Title: Development of an Assistive Device for Upper Limb Rehabilitation

Therapy

Abstract:
The purpose of this paper is to design, construct and performance test of an exoskeleton system for upper limb
rehabilitation therapy. The device aims to aid individuals with neurological impairments in restoring motor
function effectively. The exoskeleton system integrates physical and mirror therapy techniques to enhance
motor function recovery. It features a user-friendly interface allowing easy transition between therapy modes.
Patients can switch between physical and mirror therapy using an Android app. During physical therapy
sessions, the exoskeleton repetitively moves the elbow within predefined ranges. Mirror therapy involves
mirroring the movements of the healthy limb onto the impaired one. Patient safety is prioritized during
rehabilitation. The system incorporates a forced stop option enabling immediate halting if patients experience
discomfort, fatigue, or any adverse effects. The device utilizes a linear actuator to provide movement to the
elbow joint, applying different force levels at various angles for muscle conditioning. Performance metrics such
as torque output, angular velocity, response time, and positioning accuracy are evaluated and compared against
best and worst-case scenarios. Results indicate promising performance, with the device demonstrating
significant torque output (7.09 Nm max), moderate angular velocity (11 degrees/s), quick response time (1.2 s),
and minimal positioning error (2.8 degrees). This suggests its potential effectiveness in aiding upper limb
rehabilitation. Further validation through clinical trials is necessary to assess its efficacy in improving therapy
outcomes and promoting patient recovery.
 1 INTRODUCTION:
The prevalence of conditions such as stroke, peripheral neuropathy, multiple sclerosis, Parkinson’s disease,
Guillain-Barré Syndrome and amyotrophic lateral sclerosis (ALS) has led to an increasing demand for effective
upper limb rehabilitation therapies. Stroke is widely recognized as a significant global health concern due to its
serious and prevalent disabling effects worldwide. According to, Global Stroke Fact Sheet 2022, annually, the
world experiences more than 12.2 million new cases of stroke, with over 16% affecting individuals aged 15-49
years and over 62% occurring in those under 70 years of age. Recent data indicates that approximately 650
million individuals globally are aged 60 years or older, with projections forecasting a surge to 2 billion by the
year 2050[1]. Stroke damages cortical tissue, disrupting the generation and integration of neural commands[2].
As a result, many stroke survivors experience bodily malfunctions that hinder their ability to perform activities
of daily living (ADLs). Around 80% of stroke survivors experience motor impairment, which usually affects
movement on one side of the body, including the face, arm, and leg[3]. Issues with upper limb movement,
involving the arm, hand, or fingers, are often long-lasting and disabling[4].

Figure 1.1 Upper Limb Exoskeleton used for rehabilitation therapy[5]

Only about half of stroke survivors with an initially paralyzed upper limb regain some useful function after six
months[6]. Furthermore, among those with initial arm impairment, about half still face challenges with arm
function four years later[7]. Despite the remarkable progress in health sciences, challenges related to
neuromotor dysfunction leading to limitations in human limb movement persist worldwide[8]. To regain
function in a limb affected by stroke, it's essential to engage in rigorous, repetitive training sessions focusing on
maximizing voluntary motor effort while minimizing compensatory movements, aiming for muscular
coordination that closely resembles normal patterns[8]. Stroke patients require professional support for
rehabilitation activities. Therefore, upper limb rehabilitation after a stroke is often a complex process that
necessitates collaboration among the patient, caregivers, and the rehabilitation team[9]. Again the current
number of therapists is insufficient to meet the growing needs of stroke patients, particularly in rural areas.
Consequently, robot-based rehabilitation devices are being explored as a solution to address the rising number
of stroke patients and to supplement the efforts of therapists. Assistive technologies have emerged as a potential
solution to extend the reach and effectiveness of therapeutic interventions for upper limb rehabilitation in
response to the increasing challenges faced by individuals with upper limb impairments.
Manuel et al. [10] present a wearable robotic exoskeleton prototype featuring autonomous Artificial
Intelligence-based control, processing, and safety algorithms. This one-degree-of-freedom technology facilitates
elbow joint flexion and extension, offering various operation modes such as autonomous, remote control, and
leader-follower modes. The compact design, enabled by selected materials, supports personalized training
regimens crucial for effective motor rehabilitation. Laboratory tests validate minimal latency and respectable
precision, indicating the correct operation of integrated technologies. This research contributes to the
advancement of upper limb rehabilitation through the development of a small, wearable robotic exoskeleton
prototype.
Karadeniz et al. [11] describe the design, control, and testing of a 4-DOF upper limb exoskeletal robot capable
of performing four therapeutic exercises: passive, active assistive, isotonic, and isometric. A low-cost
electromyograph was integrated to measure and control real-time muscle activation during therapy. The system
supports rehabilitation for shoulder and elbow joints. A PID controller was implemented for position-controlled
exercises. Simulation and experimental results demonstrate the system's effectiveness in performing passive
exercises and its suitability for other therapeutic exercises.
Yi Liu et al.[12] developed a portable exoskeleton for elbow rehabilitation, designed to accommodate the
natural joint range of motion with adjustable actuated stiffness. The exoskeleton features five passive degrees of
freedom to support natural joint movement. An integrated variable stiffness actuator (VSA) enables independent
stiffness adjustment by altering the pivot position. This device is intended for home-based rehabilitation.
Tiboni et al.[13] propose a single-degree-of-freedom exoskeleton for the elbow–forearm region, designed to
provide effective rehabilitation therapy for individuals with upper limb motion disabilities. The control system
uses sEMG signals to drive joint movement and manage the robotic arm. An active control algorithm based on
sEMG force signals adjusts the exoskeleton's exerted force during rehabilitation. The control approach is
validated through experiments on healthy subjects simulating arm impairment.
D. Copaci et al. [14] introduce a prototype exoskeleton for evaluating and rehabilitating the elbow joint,
focusing on flexion extension and pronation-supination movements. The novelty lies in the use of bioinspired
actuators based on Shape Memory Alloys (SMAs), a novel approach in upper limb rehabilitation exoskeletons.
SMAs are metallic alloys that can recover their original shape after deformation when heated above the
transformation temperature. The prototype features two degrees of freedom (DOF) for elbow joint movement,
with SMAs enabling antagonistic movements. The system is wearable and portable, suitable for home-based
rehabilitation.
 2 MATERIALS AND METHODS:
A new exoskeleton system, composed of an iron metal structure, has been developed for upper limb
rehabilitation in individuals with impairments. This research combines physical and mirror therapy techniques
to enhance motor function recovery. The primary objective is the creation of an easy-to-use interface for
seamless therapy mode switching. To accomplish this, a control system has been developed, allowing patients to
switch between therapy modes using an Android app via Bluetooth. During physical therapy, the exoskeleton
repetitively moves the elbow within set ranges, facilitating muscle strength and joint flexibility enhancement.
Mirror therapy involves replicating movements of the healthy limb on the impaired one, recognized for its
ability to improve neural plasticity and motor learning. To ensure patient safety, the device features a stop
option for immediate therapy cessation in case of discomfort or fatigue. It utilizes a linear actuator to control
movement and apply various force levels to the elbow joint at different angles, crucial for muscle conditioning.
For simplicity, the methodology is divided into 4 steps as follows:

Figure 2.2 Work Flow Diagram

2.1 Exoskeleton Design:

 Linear
Actuator

Lever
Mechanism

Figure 2.3 SolidWorks design of the exoskeleton (dimensions in mm)

A single-degree-of-freedom (1 DOF) upper limb exoskeleton as shown in Figure 2.2 is designed to assist with
elbow flexion and extension. Linear actuator translates electrical signals into linear force, driving the movement
of the exoskeleton. The lever mechanism converts the linear motion of the actuator into rotational motion at the
elbow joint. As the actuator extends, it pushes on the lever, causing it to rotate around a fixed hinge. This hinge
connects the lever to the rest of the exoskeleton frame. The rotation of the lever appears to drive the flexion and
extension of an elbow joint.

Figure 2.4 Actual Connection Circuit

 Figure 2.5 Electrical Connection Diagram

2.2 Components and Connection:

Arduino Uno is a microcontroller board that serves as the main control unit for the device. It receives
sensor (potentiometer) data, processes it, and sends control signals to the motor driver. Figure 2.6 shows
the connection in different pins of Arduino in our device. The Arduino's GND pin connects to the GND
pin of the breadboard, and its 5V pin connects to the 5V pin of the breadboard for power supply to the
breadboard. The wiper (middle) pin of each potentiometer connects to an analog input pin of the
Arduino ((A0 for feedback potentiometer attached link in active limb (mirror therapy only)) and (A1 for
feedback potentiometer attached to the exoskeleton). The remaining pins of each potentiometer connect
to the 5V and GND pins of the breadboard.

Figure 2.6 Connection with different pins of Arduino UNO

The Bluetooth module connects to specific digital pins of the Arduino for serial communication (4 & 5
digital pins). The Arduino controls the motor driver using several digital pins. The BTS7960 motor
driver interfaces between the Arduino and the linear actuator. It controls the motor’s direction (forward
or backward) and speed (via PWM control). The motor driver ensures precise movement of the linear
actuator during rehabilitation exercises.
 Figure 2.8 Connection of feedback
potentiometer attached to the link at the
active limb.

Figure 2.7 Connection of feedback

potentiometer attached to the
exoskeleton.

Potentiometers shown in Figure 2.7 and Figure 2.8 are used to measure the angle of movement at the elbow
joint. As the elbow bends, the resistance of the potentiometers changes, providing an analog voltage signal to
the Arduino. As shown in Figure 2.7, this potentiometer is attached for position feedback of the exoskeleton,
which is first inserted into a link and then attached to the exoskeleton. Another potentiometer is used at the
active limb for real-time position feedback (Only for mirror therapy) of the active limb and sends data to
Arduino to replicate the desired position of the exoskeleton in this way mirroring the motion to the impired
limb.
In the project, the linear actuator, a pivotal component, harnesses electrical energy to produce linear motion,
which is then ingeniously transformed into circular motion for elbow flexion and extension through a lever
mechanism. Acting as a vital intermediary, the BTS7960 Motor Driver facilitates seamless communication
between the Arduino and the linear actuator, receiving control signals to regulate motor current, thereby
controlling speed and direction. The Bluetooth Module, a wireless marvel, establishes communication between
the Arduino and a smartphone application, enabling convenient mode selection and emergency stop
functionalities. Facilitating the seamless integration of electronic components, the Breadboard serves as a
temporary platform for connection sans soldering. Lastly, the Power Supply stands as the stalwart provider of
requisite voltage and current to energize the entire system, ensuring its smooth and uninterrupted operation.

2.3 Android App Interface for the Device:

 Figure 2.9 Connecting the assistive device with the apps

The app features an intuitive home screen with clear mode selection options, allowing users to choose between
"Physical Therapy" or "Mirror Therapy" modes complemented by an emergency stop feature for user comfort
and safety. Utilizing Bluetooth communication, the app seamlessly pairs with the Arduino Uno, providing real-
time status updates and bidirectional data transmission, ensuring efficient therapy sessions. Additionally, the app
requests necessary Bluetooth permissions, ensuring user privacy and compliance with data handling regulations.
Overall, these design features aim to enhance user engagement, effectiveness, and safety throughout the therapy
process.
Additionally, Motivational quotes have been intentionally incorporated to encourage and inspire users
throughout their rehabilitation journey. Users will be helped to stay positive and focused on their goals by these
messages[15],[16].

2.4 Description of the System:

The flow chart of the control system depicts the operation of the assistive device, offering both physical therapy
and mirror therapy modes.
The initial step involves connecting the device with the apps by Bluetooth and the patient selecting the therapy
mode via an Android application. This reflects a patient-centered approach, allowing individuals to actively
participate in rehabilitation. Neurologically, providing patients with choices and control over their therapy can
enhance motivation and engagement, which are essential factors in neurorehabilitation[17].
 Figure 2.10 Workflow diagram of the Control System. Figure 2.11 App interface of the device

In the physical therapy mode the device assists with repetitive extension and flexion movements of the elbow,
executing the right side of the workflow diagram as shown in Figure 2.10. Repetitive movements are crucial for
promoting neuroplasticity, the brain's ability to reorganize and form new neural connections[18]. These
movements stimulate motor pathways, helping to restore function and mobility in individuals with upper limb
impairments[19]. Additionally, setting the desired range of motion aligns with the principle of task-specific
training, focusing on targeted movements relevant to daily activities.
The mirror therapy mode replicates the active limb movement to the impaired limb, executing the left side of
the workflow diagram as shown in Figure 2.10. Mirror therapy involves connecting the exoskeleton to the
impaired limb and using a feedback link to mirror the movement of the active limb. This approach capitalizes
on the brain's mirror neuron system, which activates when observing and imitating movements[20]. By
providing visual feedback and mirroring movements, mirror therapy can help stimulate neural pathways and
improve motor function in the impaired limb.
The device's linear actuator executes control signals to move the impaired limb during physical therapy or
mirror the movement of the active limb during mirror therapy. The precise control of movement provided by the
device ensures accurate stimulation of motor pathways and promotes consistent, task-specific training.
Consistent, repetitive movement is essential for reinforcing neural connections and facilitating motor
recovery[18].
The option for the patient to stop all device movement immediately ensures safety and comfort during therapy
sessions. Avoiding overexertion or discomfort is crucial to preventing fatigue-related impairments and
optimizing neurorehabilitation outcomes[21].
The integration of technological advancements with principles of neurorehabilitation, emphasizes the
importance of patient engagement, task specificity, and safety in promoting motor recovery and functional
improvement in individuals with upper limb impairments.
Some snapshots during physical therapy and mirror therapy are given in Table 1 :
Table 1 Snapshots during therapy session.
Physical therapy Limb Position Mirror Therapy

At the extended limb

position

At Middle Position

At Flexed Position

3 RESULT AND DISCUSSION:

In this section, the tests conducted on the exoskeleton system have been examined. The force exerted at
different elbow positions, the movement of the elbow over time during therapy, its response time to commands,
and its accuracy in positioning have been assessed. These tests have played a crucial role in determining the
functionality and effectiveness of our device in assisting individuals with arm impairments in their rehabilitation
journey. Paramount goals, including ensuring safety, smooth operation, and improvement in arm movements,
have been addressed. In the subsequent sections, findings from each test will be elaborated on, and their
implications for our exoskeleton system and its users undergoing therapy will be discussed.

3.1 Torque Vs. Angle Relationship:

 Figure 3.12 Torque Vs. Angle Relationship

Figure 3.1 demonstrates how much is applied to the elbow at different angles. When the elbow is fully
straight (0 degrees), the device applies a strong force of 7.09 Nm to start movement. As the elbow bends
(flexes) to 65 degrees, the force decreases to 1.48 Nm. This gradual decrease in force as the elbow bends shows
how the device can adjust to the patient's movements. This ability is crucial in therapy because it allows the
device to challenge the muscles without causing strain or injury. The torque-angle diagram also demonstrates
how the device helps muscles adapt by providing different levels of resistance throughout elbow movement.
This is important for building muscle strength, improving joint stability, and enhancing mobility during
rehabilitation. In conclusion, the torque-angle diagram proves that the rehabilitation device is effective in
improving therapy outcomes.
3.2 Angle Vs. Time Relationship:

Figure 3.13 Angle Vs. Time Relationship

Figure 3.2 gives us useful information about how muscles and joints move over time, especially
focusing on the elbow joint's motion range. Each point on the graph shows the elbow angle at different times
during a therapy session. Starting from time 0, which is when the elbow is fully straight (0 degrees), the joint
gradually bends (flexes) to about 65 degrees as the session progresses. The steady increase in angle over time
shows that the elbow moves in a controlled and gradual manner, indicating the patient's ability to use their
muscles around the joint. This suggests that the therapy session follows a structured plan, allowing muscles and
joints to gradually get used to movement without risking strain or injury. Also, the graph gives us insights into
how muscles activate and relax over time, which is crucial for effective therapy. As the elbow bends towards 65
degrees, the rate at which the angle changes slows down, indicating that the movement is slowing as the joint
reaches its maximum bend. This slowing-down phase is important for preventing sudden stress on the joint and
ensuring smooth movements, which helps maintain joint health and function.
3.3 Performance Summary of the Rehabilitation Device
Below table is a summary of the performance metrics for our developed rehabilitation device. These metrics
provide insights into the device's capabilities and effectiveness in assisting upper limb rehabilitation therapy.
Each criterion is evaluated and compared against best and worst-case scenarios, offering a comprehensive
understanding of the device's performance across different parameters.

Summary Table of the Rehabilitation Therapy Device

Most favorable Least favorable

Criterion This Device Additional Info
outcome outcome
Torque 7.09 Nm (Max) 21 Nm[22] 3 Nm[23] Higher is Better
Angular Velocity 11 Degree/s - - -
Response Time 1.2 s (avg) - - Lower is better
Positioning
2.8 degree (avg) 0.56 degree[24] 5.6 degree[25] Lower is better
Error
Cost $82 - - Features Matters

Conclusion:
Developing an assistive device for upper limb rehabilitation therapy marks a big step forward in meeting the
diverse needs of patients recovering from upper limb injuries or neurological conditions. Our goal was to blend
physical therapy and mirror therapy into one device, offering a well-rounded solution that fits the various
aspects of recovery. Introducing two therapy modes—physical therapy for elbow extension and flexion, and
mirror therapy for unilateral stroke cases—has given patients flexible options for their rehabilitation journey.
Being able to switch between these modes using a smartphone app makes it easier for patients to stay engaged
and tailor their therapy sessions. Our device's user-friendly interface makes selecting modes a breeze,
empowering both patients and therapists to customize therapy sessions to suit individual preferences and needs.
We've also made sure to prioritize patient safety and comfort by including features like a forced stop option for
emergencies. Through performance testing, including torque vs. angle and angle vs. time plots, we've found
maximum torque of 7.09 Nm at the initiation which is well enough to bend the impaired limb at the elbow and
rotational speed of 11 degree/s.
 REFERENCES
[1] S. Mendis, “Stroke Disability and Rehabilitation of Stroke: World Health Organization Perspective,” Int.
J. Stroke, vol. 8, no. 1, pp. 3–4, Jan. 2013, doi: 10.1111/j.1747-4949.2012.00969.x.
[2] S. Mohamaddan et al., “Development of Upper Limb Rehabilitation Robot Device for Home Setting,”
Procedia Comput. Sci., vol. 76, pp. 376–380, 2015, doi: 10.1016/j.procs.2015.12.312.
[3] P. Langhorne, F. Coupar, and A. Pollock, “Motor recovery after stroke: a systematic review,” Lancet
Neurol., vol. 8, no. 8, pp. 741–754, Aug. 2009, doi: 10.1016/S1474-4422(09)70150-4.
[4] S.-M. Lai, S. Studenski, P. W. Duncan, and S. Perera, “Persisting Consequences of Stroke Measured by
the Stroke Impact Scale,” Stroke, vol. 33, no. 7, pp. 1840–1844, Jul. 2002, doi:
10.1161/01.STR.0000019289.15440.F2.
[5] M. A. Gull et al., “A 4-DOF Upper Limb Exoskeleton for Physical Assistance: Design, Modeling, Control
and Performance Evaluation,” Appl. Sci., vol. 11, no. 13, p. 5865, Jun. 2021, doi: 10.3390/app11135865.
[6] G. Kwakkel, B. J. Kollen, J. Van Der Grond, and A. J. H. Prevo, “Probability of Regaining Dexterity in
the Flaccid Upper Limb: Impact of Severity of Paresis and Time Since Onset in Acute Stroke,” Stroke, vol.
34, no. 9, pp. 2181–2186, Sep. 2003, doi: 10.1161/01.STR.0000087172.16305.CD.
[7] J. G. Broeks, G. J. Lankhorst, K. Rumping, and A. J. H. Prevo, “The long-term outcome of arm function
after stroke: results of a follow-up study,” Disabil. Rehabil., vol. 21, no. 8, pp. 357–364, Jan. 1999, doi:
10.1080/096382899297459.
[8] J. E. Harris and J. J. Eng, “Strength Training Improves Upper-Limb Function in Individuals With Stroke:
A Meta-Analysis,” Stroke, vol. 41, no. 1, pp. 136–140, Jan. 2010, doi:
10.1161/STROKEAHA.109.567438.
[9] A. Pollock et al., “Interventions for improving upper limb function after stroke,” Cochrane Database Syst.
Rev., vol. 2014, no. 11, Nov. 2014, doi: 10.1002/14651858.CD010820.pub2.
[10] M. A. Vélez-Guerrero, M. Callejas-Cuervo, and S. Mazzoleni, “Design, Development, and Testing of an
Intelligent Wearable Robotic Exoskeleton Prototype for Upper Limb Rehabilitation,” Sensors, vol. 21, no.
16, p. 5411, Aug. 2021, doi: 10.3390/s21165411.
[11] F. Karadeniz, Ö. E. Aydoğan, E. A. Kazancı, and E. Akdogan, “Design of a 4-DOF grounded exoskeletal
robot for shoulder and elbow rehabilitation,” Sustain. Eng. Innov., vol. 2, no. 1, pp. 41–65, Jul. 2020, doi:
10.37868/sei.v2i1.106.
[12] Y. Liu, S. Guo, H. Hirata, H. Ishihara, and T. Tamiya, “Development of a powered variable-stiffness
exoskeleton device for elbow rehabilitation,” Biomed. Microdevices, vol. 20, no. 3, p. 64, Sep. 2018, doi:
10.1007/s10544-018-0312-6.
[13] M. Tiboni, A. Borboni, R. Faglia, and N. Pellegrini, “Robotics rehabilitation of the elbow based on surface
electromyography signals,” Adv. Mech. Eng., vol. 10, no. 2, p. 168781401875459, Feb. 2018, doi:
10.1177/1687814018754590.
[14] D. Copaci, F. Martin, L. Moreno, and D. Blanco, “SMA Based Elbow Exoskeleton for Rehabilitation
Therapy and Patient Evaluation,” IEEE Access, vol. 7, pp. 31473–31484, 2019, doi:
10.1109/ACCESS.2019.2902939.
[15] T. Yoshida et al., “Influence of motivation on rehabilitation outcomes after subacute stroke in
convalescent rehabilitation wards,” Front. Neurol., vol. 14, p. 1185813, Jul. 2023, doi:
10.3389/fneur.2023.1185813.
[16] M. G. Mutchnick, “The Role of Motivation in Rehabilitation,” in Cognitive Approaches to
Neuropsychology, J. M. Williams and C. J. Long, Eds., Boston, MA: Springer US, 1988, pp. 169–187. doi:
10.1007/978-1-4684-5577-9_9.
[17] M. P. Barnes, “Principles of neurological rehabilitation,” J. Neurol. Neurosurg. Psychiatry, vol. 74, no.
90004, pp. 3iv–7, Dec. 2003, doi: 10.1136/jnnp.74.suppl_4.iv3.
[18] R. L. Sainburg and P. K. Mutha, “Movement Neuroscience Foundations of Neurorehabilitation,” in
Neurorehabilitation Technology, D. J. Reinkensmeyer, L. Marchal-Crespo, and V. Dietz, Eds., Cham:
Springer International Publishing, 2022, pp. 19–39. doi: 10.1007/978-3-031-08995-4_2.
[19] K. E. Musselman, M. Shah, and J. Zariffa, “Rehabilitation technologies and interventions for individuals
with spinal cord injury: translational potential of current trends,” J. NeuroEngineering Rehabil., vol. 15,
no. 1, p. 40, Dec. 2018, doi: 10.1186/s12984-018-0386-7.
[20] K. A. Garrison, C. J. Winstein, and L. Aziz-Zadeh, “The Mirror Neuron System: A Neural Substrate for
Methods in Stroke Rehabilitation,” Neurorehabil. Neural Repair, vol. 24, no. 5, pp. 404–412, Jun. 2010,
doi: 10.1177/1545968309354536.
[21] J. McLoughlin, “Ten guiding principles for movement training in neurorehabilitation,” OpenPhysio J.,
2020, doi: 10.14426/art/1260.
[22] Y. G. Kim, K. Little, B. Noronha, M. Xiloyannis, L. Masia, and D. Accoto, “A voice activated bi-articular
exosuit for upper limb assistance during lifting tasks,” Robot. Comput.-Integr. Manuf., vol. 66, p. 101995,
Dec. 2020, doi: 10.1016/j.rcim.2020.101995.
[23] B. Kim and A. D. Deshpande, “An upper-body rehabilitation exoskeleton Harmony with an anatomical
shoulder mechanism: Design, modeling, control, and performance evaluation,” Int. J. Robot. Res., vol. 36,
no. 4, pp. 414–435, Apr. 2017, doi: 10.1177/0278364917706743.
[24] S. Kumar et al., “Modular Design and Decentralized Control of the Recupera Exoskeleton for Stroke
Rehabilitation,” Appl. Sci., vol. 9, no. 4, p. 626, Feb. 2019, doi: 10.3390/app9040626.
[25] Q. Liu, J. Zuo, C. Zhu, and S. Q. Xie, “Design and control of soft rehabilitation robots actuated by
pneumatic muscles: State of the art,” Future Gener. Comput. Syst., vol. 113, pp. 620–634, Dec. 2020, doi:
10.1016/j.future.2020.06.046.