Development of a Robot-Assisted Telerehabilitation System With Integrated IIoT and Digital Twin

Upper limb dysfunction (ULD) is common following a stroke, spinal cord injury, trauma, and occupational accidents. Post-stroke patients with ULD need long-term assistance from therapists for their rehabilitation, which generally occurs at the hospital or outpatient clinic. Physical therapists are unavailable because of geographical, financial, and scheduling concerns, and continuity of care needs to be improved due to the need to travel to multiple locations for therapy. As a result, providing specific, tailored therapy programs is challenging due to the absence of feedback and real-time monitoring. An effective telerehabilitation system can address this issue and is more cost-effective for healthcare providers and patients than traditional inpatient or person-to-person rehabilitation. Remotely operating robotic devices and using advanced technology improves patient and healthcare provider safety and reduces injuries. In this study, we developed a novel telerehabilitation framework for rehabilitation robots utilizing PTC’s Industrial Internet of Things (IIoT) platform to remotely provide robot-aided therapies for individuals with ULD. With the developed telerehabilitation framework, an operator can teleoperate the rehab robots to deliver Upper-limb (UL) exercises via an Augmented Reality (AR) based graphical user interface (GUI). This AR platform communicates bidirectionally using ThingWorx IIOT. It leverages the digital twin (DT) structure facilitated by Vuforia studio to visualize the physical robot motions happening in remote places. The telerehabilitation framework was validated through a commercially available robot (xArm 5), an exoskeleton (SREx), and an end-effector type rehabilitation robot (DMRbot) developed at Biorobotics Lab, UWM. The experiment results show that the telerehabilitation system can successfully provide UL rehab exercises in 2D and 3D planes via AR. The proposed framework is developed to facilitate robust and more promising robot-aided rehabilitation sessions remotely, and it can also be applied in other medical applications.


I. INTRODUCTION
Stroke, spinal cord injury, cerebral palsy, multiple sclerosis, amyotrophic lateral sclerosis, trauma, occupational injuries, and geriatric illnesses are all prominent causes of functional impairments of the UL [1], [2]. Stroke accounts for 34 percent of all cases of ULD, followed by spinal cord injuries (27%), multiple sclerosis (19%), and cerebral palsy (8%) [3]. More than 15 million people worldwide yearly suffer from stroke [4]. A new or recurrent stroke affects approximately 795,000 Americans each year [5], [6], and the American Stroke Association forecasts that over 4% of individuals in the United States will have experienced a stroke by the year 2030 [7].
The primary method of promoting functional recovery in individuals with ULD is through rehabilitation [8] which is essential for those with temporary or permanent functional impairments [9]. Rehabilitation can alleviate acute and chronic illnesses, trauma, and injuries. This highly specialized medical branch strives to improve a patient's health while complementing other medical and surgical disciplines. Patients in rehabilitation often collaborate with a multidisciplinary team to determine the optimal treatment [10], [11]. 2.4 billion people worldwide have a health condition that rehabilitation can help, and this figure will continue to increase as it is predicted that by 2050, the elderly population will have doubled. The global need for rehabilitation services has increased due to an older population and longer life expectancy for people with serious health issues and impairments. As the health and demographics of the global population evolve, it is anticipated that the demand for rehabilitation services will increase.
Traditional forms of rehabilitation therapy demand extensive time and effort from the therapist, but there is a continuous shortage of specialists able to provide such individualized treatment [12]. Similarly, limited therapy sessions, the difficulties associated with transportation and face-to-face therapy sessions, as well as excessive expenditures frequently serve as the impetus for these persons to withdraw from rehabilitation services prematurely.
Robotic therapy, which enhances the quantity and intensity of therapy while standardizing treatment, is useful for rehabilitation. Many positive outcomes are achievable if robotic technology aid in rehabilitation [13]. These advantages include providing more intensive and personalized patient rehabilitation activities and services, increasing therapy quantity and quality, and allowing all team members to set and manage specific work parameters to make the rehabilitation process as specific and effective for the patient as possible. Regarding workout difficulty, the support supplied, and device structure, robotic and technical gadgets can provide various options with varying levels of technology. This is now achievable because of recent technological breakthroughs. The field of telerobotics emphasizes remote control of semi-autonomous robots, mainly via wireless connections such as the Internet, allowing operators to communicate with their robots from a distance [14]. This intriguing realm of remote control, typically carried out by a human operator, leverages a sophisticated network of sensors and cameras attached to the robot to facilitate real-time navigation and task execution. Diverse in their functionality, telerobots can be classified into several notable categories. Industrial telerobots are typically employed in manufacturing or production environments with minimal human intervention [15]. These tireless mechanical workers are programmed to carry out specific tasks with machine precision and relentless stamina. In healthcare, medical telerobots transform patient care and surgical procedures, enabling doctors to perform complex operations remotely [16]. Similarly, space telerobots are revolutionizing space exploration; scientists on Earth can navigate Mars rovers, demonstrating a tremendous leap in interplanetary research [17]. One notable example is the humanoid robot, designed to imitate human appearance and behavior, enabling a more natural interaction within human-centric environments [18]. Venturing beyond terrestrial applications, telerobotics extends its reach into more hostile environments. Beneath the ocean's surface, underwater telerobots are deployed to perform deep-sea explorations, salvage missions, and infrastructure inspections, penetrating the vast aquatic frontiers that are otherwise inaccessible to humans [19]. In addition, telerobots play an important role in military and rescue missions, such as bomb disposal, surveillance, and search and rescue operations [20]. Regardless of their specific applications, all classes of telerobots share a unique trait: they leverage remote control capabilities to perform tasks in environments deemed unsafe or inefficient for humans, thus bridging the gap between human limitations and technological innovation [21].
Telerehabilitation robots are distinct from robots in other application fields due to their focus on human-centric design, safety, comfort, adaptability, and user-friendly interfaces. While remote control is an important aspect, it is not the sole determinant of a robot's suitability for medical rehabilitation. These robots must possess high precision and sensitivity, offer customized therapeutic options, and effectively communicate and exchange data with patients, medical personnel, and healthcare systems. This ensures that the robot can provide effective therapy while allowing healthcare professionals to monitor patient progress and make informed decisions. To be suitable for medical rehabilitation, these robots must have specific characteristics and adhere to strict criteria to ensure safety, efficacy, and compliance with medical regulations. The current deployment of graphical interfaces, ecological circumstances, and intellectually difficult tasks can physically and cognitively engage patients throughout robotic therapy. The cognitive challenge, automatic task difficulty change, and visual and auditory feedback can all benefit the patients in a rehabilitation session.
The number of people who require rehabilitation is expanding [10], whereas the available services are insufficient. Besides, travel/transportation poses significant problems to those with upper/lower limb dysfunction and is one of the major factors of early dropout from rehab sessions. More recently [22], [23], COVID-19 has exposed the critical need for home-based rehabilitation/tele-rehabilitation: during the pandemic, many patients with acute stroke (who require immediate rehabilitation exercise) avoided or lacked access to hospital and rehab centers. Indeed, data from 187 stroke rehab centers in 40 countries across six continents show that in the 3-month period of March to May 2020 (COVID-19 era), compared with the immediately preceding three months, there was a 19% decline in stroke admissions [24]. In addition, many post-stroke patients reported feeling abandoned at home with overwhelmed caregivers [24]. Clearly, many people with stroke-induced disabilities lack access to rehabilitation services, a problem that will exacerbate with aging, even as the pandemic abates. So, we urgently need an alternative to traditional rehab and self-care. Home-based rehab/telerehab assistance tools could help future crises. Advanced rehabilitation research focusing on robot-aided therapies (RAT), home-based rehabilitation, and telerehabilitation has the potential to mitigate this problem considerably. In robot-aided telerehabilitation, signals are transmitted to control the robot, while other signals the operator receives indicate whether or not the robot carried out the instructions. Currently, teleoperation is used in many domains and helps people in real-world applications by integrating perception, control, and learning technology [25]. The existing body of research on robot-assisted telerehabilitation, although promising, is still in its early stages and has notable limitations [26], [27], [28], [29]. Studies have primarily focused on simulated telerehabilitation environments [29], intermittent data transmission from rehab robotic systems to therapists without real-time remote control [26], and virtual reality (VR) applications that are either non-immersive and less engaging [2] or fully immersive yet potentially unsafe due to users being unaware of their surroundings [30]. Nowadays, North America has the largest market for telerobotics and teleoperations, and it is anticipated to continue until 2026 [31]. Therefore, it is time for rehabilitation professionals to use this technology to create telerehabilitation services for providing robot-assisted rehab.
In this research, we developed a novel telerehabilitation framework for delivering robot-assisted rehabilitation using an Industrial Internet of Things (IIoT), Augmented Reality (AR) platform developed by PTC Inc, rehabilitation robots, and a user-friendly interface. Communication between the ThingWorx server and the robotic system is established through RESTful API. The robot-patient site is transformed into a digital entity (digital twin), allowing the therapist to monitor the robot's state remotely. This system gives therapists the ability to remotely administer a wide variety of robot-assisted rehabilitation therapies, ranging from passive to resistance exercises. In addition, the therapist can simultaneously examine the robot's precision and the patient's condition while controlling the robot. Thus, the proposed method facilitates robot-assisted telerehabilitation with a physical and cognitive challenge automatically adjusting task difficulty and real-time visual and auditory feedback, thereby overcoming current limitations such as time and effort constraints, limited therapy sessions, and challenges associated with transportation and face-to-face therapy sessions. Following the completion of development, the proposed system is validated using a commercially available robot in addition to DMRbot and SREx, both of which were designed in the Biorobotics lab.
This manuscript's remaining sections are structured as follows: The state of the art in relevant disciplines is reviewed in Section II, and detailed development of the proposed system is shown in Section III. Section IV then shows the experimental setup for validation. Section V illustrates the experiment's results and discussion in detail. Finally, Section VI presents the conclusions with future directions for this work.

II. RELATED WORKS
Cloud-based telerehabilitation system uses cloud-based technology to provide remote rehabilitation services to patients, which allows healthcare practitioners to securely access patient data and provide treatment plans, exercise plans, and progress monitoring remotely [32], [33]. One of the most significant limitations of cloud-based telerehabilitation is the potential for privacy and security breaches. As with any technology, there is the potential for hackers to gain access to confidential patient records and other sensitive information. It is important to use strong encryption and secure data transmission protocols when using cloud-based telerehabilitation. Cloud-based telerehabilitation systems may not be accessible for individuals with limited technology access. Additionally, some clients may not be comfortable using the system due to a lack of experience.
Video-based Telerehabilitation uses two-way video communication to connect healthcare professionals and patients in real-time which allows healthcare professionals to provide instruction and guidance to patients, which can range from physical therapy exercises to advice on diet and lifestyle. [34] However, the therapist does not have control over the patient's environment or the equipment used, which can limit the effectiveness of the session. Also, both systems rely on a stable internet connection to operate. If the connection drops, the system may not be able to provide the necessary services. There has been a growing interest in the use of AR (augmented reality), VR (virtual reality), and MR (mixed reality) in telerehabilitation [35]. These technologies provide a more immersive and interactive experience for patients compared to traditional telerehabilitation methods. Studies have shown that the use of AR, VR, and MR technologies in telerehabilitation has improved patient outcomes, providing a more engaging and effective rehabilitation experience [36]. They have also been found to be useful tools for remote rehabilitation, expanding access to rehabilitation services and reducing the need for face-to-face therapy [37]. The use of AR, VR, and MR technologies in telerehabilitation has the potential to revolutionize the way rehabilitation is delivered, providing a more effective and engaging experience for patients while also expanding access to rehabilitation services [38]. Despite the benefits of AR, VR, and MR in telerehabilitation, there are several limitations that need to be addressed. Some of the limitations include cost, specialized equipment, and evidence-based practice.
Existing commercially available therapeutic devices, widely known as 'continuous passive motion' (CPM) devices (Kinetec-6080 [39], E3 [40], S3 [41], KS2 [42], HP2 [43], etc.), are unable to recreate the effective therapy exercises that therapists perform on patients and are mostly limited to providing passive therapy. With increasing demand, lack of therapists, inadequate therapeutic services, and transportation problems diminishing rehabilitation services, a major focus has been given to research on robot-aided and home-based rehabilitation. Among the existing UL Rehab Robotic Devices, the most well-known Float [44], EMU [45], CLEVERarm [46], [47], EXO-UL8 [48], [49] ANYexo [50], L-EXOS [51], [52], MGAexoskeleton [53], and Harmony [54], [55] (FDA pending) are still in the development phase in research labs and are limited to provide specific rehab therapy. Yet few UE RRDs that can provide UL rehabilitation are commercially available (exoskeleton type: ArmeoPower [56] (ARMin), ArmeoSpring (T-WREX) by Hocoma, and end-effector (EEF) type: InMotion ARM by BIONIK Labs [57], ReoGo by Motorika [58], Rubidium by iDRhA [59] ARMotus EMU by Fourier Intelligence [60], BURT by Barret Technology [61], and Motore by Humanware [62]), despite extensive R&D efforts. Furthermore, these are limited to use in clinical settings and do not cover the full spectrum of UL rehab therapies in a full range of UL motion, including joint-based therapy, Cartesian position-based therapy, passive, active-assisted, and resistive therapy. For example, InMotion Arm, Rubidium, and Motore are limited to therapy in planar movement (2D) but effective rehabilitation therapy requires 3D space arm movement. Other notable commercial RRDs include Tyromotion's AMADEO [63], limited to hand/finger motion, and DIEGO [64], a cable-driven lift mechanism only for compensating UL gravity. Such existing RRDs also lack the capacity for telerehabilitation and/or personalized home-based rehabilitation. Thus, more research on robot-assisted home-based telerehabilitation is needed to increase rehabilitation services to stroke survivors to improve their health and function and facilitate their reintegration into society and community.
Robot-assisted home-based telerehabilitation has shown great promise in improving the accessibility and efficiency of rehabilitation services; however, significant limitations remain [65]. Technical issues, such as limited bandwidth, latency, and connectivity issues, can impact therapy effectiveness [66]. Some patients may be denied access due to the high cost of robotic devices [65]. To operate and troubleshoot the devices, patients, and clinicians may require additional training [67]. Ensuring patient safety in a home-based setting can be challenging, especially when clinicians are not physically present [68]. Maintaining patient motivation and engagement can be difficult, impacting adherence to rehabilitation protocols [69]. Despite these constraints, research in robot-assisted home telerehabilitation is required. It can lead to improved technology, lower costs, safety concerns addressed, and the development of new protocols and interventions tailored to a home-based setting.
This study proposed a robot-aided telerehabilitation framework ( Figure 1) designed to provide remote care and rehabilitation services to people with neurological disabilities consisting of a robotic device, such as an exoskeleton or a robotic arm, connected to a computer system that is monitored and controlled by a remote therapist. The robotic device is used to deliver physical therapy in the comfort of their home. By making use of data collected from sensors through IIoT and cameras, the patient's progress can be monitored and the therapist can adjust the robot's parameters to adapt the treatment to the patient's needs through vuforia AR environment. PTC ThingWorx is an IIoT-ready platform that provides users with a set of tools and services to quickly build and deploy connected products and applications [70]. It is designed to connect with various devices and systems regardless of protocol, allowing users to quickly create applications and connected products. It allows users to quickly connect devices and create applications that can be deployed in the real world. Also, vuforia's rendering engine [71] is optimized for high-quality, high-performance visuals, allowing developers to create stunning augmented reality experiences. Robot-aided telerehabilitation can improve the quality of life of people with disabilities, as it allows them to access rehabilitation services without leaving their homes and without relying on a caregiver.

III. DEVELOPMENT A. CONCEPT
Stroke is one of the chronic diseases that contribute to the strain on healthcare systems as they strive to provide effective care for aging populations while keeping costs under control. To address this, there has been a rise in the use of technology in healthcare services. One such technology is robot-assisted telerehabilitation, which combines aspects of robot-assisted rehabilitation and telehealthcare to provide rehabilitation services from a distance. Multiple studies have proven clinically and statistically significant improvements in the functioning of the upper and lower extremities, while multiple feasibility studies have documented improvements in motion kinematics. It has also been demonstrated that telerehabilitation robotic delivery expands access to rehabilitation services.
Utilizing technological innovations to enhance healthcare services is one way to satisfy these expectations. One of the most recent additions to technologically mediated VOLUME 11, 2023 health care is robot-assisted telerehabilitation, which combines recognized aspects of robot-assisted rehabilitation and telehealthcare to provide rehabilitation services at a distance.
This study aims to enable remote robot-assisted rehabilitation to adopt conventional therapy, meaning one-to-one interaction between therapist and patient via a robot setting where the robot will be remotely monitored and controlled by the therapist to provide therapeutic motions. The proposed system will provide a holistic, cross-platform rehabilitation experience, where therapists can remotely administer a wide range of robot-aided rehab therapy ranging from passive to resistive exercises.
The proposed framework will incorporate IIoT, rehabilitation robots, and a user-friendly interface. To acquire the back-and-forth remote communication between the robotpatient environment and therapists, connectivity between the rehab robots and ThingWorx server will be established along with the communication between ThingWorx server.
To achieve this, the robot-patient side will be converted into a digital entity (digital twin) so that the data can be sent to the therapist via the internet, allowing the therapist to perceive the robot patients' local condition remotely. The virtual data will be obtained via IIoT and Vuforia by binding the IIoT data to the Vuforia model using 3D/2D label augmentation (e.g., joint angles, torques, end-effector position, and so on) and in the rehab robot's Digital Twin (DT). Secondly, the therapist will be able to command the robot to perform therapeutic motions for the patient. Thirdly, the therapist will be able to monitor whether the robot is providing accurate motions to the patient and visualize the patient's condition simultaneously. This concept will yield a novel telerehabilitation framework (hardware and software) for future randomized controlled trials and long-term studies of robot-assisted therapies in different settings, such as clinics and homes. Figure 1 shows the schematic of the framework. Fail-safe measures will be implemented in the framework to avoid the possible loss of information. If the internet connection is lost during a rehab exercise, the connection between Vuforia Studio with the robot will be lost, and the exercise session will stop for safety. After reconnecting to the Internet, the connection between Vuforia Studio and the rehab robot will be re-established automatically.

B. DEVELOPMENT OF DIGITAL TWIN 1) COMPONENTS a: DIGITAL TWIN
A digital twin is a replica of a physical object or process that can be used to simulate and monitor its structure, behavior, and performance in real-time. Real-time data is crucial to creating a true digital twin. Augmented reality (AR) can be used to bring a digital image of a physical object into the real environment for clients and servers to interact with. Accessing the digital twin requires a scanning code, which can be found in many formats on smartphones or other devices. PTC's Vuforia Studio [72] is a web-native tool used for creating industrial 2D/3D augmented reality applications. The platform optimizes CAD models, binds them to IIoT platforms, and provides AR Studio Experience Services. Vuforia Studio can improve equipment performance and analyze how it works using real-time data from the Internet of Things. The studio interface is user-friendly, with a drag-and-drop feature for creating new apps even without programming experience. Integrating different PTC products allows for easy data transfer and use in different environments. The advantages of Vuforia Studio make it a preferred tool for research. However, a disadvantage of using digital twins is the need for real-time data, which can be challenging to collect and maintain.

b: THINGWORX
ThingWorx [73] is an Industrial Internet of Things (IIoT) platform developed by PTC Inc. Its principle is to obtain and analyze data generated by connected systems continuously. ThingWorx provides a platform to structure and organize data 70178 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.  in a way that makes sense in industrial IoT to manage physical assets from digital systems. It offers connectivity, analytics, and tools for developing applications and Augmented Reality experiences, which makes achieving efficiencies simpler and quicker, lowering the risk of activities like in-home utilities, smart cities, healthcare, and manufacturing. ThingWorx enables networking and analytics and requires devices to be authenticated before sending or receiving data. The most used authentication method with ThingWorx is an application key assigned to a particular user with the same permissions as that user. The method for accessing ThingWorx requires making a connection with the ThingWorx instance and passing the application key with ThingWorx URL to accept the request from the client. One advantage of ThingWorx is its ability to organize and structure data in a meaningful way, making it easier to manage physical assets from digital systems. However, one disadvantage is the need for authentication, which may pose a challenge for users.

c: DIGITAL TWIN MODELING PROCESS
The process of developing a new model in Vuforia studio begins with selecting a unique project name and experience service URL. Due to the requirement for ThingWorx server data for AR application integration, the experience service URL is identical to the ThingWorx server URL. After displaying the CAD model, the development environment is now displayed, and coordinates and rotations can be modified. By dragging the ''Model Item'' widget onto each connection, the robot model assembly can be converted to a hierarchical model. The component that is the bottom-most parent model in the robot hierarchy can be used as a starting point, as shown in Figure 2. Therefore, we dragged ''Model Item'' onto the base and edited the ''Component Occurrence'' by deleting the last values.
It can be seen that the base is the child model of the robot (model-1) parent model. The process must be done for each joint (link), starting with the ones at the bottom of the robot and moving up the hierarchy. Furthermore, we bound the IIoT data to the CAD model, which is shown using the 3D label Augmentation (e.g., joint angles, torques, end-effector position, etc.) and in the Digital Twin (DT) of the rehab robot, as shown in Figure 2. ThingWorx data can be bound to both 3D and 2D widgets for rehab robots (xArm 5, SREx, and DMRbot). The 3D gauge widget is associated with the particular property when it is imported from ThingWorx to Vuforia Studio by clicking ''+'' under ''External Data.'' By clicking the ''Publish'' button, Vuforia Studio projects can be accessed via the experience service URL, and users can access them from any mobile device. Figure 3 shows the flow chart to make a digital twin of the robot. The process is broken down into three main steps: (i) nest the assembly in Creo, (ii) create the model hierarchy in Vuforia Studio, and (iii) develop an application for Vuforia Studio. The development of a digital twin began with assembling the rehab robot's base to the endeffector. The 3D CAD model of the robot was assembled using PTC Creo, and PTC Vuforia Studio was used to convert the CAD model into a digital twin.

C. TELEREHABILITATION FRAMEWORK DESIGN
The telerehabilitation framework is developed using PTC ThingWorx, PTC Vuforia Studio, and PTC Vuforia View as software [74]. The digital twin acts as a replica of the physical robot whenever it transmits actual robot data. The platform for the IIoT can be set up such that it will react to changes in the data. Both parts of the system communicate with each other through the internet (shown in Figure 1). The user or operator modifies the state of the robot controller by interacting with an AR-based robot ''controller'' in a Vuforia Studio application. This controller's state change is sent from the Vuforia Studio program to the ThingWorx robot -virtual Thing, then from the ThingWorx robot to a PC that operates the rehab robot. The rehab robot is connected to a Client PC (Windows Computer), which provides various telerehabilitation exercises (trajectories) to the robot in response to orders VOLUME 11, 2023   received from the IIoT platform. ThingWorx, Experience Service, Vuforia Studio, and Vuforia View are all part of the IIoT platform. ThingWorx serves as a bridge for processing IIoT data, as shown in Figure 4 transmission between the robotic system and the Vuforia Studio application in the IIoT platform (Digital Twin of a robot) field. ThingWorx serversafe HTTPS protocol is used for bidirectional communication between the client PC and the ThingWorx server. Client PC delivers an HTTPS request to the ThingWorx server, and the ThingWorx server responds with an HTTPS response. The 'Experience Service,' which is part of the IIoT platform, provides data on rehabilitation therapy.
In this study, we used Vuforia Studio to create a digital twin of a rehab robot with a graphical user interface (GUI) as shown in Figure 5 for telerehabilitation therapy. Vuforia Studio then publishes the generated digital twin to the Experience Service URL. We utilized the Vuforia View app to access the published GUI from the Windows/Android/iOS platforms. The Vuforia View app is linked to the ThingWorx experience services URL. When engaging with the digital twin of a robot GUI, it communicates IIoT data to the robotic system. The benefit of adopting the ThingWorx IIoT cloudbased platform is that user therapists may manipulate the real robot by connecting the physical joystick to the ThingWorx server. At the same time, the associated robot's motions are visualized on the AR digital twin of robots within the Vuforia View app. This study develops several telerehabilitation control modes, including pre-defined therapy / passive therapy, joint-control mode, and joystick-control mode. In addition, when a session begins, the therapist and patients join the Video calling room using Microsoft Teams for improved communication.

IV. EXPERIMENTAL SETUP
The experimental configuration of the telerehabilitation system involves two sides of arrangements: The operator/ therapist side shown in Figure 6a and the participant side that includes communication devices and human-robot interaction as shown in Figure 6b. Here the operator and the participant can communicate over a video conference using Microsoft Teams. An operator can see the robot's motion and kinematic parameters (e.g. position, velocity, human-robot interaction forces, actuator current data, robot joint torque, and so on) while providing a variety of passive rehab rehabilitation exercises utilizing this video session. The subject sits in a position for upper-limb rehab activities while grasping the robot handles with their palm, as shown in Figure 7. The operator initiated the pre-planned movement by pressing the button on the Vuforia view user interface. As shown in Figure 8, the robotic arm moved the participants' limbs around the 3D plane during this exercise.
To provide a variety of passive rehab exercises, the therapy modes must be modified; to do so, an operator utilizes the Vuforia View app for iPad, as seen in Figure 5 (made using Vuforia Studio 3D/2D widgets). An operator employs the developed digital twin of an end-effector robot to follow the robot's movements with low latency and monitor the participant's input while interacting with the robot.
We utilized a camera, monitor, and Microsoft Teams to connect with operators during therapy sessions to record upper arm movements during rehabilitation exercises. Three healthy individuals (25-28 years old, 1.60-1.80 m tall, 70-80 kg) performed passive exercises to test the developed telerehabilitation systems. We created a use case scenario for telerehabilitation by separating the participant's and 70180 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.   operator's systems. In our tests, the robotic system and IIoT platform experienced negligible (0.16s) latency.

A. PROTOCOL
Three different serial-type robots were utilized to test and validate our developed robot-assisted telerehabilitation framework, namely xARM 5, SREx, and DMRbot (see Figure 7). Among these robots, the xARM robot is commercially available in the market for robotic research, while the SREx and DMRbot were designed and developed in the Biorobotics Lab at UWM specifically for providing upper limb rehabilitation exercises. The rationale for utilizing these VOLUME 11, 2023 70181 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.
robots is to ensure that such developed tele-platform can integrate with different robot-assisted rehabilitation systems comprising versatile communication, actuation, kinematic, and dynamic modalities.

B. REHAB ROBOTS
In addition to the dynamic's parameters of the robot, there are additional parameters with significant characteristics that are particularly important when considering robots specifically designed for medical rehabilitation. Medical rehabilitation robots require unique qualities to assure safety, efficacy, and patient comfort. Compliance control, human-robot interaction forces, flexibility, range of motion, user interface and feedback, and safety devices are all important criteria for rehabilitation robots. These precise criteria and qualities guarantee that rehabilitation robots fit the specific needs of medical rehabilitation, giving patients targeted, safe, and effective support. Our proposed method incorporates all the contents/characteristics of the above-mentioned physical data.

1) xArm 5
The xArm-5 is an end-effector type five-degrees of freedom (DoF) robotic manipulator developed by UFACTORY (see Figure 7a). It is equipped with brushless servos, harmonic reducers, a 6-axis force-torque (F/T) sensor, and position repeatability of ± 1.0 × 10-4 m. It uses Modbus-TCP and Modbus-RTU communication protocols for the robotic arm and end-effector, respectively. The total workspace area of xArm 5 is 0.7 m [75]. In our study, xArm's Python SDK is used to control the xArm 5 for advanced functionality. The workspace of xArm-5 robot has a maximum reach of 700 mm.

2) SMART ROBOTIC EXOSKELETON (SREx)
The SREx [76] is an exoskeleton robot (see Figure 7b). The SREx was designed with 7 degrees of freedom (DoF) to imitate basic upper-limb biomechanics in motion. The robot has 3 DoF shoulder motion (vertical flexion/extension, horizontal adduction/abduction, and internal/external rotation), a 2 DoF elbow and forearm motion (elbow flexion/extension and forearm pronation/supination), and a 2 DoF wrist motion support components (wrist flexion/extension and radial/ulnar deviation). Brushless DC motors with harmonic drives were used to actuate the SREx. SREx can produce isolated and composite joint motions to the upper arm as per robot operator input variables. The wrist support part contains a palm module where the user rests their palm and grabs the handle. The palm module is equipped with F/T sensor which collects human-robot interaction forces and torques.

3) DESKTOP-MOUNTED REHABILITATION ROBOT (DMRbot)
The DMRbot [77] is a 3DoFs, portable (benchtop) endeffector type therapeutic robot (see Figure 7c). This robot consists of a base, three actuators, associated links, and one force sensor. The link lengths are selected such that the robot can cover the nominal human arm therapeutic workspace. This robot has been designed as a minimum viable robot to cover 3D space motions of the human upper arm. The distal end of the robot's link contains the 6-axis F/T sensor with the handle. The person who controls the DMRbot can adjust the speed and set up motion constraints.

V. RESULTS AND DISCUSSION
The proposed telerehabilitation framework has been evaluated using three different robots an end-effector(DMRbot), an exoskeleton(SREx), and a commercially available endeffector robot(xArm 5). DMRbot and SREx were developed in the Biorobotics lab and all the robots work with a different communication protocol. The validation of the proposed system is considered in light of standard criteria for 70182 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. telerehabilitation. The data collected from monitoring the robotic arm's movement is presented to show each joint's joint angles, torques, and velocities/speeds, as well as the force exerted by the participant's wrist and upper limb along the x-axis, y-axis, and z-axis. Note that, to assess the patient's condition (such as spasticity) [78] and rehab progress, the therapist needs to measure the patient's active and/or passive range of motion [79], the velocity of movement, and resistive force (human-robot interactive force) [80]. On the other hand, motor current and robot joint torque are measured to provide feedback input to the robot controller to generate the actuating control commands for robot motion to follow a desired exercise [81].
The passive mode of telerehabilitation exercises has been evaluated with the rehab robots while remotely controlling, monitoring, and running the experiments. Passive mode targets the massed and repetitive practice of the exercises. In this mode, the rehab robots move the participant's arm in 2D/3D space to increase the range of motion and reduce spasticity. Exercising this way can improve the effectiveness of therapy and speed up recovery. These are the robot's joint-based trajectories, which are generated by simulating the motion of the participant's arm. The operator can change different exercise methods/modes during experiments using the Vuforia View user interface, as illustrated in Figure 5. Each kind of exercise has its own user interface.
In Figure 8a the operator can monitor participants' upperlimb motion in the Vuforia studio application while using DMRbot to take multi-joint exercises. A participant is seated on the chair, holding the robot's end-effector (handle) and facing the MS team's screen to see the exercise motion. In this exercises, Joint-1 moves from its initial position 0 • to −30 • , then −30 • to 15 • and finally returns to 0 • . Joint-2 moves from its initial position 0 • to −35 • , and stays at it for around 8 sec, then moves from 35 • to −35 • and finally returns to its initial position. At last, joint-3 moves from its initial position 0 • to 80 • and finally returns to its initial position. The experiment uses three repetitions.
To provide forearm supination-pronation exercise shown in Figure 8b During the experiment with SREx, the operator provided Elbow flexion-extension motion to a healthy participant remotely via Vuforia View interface (see Figure 5c) where the operator clicks and drags the elbow motion-associated slider (4th slider on the top left corner)from a remote PC. Leveraging the telerehabilitation system the SREx provided passive elbow movements according to the operator's input. The elbow joint motions can be seen in Figure 9, where 1st row shows the joint-associated angle that ranges from 30 • to 110 • , During the elbow motion peak velocity reaches around 27 • /s. The kinematic parameters which include human-robot interaction forces and torques can be seen from the 3rd and 4th rows. These kinetic and kinematic values are transmitted from the SREx system to the thingWorx server and subsequently to the Vuforia app view. The operator can observe these values while providing joint-specific therapy to the remote participant and adjust the therapeutic motion accordingly.
The proposed system can provide interactive one-on-one real-time telerehabilitation exercises as well as passive telerehabilitation exercises that use predefined trajectory rehab exercises. Teleoperation is used to control the movement of the shoulder, elbow, and forearm joints of the people taking part in the experiments. The main benefit of this research is that our proposed telerehabilitation framework is tested using a variety of rehab robots currently in use for therapy for UL rehabilitation. The robot Data were collected while monitoring the AR robotic arm's movement. Based on the data collected, Figure 10, and Figure 11 show each joint's joint angles, torques, and velocities/speeds. The PID control's switching frequencies cause speed fluctuations. The compatibility of the proposed system with different communication protocols was tested in the experiments described above. The results showed that the system was able to communicate with all three robots, and the rehabilitation exercises were successfully executed by the robots. This demonstrated the effectiveness of the communication interface developed for the system.
RESTful APIs offer a set of well-defined methods for handling and reporting errors. By leveraging HTTP status codes and error messages, the proposed system can effectively identify and address issues related to the indexing problem, such as data mismatches or communication delays between master and slave. The combination of HTTP requests and RESTful APIs with technologies such as Web-Socket or Server-Sent Events enables real-time updates and communication between the master and slave. This feature helps mitigate time-based indexing problems and ensures the system remains responsive and up-to-date. The considered rehab robot in the proposed system is equipped with a local controller (loop time: 5 ms) [82] that employs telerehabilitation to update its position values. Our proposed system enables the robot to receive real-time updates on its position from a remote location with a latency of 160 ms. In the event of a loss of communication between the robot and the remote operator, the robot's local controller is designed to 70184 VOLUME 11, 2023 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. persist the last available position values. Literature shows that the delays between 100 ms and 250 ms did not make a significant difference from no delay in the performance of a robotic coronary telestenting system [83]. This suggests that the 160 ms delay falls within the range of tolerable latency for teleoperated medical systems. The 160ms latency observed in the proposed telerehabilitation system is well below what we observed in the above-mentioned robotic coronary telestenting system, and does not impact overall performance.

VI. CONCLUSION
We developed a novel framework to provide various passive telerehabilitation exercises to persons with upper limb impairment. The proposed system is built using many cuttingedge technologies, including the PTC ThingWorx IIoT platform, the Vuforia Studio Augmented Reality platform, the PTC ThingWorx experience services, and the Vuforia View app for iPad. The key advantage of this study is the use of different rehab robots used in therapy for UL rehabilitation to evaluate our proposed telerehabilitation framework. Furthermore, this research makes use of the digital twin structure, made possible by Vuforia studio, to see real robot movements occurring in remote locations. The experimental results indicated that the proposed IIoT and augmented reality-based framework for human-robot communication can be utilized for various telerehabilitation applications. The future heading of this research could involve further development and testing of the proposed telerehabilitation system in clinical settings, including patient trials and active-assisted rehabilitative therapy. Additionally, the system's effectiveness and usability could be compared to traditional in-person therapy, and any differences or advantages could be evaluated. Furthermore, the system's integration with other technologies, such as mixed-reality along with machine-learning algorithms, could be explored to enhance the user experience and improve therapeutic outcomes. Overall, the proposed telerehabilitation system has the potential to revolutionize the field of rehabilitation by offering a flexible, accessible, and effective alternative to traditional therapy.