1. Introduction
In today’s social context, where several factors such as sedentarism, diet, and stress can cause traumas or impairments in the population, this is at the root of the increasing number of people suffering from various forms of disability. According to statistics reported in 2017, the most significant cause leading to various forms of disability is due to stroke [
1]. Similarly, in 2012, according to a report by the American Health Association (AHA) in the United States of America, around 4 million people were suffering from a range of injuries due to stroke [
2]. Other studies show that more than 80% of people who have suffered a stroke remain with some movement impairment in their upper limbs [
3,
4], hampering or limiting the performance of activities of daily living (ADL). They need repeated rehabilitation exercises over a long period in order to be able to regain their original mobility [
5,
6]. Due to the increase in the number of people needing post-stroke rehabilitation, as well as limited specialist medical staff to treat stroke [
7,
8], there is a need to identify effective methods to help them regain autonomous movement. This “self-help” type of training has the potential to reduce some of the burdens on the healthcare system, namely by offering stroke patients the opportunity to perform rehabilitation training alone at home or at their desired place.
Both commercially and in the literature, several robotic variants meet rehabilitation or assistive needs by guiding repetitive movements with different ranges of motion in the hands, elbow, or shoulder joints. Available variants such as PEXO [
9] or TENEXO [
10], HANDEXOS [
11], or the exoskeleton-type device proposed by Ho, N. et al. [
12], use rigid joint links driven by DC motors. These variants have advantages in terms of good modeling and control capability providing accuracy, but they also have some less favorable aspects in terms of size, shape, weight, and comfort of the devices. These are important criteria for the success of such a device. Based on these considerations, over the past decade, researchers have been concerned with developing various robotic systems (with rigid or soft elements) that possess features inspired by the living world, which provide a transition from nonredundant to redundant, hyper-redundant, continuous, and soft systems to increase the flexibility, adaptability, and biocompatibility of the systems [
13,
14]. Wearable hand rehabilitation/assistance devices and other devices using soft actuators represent an important research direction in the field of soft robotics and there is a large diversity of such devices. In part, soft robotic devices for hand rehabilitation use fluidic soft actuators such as the handheld device of the Wyss Institute at Harvard University called “Wyss Soft Robotic Glove” [
15] or the commercially available devices of Syrebo called “Syrebo Hand” [
16].
Analyzing the literature, we have also found several publications addressing handheld devices with soft actuators for hand rehabilitation/assistance at different levels of implementation, some of which will be presented shortly. Hong Kai. Yap et al. developed a portable glove-type device based on pneumatically actuated fluid actuators made of flexible thermoplastic materials and coated with thermoplastic polyurethane (TPU) that have the ability to provide bidirectional assistance of both flexion and extension movements [
17]. Another approach integrating bidirectional fluid actuators is presented in the work of Heung KHL et al. who proposed a wearable glove designed for everyday activities (ADL). The actuators are made of elastomer with two inner chambers that are individually pressurized to actively assist both flexion and extension of the fingers. Their structure is uniform throughout their length, and they are fitted with a flexible sensor that monitors the level of flexion of the actuators, whose insertion is between the two chambers. The authors proposed an analytical model that aims to quantify the output force of the actuators in the case of squeezing in relation to the relationship between the input pressure, the flexure angle, and the output force. All these analyses have been validated numerically and experimentally [
18]. Jiangbei Wang et al. proposed a wearable lower limb rehabilitation device based on 5 PneuNets fluid actuators that actively assist flexion and passively assist extension. The actuators are designed according to the anatomical characteristics of human fingers, with the actuators having three joint segments corresponding to the three finger joints (2–5), two for the thumb, and four rigid segments corresponding to the phalangeal and metacarpal bone segments, respectively. This approach allowed the actuator characteristics to be designed to have kinematics and a range of motion similar to those of human fingers, a feature achieved through numerical and experimental modeling [
19]. The literature review also identified hybrid wearable glove actuation solutions using both pneumatic and cable actuation. In the article by Lucas Grez et al., such a hybrid drive wearable exoskeleton device with variable stiffness for ADL activities is proposed. Furthermore, the glove is designed to assist both flexion/extension and abduction/adduction movements of the fingers. Their tests aimed to improve the grip of various objects by using an additional telescopic finger that is pneumatically actuated. According to the results presented by the collective, the robotic glove in the configuration presented has considerably improved patients’ grip ability [
20]. Computerized textile manufacturing technologies have been integrated into the creation of wearable actuators and devices to be incorporated into clothing. Hend M. Elmoughni et al. proposed a pneumatic actuator produced entirely by computerized knitting without using adjacent cutting or sewing technologies. Different knitting configurations were analyzed, and a grip assist device was invented able to perform active flexion movement at a pressure of 150 kPa and grip objects weighing up to 125 g [
21]. In the specialist literature, wearable devices intended for the rehabilitation of the thumb have also been identified. Paxton Maeder-York et al. have constructed a wearable device intended to rehabilitate the flexion movement of the thumb through active control. The device uses a pneumatically actuated fluidic actuator composed of elastomer and reinforced with fibers to increase force characteristics. The configuration of the reinforcement braid is different to be able to achieve the kinematics of the human thumb as faithfully as possible [
22]. Another article focusing on the thumb is by Yuanyuan Wang et al., which proposes two different approaches and analyzes them comparatively on the realization of flexion/extension and abduction/adduction movements. The two variants are based on soft fluidic actuators, the first variant being one that uses two elastomeric and fiber-reinforced actuators for the thumb and index finger corresponding to the flexion and extension motion of the fingers and a fan-shaped fiber-reinforced soft fluidic actuator. Once it is pressurized, it makes an angular movement driving the thumb into an abduction/adduction movement. The second configuration uses a reinforced soft actuator that performs flexion/extension motion of the finger and a fiber-reinforced actuator with three internal cavities that can perform both flexion/extension and abduction/adduction motion. Results show that each variant conveys advantages in some aspects of rehabilitation and a combined configuration would lead to substantial improvement [
23]. Panagiotis Polygerinos et al. proposed a control system for a rehabilitation/assist glove that is based on the capture of myoelectric signals (EMG) from the forearm muscles through surface electrodes. The glove is made of elastomeric material fluid actuators that assist in flexion and extension movements of the hand phalanges [
24]. The analysis identified a wide variety of soft wearable systems intended for hand rehabilitation or assistance. Most of these use air-pressurized fluidic actuators as actuation methods in a variety of configurations and features to meet the patient’s needs for strength, range of motion, and comfort in the recovery process. Also, most devices focus on assisting finger flexion/extension movements and less on abduction/adduction ones. The main component of these soft wearable exoskeleton devices are soft actuators, and with the accelerated development of the field of soft robotics in the last decade, a multitude of types of fluidic actuators have been designed for various applications. However, with the development of the field, their complexity in terms of manufacturing, actuation, modeling, and control has been of spectacular growth. Based on these considerations, the present work focuses on the creation and validation of a wearable device that is easy to wear and use and foremost has low manufacturing costs.
Generally, in the rehabilitation field, devices that provide sufficient biocompatibility and biomimeticity are predominantly used due to the provided comfort and lightweight. In terms of soft actuator systems, the most common types in the rehabilitation field are those based on fluid actuation, shape memory alloys, and dielectric elastomers [
25]. Regarding the modalities of intention detection and sensors used in the control of the device, the majority of publications use control based on the capture of EMG muscle signals from the forearm muscle area, with this method being the most predominant in wearable devices. Another prevailing method is based on pressure and flex sensors [
26].
The use of soft robotic systems in portable devices has the great advantage of offering patients the possibility of rehabilitation/assistance in the comfort of their own homes. Moreover, the use of fluid actuators made of elastomeric materials grants patients a safe interaction due to highly elastic soft materials and a large number of degrees of freedom, which implicitly provides the possibility of performing a wide range of movements, the low cost of producing such a device, and the corresponding high portability and manufacturing devices customized to the patient’s anatomical data. Based on the aspects presented above, this work addresses the realization of a low-priced portable device intended for rehabilitation/assistance of hands at the patient’s home for people who have suffered a stroke or other neurological diseases following which they are left with a range of mobility impairments. The wearable device is made with PneuNets fluid actuators that assist in the flexion and extension motion of the phalanges, specifically the DIP (distal interphalangeal) joint, PIP (proximal interphalangeal) joint, and MCP (metacarpophalangeal) joint at fingers 2–5 and IP (interphalangeal) joint and MCP (metacarpophalangeal) joint at finger 1 (thumb finger) in
Figure 1. The actuators are positioned on the dorsal side of the hands, being actuated from a single actuator source, and the control is based on EMG signal capture with a commercial module with surface electrodes positioned in the forearm muscle area. This gadget also includes a pressure sensor and a flex sensor to monitor the air supply pressure and identify the amplitude of phalangeal movement. Its weight is 180 g, putting it in the optimal range for such devices [
27,
28].
2. Materials and Methods
2.1. Actuator Design and Fabrication
PneuNets actuators belong to the bellows actuator category, in which the specific bending movement is achieved by the deformation of the inner cavities. The actuator used benefits from a uniform bending movement over the entire surface, and it generally possesses the advantage of a fast response when pressurized, usually at low actuating pressure values. To achieve uniform deformation, the existing inextensible layer limits axial expansion during pressurization. Therefore, the bending motion is actively controlled during actuator pressurization. The internal pressures cause bending and lead to the production of mechanical energy while also accumulating elastic energy. As for the extension movement, it is passively controlled by different levels of pressurization as well as by the elastic energy accumulated by the actuator through bending movement, more precisely at the moment of pressurization. The choice of this type of actuator with this geometry was based on a number of advantages related to the simplicity of construction, low cost of realization, and meeting the needs of force and range of movement in the case of finger actuation, where the forces are relatively small. In addition, this type of configuration has proven over time to be the appropriate variant for such cases, receiving validation from the scientific community.
The 3D design of the actuators was carried out in the SolidWorks 2021 software in two different lengths. The actuators were chosen to be made in two different lengths due to the different anatomical dimensions of the human fingers. The main dimensions of the actuator and its components can be found in
Figure 2.
The four elements that make up the actuator are the main body, through which the specific bending movement is achieved; two layers of silicone that enclose the main body; and between these two 3 mm thick layers of silicone, there is an inextensible layer of paper inserted to limit the axial extension, helping to achieve the bending movement. The dimensions of the main body are l—8 mm, d—2 mm, t—2.5 mm, and b—15 mm. These sizes were determined by numerical and experimental finite element modelling in the section below. They are identical for all 5 actuators and, in addition to these dimensions, the actuators have a width of 20 mm, while the lengths are 95 mm for fingers 2 (index), 3 (middle), and 4 (ring), and 75 mm for fingers 1 (thumb) and 5 (little), respectively. These dimensions were determined experimentally, encompassing the DIP, PIP, and MCP joints at which the flexion/extension movement is performed.
In the field of soft robotics, the most commonly used manufacturing processes are casting and 3D printing. In most situations, these two techniques are used collectively, with 3D printing being the technology for producing molds [
29]. In this case, too, we have used casting as the main process to manufacture the actuators, and the molds were fabricated by 3D printing technology using polylactic acid (PLA) as the material. The 3 component molds needed to make the actuators were previously 3D modeled and 3D made using a CraftBot printer with a single extruder. The material from which the actuators were made is a two-component RTV ZA 22A silicone with a Shore hardness of A 22, composed of two elements A and B (base and catalyst), which are dosed in equal quantities. The specifications of this material are given in
Table 1.
The manufacturing process by casting is based on a rigorous realization process. Following the mixing of the two components, which have been weighed beforehand, in equal quantities, a series of air bubbles accumulate in the material, which, once they enter the actuator, can alter its behavior or, worse still, cause it to break during pressurization. Therefore, an intermediate process was carried out to remove the air bubbles utilizing a vacuum system. Inside the vacuum plant, the silicone was left for 2 min at a negative pressure of −85 kPa. The manufacturing process is shown in
Figure 3.
After the homogenized silicone was removed from the vacuum plant, it was poured into the 3D-printed mold, both for the main body of the actuator and for the two silicone layers that enclose the actuator. Two hours later, the actuator was removed from the mold and assembled. As an inextensible layer between the two silicone layers, we have used one made of paper.
2.2. Material Properties and Finite Elements Methods (FEA)
RTV ZA 22A is a high-compliance hyperelastic material that possesses the ability to elastically deform several times more than its initial value at the moment of stress. Determining the mechanical characteristics of the material is important to establish and validate the real and simulated behavior employing the FEA of the actuators used. For this purpose, the tensile mechanical characteristics of the RTV ZA 22A material were determined using the ASTM D12 (method A) standard [
30]. The dimensions of the specimens are 3 mm in thickness, a total length of 115 mm (where the length of the calibrated part is 33 mm), a width at the ends of 25 mm, and a width of the calibrated part of 6 mm. The tests were performed on a sample of 5 specimens on the Titan 2 Universal tensile testing machine (software version 7.0.4.14642) at a speed of 50 mm/min and with a cell force of 600 N, as shown in
Figure 4b. From the experimental data obtained, for the accuracy of the results, outliers were removed. Based on the experimental results taken from the machine software, the characteristic curve in the form of engineering stress and engineering strain was plotted, with the curve representing the average of the values of the 5 specimens tested. The test results are shown in
Figure 4a.
The experimental data obtained from the tensile tests provide the possibility of determining a number of characteristics such as material elongation, Young’s modulus of elasticity, and material constants needed to perform finite element simulation of the actuator. This is essential to determine the approximate behavior of the actuators tested. Also, another element to consider was to identify the actuating pressure required to achieve amplitude of motion values as close as possible to the phalanx amplitude of motion values. In the case of the PIP joint, the value is approximately 120°. The conformity between the two values leads implicitly to greater efficiency and to a shorter recovery time.
Obtaining this information was achieved using the finite element program Abaqus (Dassault Systems—6.13-1), and the simulation is identified in
Figure 5a. Modeling of the PneuNets actuators was performed using the Yeoh hyperelastic model [
31], for non-linear and incompressible materials based on the experimental data of the graph in
Figure 4 engineering stress and engendering strain. The strain energy function for determining the 2nd-order material constants is shown below in Equation (1):
where W is a strain energy density, C
i0 are coefficients of the polynomial function,
is the first strain invariant, and J is the determinant of the elastic deformation gradient. The values of the Yeoh material constants of degree 2 obtained are C
10 = 48.673 (Pa) and C
20 = 962.03 (Pa). From the simulations, it can be seen compared to
Figure 5b that the simulated bending amplitude is approximately equal to the actual bending one.
2.3. Relationship between Pressure/Angle/Force in Bending Phase
To identify the characteristics that hold the actuating pressure, the amplitude of movement, and the force that the actuator can develop at different input pressures, an experimental test was carried out. Thus, two graphs (
Figure 6a,b) were made in which the output characteristics, namely the bending angle and the force as a function of the input pressure, were determined.
Determination of the bending angle so that the value is as close as possible to the amplitude of the natural motion was accomplished according to the inlet pressure. A resistive flex sensor was used and integrated between the actuator layers. This sensor provides relatively accurate information about the angular position of the actuator as a function of the pressure inside the actuator chamber. We have used the Honeywell ABPDANV150PGSA3 pressure sensor, which has a pressure range of 0–150 psi. From the graph in
Figure 6a, it can be seen that at a pressure of 0.06 MPa, an actuator deflection value of 117° is obtained, which represents an amplitude similar to that of the PIP joint. Regarding the output force as a function of the input pressure, it was measured using a force-sensing resistor. The force measurement occurs at the distal end of the actuator, and the characteristic curve is shown in
Figure 6b, where it can be seen that at an input pressure of 0.06 MPa, the output force of the actuator measures 1.3 N.
2.4. Pneumatic Glove Control System
The control of the soft glove is presented in
Figure 7. This control unit was realized at a low cost, which was possible due to the efficiency and use of alternative equipment that meets the specific requirements of the study but also has a relatively low cost. The approximate cost of making this control unit is 100 €. This control unit allows the supply of the simultaneous actuators through a pneumatic control system, which is composed of an electro-pneumatic valve (SMC—SYJ5120—24 V) that allows the pressurization and depressurization of the actuators. The pressure sensors (Honeywell—ABPDANV150PGSA3) and the pressure supply are taken from an air compressor. As for the control system, it consists of a development board equipped with a microcontroller (ATmega 328P), control relay, EMG module with three surface electrodes, a display for exhibiting information related to the number of repetitions performed by the patient, the pressure inside the actuators, and data from the flex sensor integrated into the actuator layers (grade °). Closed-loop control in the case of rehabilitation is achieved by the flex sensor closing the feedback loop. In the case of ADL, the feedback loop is closed by the pressure sensor, which provides pressure data inside the actuators. Since several voltage levels are required to supply the components, both pneumatic and control, 12 V and 24 V voltage regulators were used, respectively.
4. Conclusions and Future Works
Due to the social context in which an increasing number of people suffer strokes or other neurological diseases that leave them with various impairments affecting mobility, especially that of the upper limbs, it is necessary to carry out rehabilitation training to regain mobility and perform activities of daily living (ADL) anew. In this particular context, the present work addressed the realization of a wearable exoskeleton-type device for the upper limb to assist the patient during rehabilitation exercises, as well as the possibility of facilitating different daily activities, providing the possibility of re-enabling one’s personal environment. The device provides active rehabilitation/assistance, being based on capturing the patient’s muscle intention within the forearm muscles. PneuNets actuators provide flexion motion actuation of the phalanges (DIP, PIP, MCP) for fingers 2–5 and (IP, MCP) for finger 1, and the extension motion is passively controlled by depressurizing the actuators. Two different actuator lengths were used depending on the anatomical dimensions of the fingers. A rigorous procedure was followed for the manufacture of the actuators, based on a series of steps to ensure that the actuators would be consistent and reliable over a long period. The two main manufacturing technologies in the field of soft robotics were used, namely casting and 3D printing to produce the casting mold. Two-component elastomeric material (RTV ZA 22 A) with a Shore hardness of 22 A was subjected to uniaxial tensile testing until breakage to determine the material’s mechanical characteristics and to perform FEA simulation based on material constants. It was taken into account that the actuator performs the bending motion with an amplitude approximately equal to the amplitude of motion of the PIP joint, so the actuator behavior was determined as a function of the input pressure, and the degree of bending of the actuator as well as the output force developed by the actuator were monitored with a flex sensor. At a pressure of 0.06 MPa, the bending angle is 117° and the force developed is 1.3 N. The control unit of the wearable device is mainly based on microcontroller-based information processing electronics and flow control elements such as pneumatic valves that distribute fluid to actuators. The exoskeleton-type wearable device has a relatively elementary structure consisting of a textile glove with a silicone rubber on the palmar area to improve grip and PneuNets actuators, which are positioned on the back of the hands using nylon fibers. This structure is designed to be effortless to use and wear while being comfortable for the patient. Moreover, the cost of its realization is really low. The wearable device is mainly intended for active rehabilitation of the patient’s hands, but it is also possible to assist in various daily activities. The tests were performed on healthy subjects and were based on a series of flexion/extension movements of the fingers in the three joints (DIP, PIP, MCP, and IP), as well as through the use of different objects with different geometries and masses. These movements are based on the capture of the patient’s muscle intention through EMG surface electrodes.
As a future direction, the authors intend to increase the modularity and flexibility of the control unit, as currently, this unit brings some limitations in terms of its volume as well as its portability, also limiting the patient’s room for action in case of assistance. Another issue the authors are considering as a future direction is related to the integration of the wearable device with virtual reality (VR) and the LeapMotion controller. By using VR and the LeapMotion controller, a series of specific rehabilitation games can be designed tailored to the patient’s needs. The LeapMotion controller as an immersive element in VR provides the possibility for the patient to interact with different objects in VR, improving motivation and the level of interaction. Another aspect that the authors wish to develop is the integration of the wearable glove into a wearable upper limb device that assists both the wrist (flexion/extension and abduction/adduction) and the elbow joint in flexion/extension movement. It is also intended to validate the device by performing tests on patients with hand movement impairments due to stroke or other neurological diseases.