation, Salerno et al. and Mintchev et al. used existing electromechanical systems to achieve simultaneous sensing and actuation at around 1 W power [36], [37].Building on top of this, Giraud et al. used more powerful sensors and piezoelectric motors, achieving high vibration capability and force sensing as seen in the figures [17].
These three studies spearhead closed-loop feedback control integration.These employ either bang-bang or Proportional-Integral-Derivative (PID) controllers with manually adjusted control parameters to control their devices' output forces/torques.Together with William et al.'s work in the previous section, these four works stand out in Figure 5 as controllable and observable devices.In particular, Giraud et al.'s work is notable as a controllable device with multiple sensing and actuation capabilities, especially for force output and vibration measurement.
In terms of custom-made sensing and actuation, Li et al. designed their device to alternate between an electrostatic actuator and a capacitive sensor [55].Also, Yue et al. combined their micro-fiber pressure sensor with pneumatic actuation to achieve simultaneous force sensing and actuation [40].While combined technologies are rarer for kirigami-based devices, it is equally feasible, as demonstrated by Won et al. in combining an EMG and EEG-sensing electrodes into a thermal actuator [34].In theory, these two works can also benefit from closed-loop feedback control.

V. DESIGN METHOD DEMONSTRATION
We demonstrate a design framework where origami plays a central role in simulating haptic feedback, essential in the medical training of dentists.A large part of dentistry training relies heavily on developing high tactile awareness during several standard dental procedures [59], such as simple filling procedures, root canal therapy, application of anesthesia, and the use of periodontal probes during gum examinations for pockets.All of these procedures are subjectively assessed through the tactile feedback of the dentists, which strongly depends on the quality of their training and experience.Similar to driving schools that have heavily embraced the use of virtual simulators to provide affordable yet effective training hours, medical training has also started to incorporate simulating tools, such as virtual, augmented, and mixed reality training modules [60].However, predominantly, these modules are aimed at producing intricate visuals and 3D models, with little emphasis on tactile feedback.
Haptic technologies are thus essential in devising training systems for dentists [59], [61].Application of anesthesia is among the routinely performed procedures that precede any dental operation.However, relevant training is limited to dummy prototypes or live subjects, mostly students themselves or through a few training sessions on patients visiting university hospitals.To provide unlimited training hours to students, we present, in terms of our workflow, a case study of designing an instrumented syringe.The syringe comes with origami cartridges in place of fluid in the syringe barrel (Fig. 6a), integrated with a compatible interface with virtual reality (VR) systems (e.g., Geomagic Touch [62]).Design input-wise, the objective was to enable students to obtain the exact tactile feedback of an actual syringe and its interaction with the patient's gum during anesthesia application.For origami pattern generation, as shown in Fig. 6a, we employed an origami column made from an opposite chirality stack of Kresling springs.They are highly versatile and capable of full force-displacement profile customization.This is the first implementation of a haptic device that is designed based on the Kresling origami pattern.
Through polyjet additive manufacturing, we have recently been able to transform non-functional paper-based models into reliable and functional models using a two-phase composite [58], [63].Soft and rubbery materials are used for creases, which correspond to the black phase in fig.6b).Further, hard materials are used for the stiff panels, which are the yellow facets in fig.6b).The shape and behavior of the Kresling unit depend on its geometric parameters, namely the initial predeformation height u o , the initial predeformation rotation angle, φ o , the radius R, the overall sheet thickness t, and the crease width w (in black).Depending primarily on (u o /R, φ o ), Fig. 6c shows two very different quasi-static restoring force responses that can be exhibited by the Kresling unit spring: quasi-zero-stiffness (QZS) behavior and bistability [58].
For the required tactile feedback design input, the maximum recorded pressure measured in the syringes during the actual application of anesthesia is around 2000 mm Hg (i.e., 266.5 kPa) [64].Based on the typical plunger's surface area, the feedback force after the onset of sliding of the plunger when pressed is a flat profile plateauing around 1.6 N (see blue curve in Fig. 6d).In order to achieve the opitmal Kresling stack pattern, we iterate over a few cycles of simulation-based pattern generation and analysis.To simulate the tactile feedback of actual syringes, we employed a high-fidelity finite element simulation to calibrate the geometric parameters of the Kresling unit to match the required force [58].We found that a Kresling unit with (u o /R = 0.9, φ o = 52.5 o , t = 0.75 mm, w = 1 mm) can simulate the intended force profile.Both the experimental response and simulation match well (Fig. 6d).Finally, a stack of 12 units arranged with opposite chirality is designed.
The prototyping stage for the origami pattern follows the 3D printing method from above.More details on the choice of materials and fabrication methods, as well as the syringe assembly, can be found in [65].The syringe housing the origami spring comprises four parts: (i) a syringe and the origami column with slots to house, (ii) two linear encoders, with (iii) press-fit finger flanges, and (iv) a plunger.Figure 6e shows the assembly process of these parts to yield the final device (Fig. 6f).The distal end of the syringe contains an audio jack connection that is compatible with VR interfaces.
The haptic characterization of the syringe were ordered by system, task, and user-based characterization.For systembased characterization, the plunging resistance of the assembled syringe was measured and compared with the required tactile feedback values.The characterization experiment is detailed in [65].For task-based characterization, as seen in figure 7, we attached the syringe to the touch haptic device, and then used with a VR simulation and Meta Quest Pro to verify if we can conduct the training with the setup.The syringe is connected to a Geomagic haptic interface.The entire position of the syringe is reflected in the VR environment using the Geomagic device.The linear encoders track and relay the location of the plunger to the Geomagic device, so as to reflect the plunging kinematics within the virtual reality environment.Besides capturing the haptic feedback of touching the different parts of the oral region in VR, we verified that students can experience the tactile feedback during the application of anesthesia.In order to meet the taskbased characterization, we reiterated the origami generation and analysis phase to modify stack size from 12 to 10, as well as the prototyping phase by reprinting the syringe enclosure over a dozen versions to ensure smooth plunging motion.Finally, for user-based haptic characterization, we conducted a preliminary user study in [65] .In the user study, dental experts observed that the syringe interface felt realistic and that it would significantly improve the training productivity.
Finally, we identified challenges for each workflow stage.For the design input stage, our dental experts suggested that the most crucial feature was to accurately track the visual movement of the syringe plunger on the VR interface.This prompted us to re-evaluate our design input conditions.For pattern generation and analysis, the challenge was to experimentally and comprehensively verify the relationships between the geometric parameters and its mechanical properties [58], [63].The prototyping and haptic characterization stage had the most extensive challenges.One challenge was that the syringe interface is larger than the actual syringe.While the added size is coming partly from the Touch Haptic Device's end effector size and socket length, this is also coming from the origami spring fabrication method.While slimmer springs can be designed, our polyjet 3D printer cannot print the thin origami walls required for this design.Due to its thickness, the spring needs to be longer than the actual syringe to produce the same range of motion.Also, unlike the concentrically symmetric simulation models, real life springs have minor asymmetries that cause buckling for taller stacks.With a combination of thin rubber creases and buckles, the syringe is prone to wear and tear.In terms of reaching the output stage, there needs to be more workflow iterations to better satisfy our dental experts.While there were design workflows for general origami engineering applications and for traditional haptic interfaces, there are no design workflows for origami haptic interfaces.Thus, the authors proposed an integrated design workflow to bridge this gap.Unlike its predecessors, the proposed workflow captures the iterative nature of device design based on both origami pattern analysis and haptic characterization.The workflow was explained in detail in terms of input profile, origami pattern generation and analysis, prototyping, and haptic characterization.These are demonstrated through the development of the Kresling origami-inspired syringe interface.
The authors also presented the currently available origami-inspired haptic technologies.Some technologies sense forces up to 70 N.This is more than enough for most haptic motion sensing.Granted, such mechanisms use somewhat bulky electromechanical sensors.Rotation sensing, on the other hand, can go up to 150 • , and is flat and flexible as it is based on kirigami.Many others sense strain, with up to almost 400% measurable elongation, EMG signals at mV range, temperature between 22 and 62 • C, and pH between 2 and 13.
For actuation, many technologies provide force and torque up to 20 N and 25 Nmm, respectively.These are more than enough for most haptic use cases.These technologies typically required bulky pneumatic actuators.Actuation through electromechanical means results in less than a few Newtons of force or a few Newtons millimeter of torque.In exchange, however, these tend to excel at vibration frequency, with one device vibrating up to 300 Hz.Some devices have managed to balance multiple performance metrics well, such as Giraud et al. work [17].In addition, some works can electromechanically actuate based on motor sensor-based closed-loop feedback control.

B. CHALLENGES 1) LIMITATION AND VOIDS IN DESIGN TOOLS
The current proposed workflow separates origami pattern generation and prototyping so that the origami pattern analysis and haptic characterization of prototypes are separate.While this was meant to simplify the design process, incorporating the prototype phase into origami pattern generation, including materials and sensor/actuator integration, for more comprehensive simulation analysis, would reduce the burden of physical testing.
In terms of the design tools, there is currently no known tool that can complete a full stage origami interface development.Furthermore, the design tools for each of the four design stages are limited.For the pattern generation stage, most tools do not support inverse design.For example, a given software cannot create optimal designs given an input that it never encountered before.For the analysis stage, there is limited structural and kinematic analyses, as aside from Crane that recently appeared, the design tools used here are based on simplified mathematical models and simulations.In fact, there is no consideration of non-rigid folding.For the prototyping stage, much more exploration is possible for materials and fabrication options.For instance, the tool should be able to simulate multi-layered origami sheets made of different materials.This would mean better origami fabrication methods as well, as discussed through the limitation of producing thin and robust origami design in Section V. Especially for devices that have both sensing and actuation, there should be a thorough control method implementation and parts integration feature as well.For the haptic characterization stage, while there is system-centered evaluation through instrumentation (i.e.embedded sensing), only a few works conduct task-centered evaluation, and none of the works perform user-centered evaluation.Also, there are no commercial-grade open-source performance measurement tools as indicated in Figure 4.

2) LIMITATIONS AND VOIDS IN CURRENT TECHNOLOGY
While it was possible to compare devices against a single performance metric, it was harder to rank their overall performance, as many of the devices were developed with specific applications in mind.In fact, no work has attempted to develop a general purpose origami haptics device.For sensing, despite the wide range of capabilities, for all of the technologies, high sensing performance of one metric comes at the cost of other sensing metrics, of which its inverse relation was discussed in Figure 5.For actuation, aside from force/torque, some technologies served as rapid heaters.Similar to the trend for sensing technologies, high actuation performance of one metric comes at the cost of other actuation metrics.Thus, there is a clear void in creating devices that excel at multiple sensing and actuation metrics.On another level, devices with both sensing and actuation should utilize closed-loop feedback control.In particular, there is a lack of closed-loop feedback control implementation on non-electromechanical motor mechanisms.

VII. FUTURE RESEARCH PERSPECTIVES A. DESIGN PERSPECTIVES
Based on the limitations discussed above, future design-based research direction can include, but not be limited to, the following; • Development of a unified workflow scheme that can handle all stages from input to output.
• Use of machine learning and artificial intelligence (AI), especially in the origami-inspired haptic design generation.
• Mathematical modeling of how different design features in each of the workflow stages map to origami-inspired haptic behaviors.
• Design of high DoF haptic interface due to modular design and fabrication.
• Design library for materials, sensors, actuators, kinematics, and control.
• Simulation of actuator and sensor-integrated origamiinspired haptic design.
• Standardization of what software, hardware, and human experiment protocols to use for haptic characterization.

B. TECHNOLOGY PERSPECTIVES
In the long term, devices with multiple balanced capabilities should be developed.This means developing haptic interfaces with multiple modalities of haptic actuation, such as vibration, force, heat, etc.Some ways of achieving this are as follows: • Exploration of novel materials that offer embedded sensing/actuation capabilities, in addition to the origami structure.These include soft materials that allow elastic and/or curved folding, as well as self-actuating materials such as shape memory polymers.
• Benchmark the standardization for comparing the performance of these origami-inspired interfaces.As different applications have various requirements, it may be worth creating application-specific benchmarks.For example, wearable applications would value devices with essentially flat size profiles and low power consumption.Conversely, desktop tactile displays can tolerate devices with high power and medium size profiles.This in itself could be an independent research topic.As noted by Kern et al. these standardization are beneficial [13].

FIGURE 4 .
FIGURE 4. Design tools relevant throughout the origami haptics design workflow flowchart.The large rectangular blocks Also, each of the design workflow stage highlighted in light gray .correspond to the blocks from Figure 2. Blocks with white fill color indicate open source tools, while those in grayindicate closed source tools, both of which have been used in origami haptics.The dark gray fill color indicates existing tools that are yet to be used in origami haptics.Some of the blocks were taken from[10].Note that each tool's capability comes with limitations, and many of the FEA tools are commercial.

FIGURE 5 .TABLE 1 .
FIGURE 5. Literature map for a sensing parameter space (Maximum strain-rotation-force), b actuation parameter space (Maximum vibration frequency-torque-force), together with their respective raw data and reference(s) tabulated to the right of each 3D scatter plot.Note that the raw data for b was too scattered and non-uniform to be captured by the fitting surface.

FIGURE 6 .
FIGURE 6. Case study demonstration and Kresling spring-based pattern: design and fabrication.(a) An actual syringe used during the application of local anesthesia using anesthetic fluid, where fluid interaction with the plunger is substituted with a dry Kresling-based column.(b) A paper-based model transformed into a functional model through additive manufacturing.Kresling spring behavior of (c) quasi-zero stiffness (QZS) and bistability [58].(d) Calibration of the Kresling-based column to match the actual plunging resistance exerted on the thumb of dentists.(e) Fabrication and assembly of final design parts.(f) A functional and instrumented design that is compatible with a virtual reality (VR) interface.

FIGURE 7 .
FIGURE 7. The final syringe design incorporated with the VR interface for training dental students.