Design and preliminary bench-level validation of a low-cost 3D-printed hip–knee exoskeleton for lower-limb rehabilitation training

Abstract

Lower-limb impairment caused by neurological injury, musculoskeletal disorders, or reduced mobility often requires repetitive rehabilitation exercises, but continuous therapist-assisted therapy may be difficult to access in low-resource settings. In this study, we developed and preliminarily evaluated a low-cost, lightweight, portable hip–knee exoskeleton prototype capable of generating controlled hip and knee joint motion for rehabilitation-oriented range-of-motion training. The prototype was constructed mainly of 3D-printed polylactic acid parts and powered by a pair of high-torque servo motors in the hip and knee joints. A Bluetooth-connected Android app was programmed to offer two modes: rehabilitation mode for independent joint control and a gait-inspired sinusoidal mode that generates coordinated hip and knee joint trajectories. The cost of the materials used to construct the prototype is estimated to be USD 155 with a weight of about 2.1 kg. Joint motion performance was assessed through time-angle measurements at three predefined speed settings. The hip joint (approximately 60° movement) was achieved in 2.2, 1.6, and 1.1 s and the knee joint (approximately 140° movement) was achieved in 3.0, 2.0, and 0.6 s for slow, medium and fast modes, respectively. The trial-to-trial variation in commanded servo position remained below ±1° under the tested conditions. The results demonstrate the feasibility of generating repeatable commanded hip and knee joint motion under controlled laboratory conditions. However, further investigations involving quantitative load-bearing evaluation, safety assessment, user studies, and clinical trials are required before the system can be considered for practical rehabilitation applications.

1. Introduction

The functional integrity of the musculoskeletal and neuromuscular systems of the lower limb is critical for maintaining proper gait and overall mobility, enabling essential activities such as walking, running, standing, and postural control. Injury or loss of a lower limb can profoundly affect an individual’s quality of life, often leading to reduced independence in performing daily tasks. Lower limb damage can come from various causes like accidental fracture, ligament injuries, spinal cord injuries, stroke, osteoarthritis, diabetes etc. Statistics from a study of 365 million people show that the knee is the most frequently affected joint, followed by the hip and the hand. ¹ Currently, 344 million people living with osteoarthritis could benefit from rehabilitation. ²

Moreover, a study shows nearly 6.6 million people have injured ligaments in America during 1999 to 2008, for a rate of 2.29 knee injuries per 1,000 population. ³ According to a study of 2017, reported that nearly half of the patients required orthotic support during rehabilitation, with many still depending on these devices upon discharge. ⁴ Rehabilitation for lower limb impairments often involves physical therapy, which aims to restore function and strength through targeted exercises. Figure 1 shows a reference lower limb exoskeleton suit showing external joint supports and actuator placement for hip and knee mobility assistance.

Figure 1.

A reference lower limb exoskeleton suit showing external joint supports and actuator placement for hip and knee mobility assistance. ⁵

Figure 2 represents the six stages that an individual pass after a stroke. These stages help doctors and therapists to understand what kind of treatment or therapy is needed for each patient. ⁶ The first stage is called flaccidity, when muscles are too weak and floppy to move the affected limb. In the second stage, called spasticity, muscles are strong and tight enough to help the person make small movements. In the third stage (marked spasticity), more muscle starts to tighten, and the person can do more movement in a repeated fashion. By the fourth stage (characterized by less spasticity), more muscle tissue becomes visible, and the person can engage in a new way of movement. In the fifth stage (selective control of movement), the person can perform complex movements, and by the sixth stage (isolated/coordinated movement), the execution of movements becomes smooth. In this stage, additionally, the person can move each joint separately, like before the stroke. After that the person can do normal movement and do daily work easily. In the present work, the proposed exoskeleton is primarily intended as an engineering prototype for individuals requiring controlled hip and knee range-of-motion exercises during the early stages of lower-limb rehabilitation. The current design is particularly suited for rehabilitation-oriented joint-motion training following orthopedic injuries or neurological conditions that require repetitive flexion and extension exercises under supervised conditions.

Figure 2.

The key stages of physical rehabilitation, outlining progression from early mobilization to functional recovery for patients with lower limb impairments. ⁶

The rehabilitation program helps patients progress through different stages under the supervision of a therapist. However, traditional rehabilitation is labor-intensive and requires continuous therapist involvement, which significantly increases treatment costs. To address this limitation, the use of a lower-limb exoskeleton can provide an effective alternative solution.

To date, many types of exoskeleton have been developed, they offer different types of functionality to adapt different condition of the patient. Yu et al., ⁷ developed a lightweight hip exoskeleton using quasi-direct drive (QDD) actuation, with a primary focus on designing an efficient actuator for exoskeleton applications. The actuator is capable of generating up to 36 N.m of torque and operates with a control bandwidth of 12.5 Hz. This relatively high bandwidth indicates that the actuator can respond rapidly to changes in motion. As a result, the actuator is well suited for walking-assistive devices, as it can quickly adapt to variations in human gait and maintain smooth real-time synchronization during walking. This capability is particularly important in rehabilitation, as it allows patients to move their limbs more freely when exerting voluntary effort while still receiving assistance when required. Although the actuator demonstrated excellent torque control and fast dynamic response, the study mainly focused on actuator development and did not present a complete low-cost exoskeleton system with a simple user interface for rehabilitation applications.

Wang et al., ⁸ designed a exoskeleton using an electrohydraulic servo system. It is design to application of walking assistance control for lower limb. It also has a system that can accurately control of torque and precise control for walking assistance while maintaining stability and comfort for user. This work contributes to advancing of development of lower limb for rehabilitation and mobility assistance but it has lack of real-world testing and active assistance implementation. Although the electrohydraulic actuator provides accurate torque control, the system is mechanically complex and mainly focuses on actuator performance rather than the development of a lightweight, portable, and low-cost rehabilitation device suitable for home-based use.

Wang et al., ⁹ developed a wearable lower limb exoskeleton for mobility assistance and rehabilitation. The author focuses to build a lightweight adaptable structure that can support normal human gait interaction and also provide enough actuation and control. The system has a real time feedback to ensure safe actuation. Its result shows strong potential for application in rehabilitation and human mobility support but it has limited control integration. Although the proposed system demonstrates a lightweight mechanical design, the control architecture remains relatively complex and does not emphasize a simple mobile-based interface for user-friendly rehabilitation.

Gao et al., ¹⁰ develop a lower limbs exoskeleton that focuses on optimized compliance control algorithm system. The algorithm helps to adapts the user’s movement and minimize resistance. The work result shows its control algorithm ensures smooth and responsive gait assistance. However, the study mainly focuses on advanced control algorithms, increasing computational complexity and making the system less suitable for low-cost rehabilitation platforms.

Although the exoskeleton demonstrated promising mobility assistance, the use of metallic structural components may increase fabrication cost and system weight compared with lightweight 3D-printed designs.

The reviewed studies demonstrate significant progress in actuator design, control strategies, and wearable exoskeleton develop While these examples represent major steps forward, numerous other studies continue to refine aspects such as material selection, actuation methods, and human–machine interaction—driving the field closer to practical, everyday use. Rehabilitation robotics has made great advancement throughout time, especially lower limb robotics. Development like lightweight design, enhance mobility assistance and intuitive control. Although there has been many research’s on lower limb exoskeleton for rehabilitation and mobility assistance there is still some way of improvement. Many systems are either too bulky and expensive or too limited for adaptability for individuals who have less resources. Also, majority of the system depend on single mode control and they have not any user-friendly control system. Affordability, portability, and personalization are key essential for long term rehabilitation especially when it is for home-based rehabilitation system. Without solving these problems exoskeleton rehabilitation technology remains only for clinical settings.

Recent studies published in Robotics further show that lower-limb rehabilitation exoskeletons are evolving toward anatomically aligned, adjustable, and wearable systems. Hasan and Alam ¹¹ presented a 4-DOF lower-extremity exoskeleton with anatomically aligned joints, modular adjustability, and adaptive sliding-mode control to improve biomechanical compatibility and trajectory tracking. Goo et al. ¹² reported a human-factors assessment of an adjustable pediatric lower-limb exoskeleton and showed that lightweight construction, rapid reconfiguration, and ease of donning and doffing are important for practical usability. In addition, Kóra et al. ¹³ clinically evaluated the powered ReStore soft exo-suit for stroke rehabilitation and reported improvements in walking-related functional outcomes, highlighting the translational value of lightweight wearable robotic assistance. In parallel, researchers have also explored mobile-application-based interfaces and additive-manufactured structures to improve usability and reduce fabrication cost.^14–25

Cost-effectiveness is another critical design factor in the rehabilitation robotics of the lower limbs. Recent research has shown that the addition of better biomechanical alignment, more degrees of freedom, more sophisticated sensing, and more customized wearable module functions is effective to improve performance, but may also bring more structural complexity, reconfiguration burden and development cost.^11,12 Recent review evidence has also stressed the need for future exoskeletons to be kept lightweight, wearable and clinically deployable with a balance between compliance, output performance and control complexity. ²⁶ On this basis, the present study takes simplified hip-knee architecture, 3D printed PLA, and electronic control hardware, which costs less, to realize a low-cost prototype to conduct preliminary rehabilitation-oriented training of the joint.

Although significant advances have been made in the development of lower-limb exoskeletons, multiple factors are still hindering its integration into daily rehabilitation, especially in resource-poor settings. Many current designs include sophisticated sensing systems, costly actuators, metallic frames, or complex control strategies, which may lead to additional weight, cost, and complexity. Moreover, while some reported exoskeletons focus on rehabilitation joint exercises or walking assistance, fewer low-cost systems implement both independent joint-based rehabilitation and pre-programmed coordinated hip–knee motion in a lightweight and easy-to-use system.

In this work, we propose a low-cost hip–knee exoskeleton prototype primarily fabricated using 3D-printed PLA and Arduino-Bluetooth-Android-based control. The objective of this study is not to demonstrate clinical rehabilitation outcomes, but to present and evaluate an engineering proof-of-concept prototype capable of generating controlled hip and knee joint motion for rehabilitation-oriented applications. Clinical rehabilitation effectiveness, walking assistance, and human-subject evaluation are beyond the scope of this study and are reserved for future investigations. The key contributions are: (i) the development of a lightweight and adjustable hip–knee structure; (ii) the use of a mobile-phone-based two-mode control interface; (iii) the development of independent manual joint control for rehabilitation exercises and a pre-programmed gait-inspired sinusoidal hip–knee trajectory generator; and (iv) the preliminary evaluation of joint-motion repeatability on a healthy volunteer at three predefined speed levels.

2. Materials and methods

To overcome such critical issues as affordability, versatility and provision of simple and user-friendly interface, the present work aims to develop a lower limb exoskeleton which is able to overcome such limitations. The proposed system focuses on achieving low fabrication cost while maintaining sufficient mechanical functionality for preliminary engineering evaluation (Figure 3). The prototype was designed based on five engineering design specifications defined for the early-stage development to lower-limb rehabilitation engineering at the early stages. First, the system will have to offer controlled hip and knee joint motion for repetitive range-of-motion exercises. Second, the device must be lightweight and affordable to aid accessibility in low-resource environments. Third, the mechanical structure must have an adjustable mechanism to fit in the various lower-limb sizes during the prototype fitting. Fourth, the control interface must be user or therapist friendly by using a mobile application. Fifth, repeatable joint-level motion should be offered by the system before progressing to load-bearing or human-subject evaluation.

Figure 3.

Highlighted special features of the proposed exoskeleton, including dual-mode operation, modular design, mobile app control, and lightweight 3D-printed PLA structure.

For the main structural parts, PLA was used as the fabrication material. PLA was chosen because of its low cost, biocompatibility and the ability to produce complex geometries in the Fused Deposition Modeling process. The filament used had a diameter of 1.75 mm, a nominal density of 1.24 g/cm³, and an approximate tensile strength of 60 MPa.^21,27–30 These properties of the material make it suitable to build a lightweight and low-cost exoskeleton structure designed for rehabilitation purposes.

All the structural parts were printed with a Creality Ender-3 Pro 3D printer. To achieve a smooth surface finish and reliable print quality, the nozzle temperature was set to 200 °C. The infill density was set at 80%, and the printing speed was set to 50 mm/s, to ensure that the structure possesses enough strength and the overall weight is low.

Two servo motors with high torque were used to move the main joints of the exoskeleton. An ASME-05B servo motor rated for 380 kg.cm (37.27 N.m) torque was used for hip actuation while an RDS51150 servo motor rated for 150 kg.cm (14.71 N.m) torque was used for knee actuation. Both the motors were powered from a 12.1V DC lithium-ion battery. Wireless communication between exoskeleton and control application was done using HC-05 Bluetooth module and central processing unit was made with the Arduino Uno microcontroller to control the system.

The developed system has two control modes. In manual rehabilitation mode, the hip and knee joints are driven independently using the mobile application and allow for slow and controlled flexion and extension exercises for therapist-guided or user-guided range-of-motion training. In gait-inspired trajectory-generation mode, the controller generates coordinated periodic reference trajectories for the hip and knee using predefined sinusoidal functions. In the present prototype, this mode should be interpreted as programmed gait-pattern generation rather than intention-driven assistance, because the system does not yet employ dedicated sensing for gait-phase estimation or user-intent detection.

2.1. Exoskeleton design & fabrication

Figure 4 showing the complete CAD model. The exoskeleton consists of four main parts: the hip mount, hip connector, knee mount, and knee connector. The exoskeleton has four major components: the hip mount, hip connector, knee mount, and knee connector. All of the components were designed using SolidWorks and made using 3D printing technology. The hip mount is used to secure the exoskeleton to the user’s waist and also provides a mounting interface for the servo motor which is responsible for actuating the hip joint. The hip connector links the hip mount to the knee section, allowing smooth transmission of rotational motion between the two segments. It also incorporates a servo coupling interface along with multiple adjustment holes, which enable fine tuning of alignment and fit during assembly.

Figure 4.

Complete CAD model of the proposed lower limb exoskeleton developed in solidworks, highlighting adjustable frame segments and motor integration.

The knee mount is attached around the user’s thigh and is connected to the hip mount through the adjustable hip connector, providing stable support for knee joint movement. Finally, the knee connector completes the exoskeleton. It links the lower part of the leg with the exoskeleton. Figure 5 showing all the four parts with dimension.

Figure 5.

All the four parts of the exoskeleton with their detail dimension, all dimensions are millimeters.

A key feature of the system is the screw-based adjustable-length mechanism between the upper and lower frame connectors, which provides approximately 5 cm of length adjustment to accommodate different user anthropometries as shown in Figure 6.

Figure 6.

Illustration shows the screw-based length adjustment mechanism between hip connector and knee connector up to 5 cm, that allows user to adjust length according to their height.

The exoskeleton (Control, Structure & interface and its Application Areas) was developed as a lightweight hip–knee prototype for controlled joint-motion generation intended for rehabilitation-oriented applications. The system incorporates an adjustable mechanical structure that accommodates users of different heights and serves as an engineering platform for investigating programmable hip and knee joint trajectories in future rehabilitation and assistive applications (Figure 7).

Figure 7.

The central image shows the physical prototype worn by a test subject, highlighting major structural parts such as the hip mount, servo motors, thigh and knee connectors. The surrounding panels illustrate core system features categorized as actuator, control, structure & interface and its application areas.

2.2. Components and connection

The exoskeleton’s electronic system is built around an Arduino microcontroller, which controls two high-torque servo motors: the RDS51150 ³¹ for the knee and ASME-05B ³² for the hip. A Bluetooth module HC-05 ³³ connects the Arduino to a custom Android app (developed with MIT App Inventor,^34–36 allowing wireless control of joint motion and speed. The app has basic controls for hip and knee movement for independent motion. Commands received via Bluetooth are used by the Arduino to generate PWM signals to make real-time actuation.

The control system of the exoskeleton is based on an Arduino Uno (Figure 8), which controls two servo motors: the RDS5115 to control the knee and the ASME-05B for the hip. To prevent overloading the Arduino’s onboard regulator, the servos are powered separately using a 12–12.7 V DC battery. A common ground connects the Arduino, servos, and power supply, ensuring signal integrity across the system.

Figure 8.

Showing the circuit connections of servos, HC-05 bluetooth module and power system with Arduino uno.

2.3. Android app user-interface

In order to have the ability to control it remotely and have an easy and user-friendly control system, a mobile application was created using MIT app inventor. The application interface has five main control buttons (Figure 9) for controlling the hip and knee servos. In manual rehabilitation mode, two buttons are dedicated to the hip joint and another two to the knee joint so that each servo can be independently controlled. A separate speed control button is also provided for adjusting the speed of the motion with three levels to choose from - slow, medium and fast.

Figure 9.

This is the user interface of the mobile application. It has Bluetooth connection buttons, hip and knee control button with different speed mode and two mode selection button.

There are also two mode-selection buttons “Manual Rehabilitation” and “Mobility Assistive” in the application to switch between the modes of operations based on the patient’s needs. The hip and knee servos can be adjusted manually for target exercises to help with therapeutic efforts. During the gait-inspired trajectory-generation mode, both servos are controlled following the programmed periodic reference trajectories to produce a coordinated motion of hip and knee. This mode is designed to only showcase trajectory-generation functionality in the laboratory and should not be interpreted as adaptive walking-assistance. The application interface is used to manage Bluetooth connectivity. When the device is disconnected, the Device Connected status message is “Search Device” and when successfully connected it changes to “Device Connected”, giving clear feedback to the user when using the device.

2.4. Description of the system

The system starts with the connecting Bluetooth between the device and the mobile application (Figure 10). Then user have to select between two mode Manual rehabilitation mode or Mobility assistive mode. In rehabilitation mode two joint can be remotely control using dedicated button. Hip and knee servo can control using button and targeted movement can be done. In the mobility assistive mode two servo works automatically with single press of hip or knee buttons and mimics normal human walking motion. To improve operational safety, software limits were implemented for both the hip and knee joints to prevent excessive angular displacement. In addition, the Android application includes an emergency stop function that immediately stops actuator motion whenever required by the user or therapist.

Figure 10.

Flowchart shows the work flow from user input in the mobile application to the exoskeleton device.

2.5. Kinematic modeling and Denavit–Hartenberg parameters

To describe the kinematic structure of the proposed lower-limb exoskeleton, the hip–knee assembly was modeled as a planar two-degree-of-freedom serial manipulator in the sagittal plane. The mechanism consists of two rigid links connected by two revolute joints corresponding to the actuated hip and knee joints. The first link represents the thigh-side segment of the exoskeleton, extending from the hip joint center to the knee joint center, whereas the second link represents the shank-side segment, extending from the knee joint center to the distal end of the lower support segment. The nominal link lengths of the prototype were L₁=19.5 cm and L₂=18.5 cm, corresponding to 0.195 m and 0.185 m, respectively.

As shown in Figure 11, three coordinate frames, {0}, {1}, and {2}, were assigned at the hip joint, knee joint, and distal endpoint, respectively. The base frame {0} was fixed at the hip joint. The generalized coordinates of the mechanism were defined as the hip joint angle θ1 and the relative knee joint angle θ₂. In the present frame assignment, θ₁ is measured from the base-frame x₀-axis to the first link, whereas θ₂ is measured from the virtual extension of the first link to the second link at the knee joint, as illustrated in Figure 11. Therefore, θ₂ represents the relative rotation between the thigh-side and shank-side links rather than an absolute angular orientation with respect to the global frame.

Figure 11.

Two-DOF Denavit–Hartenberg (D–H) kinematic model of the hip–knee exoskeleton.

Because the exoskeleton motion is restricted to the sagittal plane, all joint axes are parallel and normal to the plane of motion. Accordingly, the link twist parameters αi−1 and link offsets di are zero for both joints. Under these assumptions, the DH parameters of the exoskeleton can be expressed as listed in Table 1.

Table 1.

D–H parameters of the proposed planar two-link hip–knee exoskeleton.

Link i	Twist α i−1	Length L i-1	Offset d i	Joint angle ${\mathit{\theta }}_{\mathit{i}}$
0→1	0	0.195 m	0	${\theta }_{\mathbf{1}}$
1→2	0	0.185 m	0	${\theta }_{\mathbf{2}}$

Using the standard DH formulation, the homogeneous transformation matrix from frame i−1 to frame i is given by

$${}_i{}^{i-1}T=\left[\begin{array}{llll}\mathit{cos}\hspace{0.17em}{\theta }_i& -\mathit{sin}\hspace{0.17em}{\theta }_i& 0& L_{i-1}\hspace{0.17em}\mathit{cos}\hspace{0.17em}{\theta }_i\\ \mathit{sin}\hspace{0.17em}{\theta }_i& \mathit{cos}\hspace{0.17em}{\theta }_i& 0& L_{i-1}\hspace{0.17em}\mathit{sin}\hspace{0.17em}{\theta }_i\\ 0& 0& 1& d_i\\ 0& 0& 0& 1\end{array}\right]$$ (1)

Based on these parameters, the homogeneous transformation matrix relating the base frame {0} to the knee frame {1} is expressed as:

$${}_1{}^{0}T=\left[\begin{array}{llll}\mathit{cos}\hspace{0.17em}{\theta }_1& -\mathit{sin}\hspace{0.17em}{\theta }_1& 0& L_1\hspace{0.17em}\mathit{cos}\hspace{0.17em}{\theta }_1\\ \mathit{sin}\hspace{0.17em}{\theta }_1& \mathit{cos}\hspace{0.17em}{\theta }_1& 0& L_1\hspace{0.17em}\mathit{sin}\hspace{0.17em}{\theta }_1\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right]$$ (2)

Similarly, the transformation matrix from the knee frame {1} to the end-effector frame {2} is defined as:

$${}_2{}^{1}T=\left[\begin{array}{llll}\mathit{cos}\hspace{0.17em}{\theta }_2& -\mathit{sin}\hspace{0.17em}{\theta }_2& 0& L_2\hspace{0.17em}\mathit{cos}\hspace{0.17em}{\theta }_2\\ \mathit{sin}\hspace{0.17em}{\theta }_2& \mathit{cos}\hspace{0.17em}{\theta }_2& 0& L_2\hspace{0.17em}\mathit{sin}\hspace{0.17em}{\theta }_2\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right]$$ (3)

The overall forward kinematic model, which maps the joint-space variables (θ1, θ2) to the Cartesian task space, is obtained by multiplying the individual transformation matrices:

$$\left({}_2{}^{0}T={}_1{}^{0}T{}_2{}^{1}T\right)$$ (4)

Extracting the positional components from the resulting combined matrix yields the parametric equations for the Cartesian coordinates (x₂, y₂) of the exoskeleton’s distal end:

$${\mathrm{x}}_2={\mathrm{L}}_1\hspace{0.17em}\mathit{cos}\hspace{0.17em}{\theta }_1+{\mathrm{L}}_2\hspace{0.17em}\mathit{cos}\left({\theta }_1{+\theta }_2\right)$$ (5)

$${\mathrm{y}}_2={\mathrm{L}}_1\hspace{0.17em}\sin \hspace{0.17em}{\theta }_1+{\mathrm{L}}_2\hspace{0.17em}\mathit{sin}\left({\theta }_1{+\theta }_2\right)$$ (6)

These kinematic equations determine the spatial positioning of the device in the manual rehabilitation mode and the gait-inspired trajectory-generation mode. By inputting the experimentally defined angular ranges (θ₁: 0°- 30° for the hip; θ₂: 0°-65° for the knee), this analytical model confirms that the generated workspace safely encompasses the target physiological trajectories required for biomimetic lower-limb rehabilitation.

2.5.1. Torque analysis of the hip–knee exoskeleton

To further evaluate the actuation feasibility of the proposed hip–knee exoskeleton, a simplified dynamic and gravitational torque analysis was performed. The exoskeleton was modeled as a planar two-link mechanism operating in the sagittal plane, consistent with the kinematic model described in the previous section. The first link represents the thigh-side segment and the second link represents the shank-side segment. Since the present prototype is intended for controlled rehabilitation-oriented joint-motion training rather than load-bearing walking, the analysis was performed to estimate the torque required for unloaded or partially supported range-of-motion exercise.

The generalized coordinate vector of the system is defined as

$$q=\left[\begin{array}{l}{\theta }_1\\ {\theta }_2\end{array}\right]$$ (7)

where θ₁ is the hip joint angle and θ₂ is the relative knee joint angle. The nominal link lengths of the prototype were taken as

$${\mathrm{L}}_1=0.195\hspace{0.17em}\mathrm{m},$$

$${\mathrm{L}}_2=0.185\hspace{0.17em}\mathrm{m}$$

where L₁ is the hip-to-knee link length and L₂ is the knee-to-distal-support link length.

For estimating the lower-limb inertial and gravitational loading, Dempster’s classical anthropometric segment mass proportions were used. According to Dempster’s cadaver-based anthropometric model, the thigh, shank, and foot masses correspond approximately to 10.0%, 4.7%, and 1.4% of total body mass. ³⁷ Therefore, for the representative healthy adult used for qualitative prototype fitting in this study, with body mass M=85 kg, the estimated segment masses are

$${\mathrm{m}}_{\mathrm{T}}=0.100\mathrm{M}=8.50\hspace{0.17em}\text{kg}$$

$${\mathrm{m}}_{\mathrm{S}}=0.047\mathrm{M}=4.00\hspace{0.17em}\text{kg}$$

$${\mathrm{m}}_{\mathrm{F}}=0.014\mathrm{M}=1.19\hspace{0.17em}\text{kg}$$

where m_T, m_S, m_F represents the thigh, shank, and foot masses, respectively. In the simplified two-link model, the foot mass was treated as a distal lumped mass at the end of the shank-side segment because the present prototype does not include an actively controlled ankle joint. The center of mass of each link was approximated at 43.3% of the corresponding segment length from the proximal joint, following Dempster’s anthropometric convention. Thus,

$${\mathrm{r}}_1=0.433{\mathrm{L}}_1=0.084\hspace{0.17em}\mathrm{m}$$

$${\mathrm{r}}_2=0.433{\mathrm{L}}_2=0.080\hspace{0.17em}\mathrm{m}$$

where r₁ is the distance from the hip joint to the center of mass of the thigh-side segment and r₂ is the distance from the knee joint to the center of mass of the shank-side segment.

The dynamic equation of the planar two-link hip–knee system can be expressed as

$$\tau =M\left(q\right)\ddot{q}+C\left(q,\dot{q}\right)\dot{q}+G\left(q\right)$$ (8)

where τ is the actuator torque vector, M (q) is the inertia matrix, C (q, $\dot{q}$ ) represents Coriolis and centrifugal effects, and G (q) represents the gravitational torque vector.

For the simplified point-mass model, the inertia matrix terms are expressed as

$$M_{11\hspace{0.17em}}=m_T{r}_1^2+\left(m_s+m_F\right)L_1^2+\left(m_S{r}_2^2+m_F{L}_2^2\right)+2L_1\left(m_s{r}_2+m_F{L}_2\right)\cos \hspace{0.17em}{\mathrm{\theta }}_2$$ (9)

$$M_{12\hspace{0.17em}}=M_{21\hspace{0.17em}}=m_S{r}_2^2+m_F{L}_2^2+L_1(m_s{r}_2+m_F{L}_2)\cos \hspace{0.17em}{\mathrm{\theta }}_2$$ (10)

$$M_{22\hspace{0.17em}}=m_S{r}_2^2+m_F{L}_2^2$$ (11)

The gravitational torque components are given by

$$G_{1\hspace{0.17em}}=g[(m_T{r}_1+(m_s+m_F)L_1)\cos \hspace{0.17em}{\mathrm{\theta }}_1+(m_S{r}_2+m_F{L}_2)\cos ({\mathrm{\theta }}_1+{\mathrm{\theta }}_2)]$$ (12)

$$G_{2\hspace{0.17em}}=g\left[(m_S{r}_2+m_F{L}_2)\cos ({\mathrm{\theta }}_1+{\mathrm{\theta }}_2)\right]$$ (13)

Where, g= 9.81 m/s². The maximum gravitational torque occurs when the corresponding limb segment is approximately horizontal, because the moment arm with respect to gravity becomes maximum. Therefore, the worst-case simplified static condition was evaluated with the limb segments approximately horizontal, corresponding to maximum gravitational moment arm. Using the above anthropometric and geometric assumptions, the estimated maximum gravitational torque at the knee joint is

$${\tau }_{knee,\mathit{max}\hspace{0.17em}}=g\left(m_S{r}_2+m_F{L}_2\right)=5.30\hspace{0.17em}N.m$$ (14)

Similarly, the estimated maximum gravitational torque at the hip joint is

$${\tau }_{hip,\mathit{max}\hspace{0.17em}}=g\left[\left(m_T{r}_1+\left(m_s+m_F\right)L_1\right)+m_S{r}_2+m_F{L}_2\right]=22.26\hspace{0.17em}N.m$$ (15)

The estimated torque requirements were compared with the nominal servo motor torque ratings used in the prototype. The ASME-05B servo used for hip actuation has a rated torque of 37.27 N.m, while the RDS51150 servo used for knee actuation has a rated torque of 14.71 N.m. Therefore, the corresponding nominal torque ratios are.

$$\text{Hip}\hspace{0.17em}\text{torque}\hspace{0.17em}\text{ratio}=37.27/22.26=1.67$$

$$\text{Knee}\hspace{0.17em}\text{torque}\hspace{0.17em}\text{ratio}=14.71/5.30=2.78$$

These values indicate that, under the simplified unloaded sagittal-plane model, the selected actuators provide sufficient nominal torque capacity for preliminary range-of-motion operation. However, these ratios should not be interpreted as clinical safety factors because the servo ratings represent nominal motor specifications and may differ from continuous torque capacity under real operating conditions.

A summary of the estimated torque requirements is presented in Table 2.

Table 2.

Estimated gravitational torque demand for the proposed hip–knee exoskeleton based on Dempster’s anthropometric mass proportions.

Parameter	Hip joint	Knee joint
Actuated segment contribution	Thigh + shank + foot	Shank + foot
Estimated gravitational torque demand	22.26 N.m	5.30 N.m
Rated actuator torque	37.27 N.m	14.71 N.m
Nominal torque ratio	1.67	2.78

Because subject-specific anatomical segment lengths, exoskeleton-link mass distribution, and direct torque measurements were not available, the torque analysis should be interpreted as a first-order analytical estimate rather than a measured actuator-load validation. The calculated values do not represent clinical load-bearing capacity or verified joint torque during human use.

2.6. Experimental validation procedure

A brief wearable fitting and functional verification was performed Figure 12 for preliminary human-subject testing to assess joint-motion repeatability. The prototype was fitted to a healthy volunteer with a height of 173 cm and a body weight of 85 kg. The hip and knee joints were actuated individually through the Android application under three preset speed settings (slow, medium, and fast). For each operating condition, the experiment was repeated three times to evaluate the repeatability of the joint motion.

Figure 12.

A brief wearable fitting and functional verification was performed on a healthy adult volunteer to check mechanical fit and basic commanded motion. No clinical, biomechanical, gait, load-bearing, or physiological data were collected.

The joint angle was obtained from the commanded servo position displayed through the Arduino Serial Monitor. The motion completion time was measured using a digital stopwatch from the initiation of joint movement until the programmed target angle was reached. The collected measurements were used to prepare the time–angle graphs shown in the Results and Discussion section. For each testing condition, three independent trials were conducted, and the reported results represent the mean values, while standard deviations were calculated to quantify experimental repeatability.

2.7. Safety considerations

As the proposed exoskeleton is designed for rehabilitation-based lower-limb exercises, several fundamental safety features have been incorporated into the prototype. The hip and knee joints operate within predefined software angle limits to prevent excessive flexion or extension beyond the programmed range of motion. In addition, the Android mobile application includes an emergency stop button that immediately stops all actuator motion whenever required by the user or therapist. Besides the software-defined angle limits, mechanical stoppers were added to the hip and knee joints. These stoppers physically limit the hip flexion to 60° and the knee flexion to 140°, providing an additional safety mechanism to prevent excessive joint motion during operation. The prototype should not be used in individuals with severe spasticity, contracture, pain, joint instability, or restricted passive range of motion until torque sensing, calibrated validation of the mechanical stops, therapist-supervised safety protocols, and formal clinical safety evaluation are completed.

The mechanical frame was designed with adjustable hip and knee connector lengths to improve alignment between the exoskeleton joints and the user’s anatomical joints, thereby reducing unnecessary joint misalignment during operation. The wearable parts were attached with adjustable leather straps that hold the device securely, making the system more comfortable during preliminary fitting.

High-torque servo motors were selected to provide sufficient actuation capability for hip and knee motion. In the present study, all tests were conducted under controlled laboratory conditions and within predefined software motion limits. However, the current prototype does not include active torque limiting, force sensing, automatic fault detection, mechanical overload protection, or redundant fail-safe mechanisms for communication or actuator failure. Therefore, the system should be regarded as an engineering proof-of-concept rather than a clinically deployable rehabilitation device. The prototype should not be used in individuals with severe spasticity, contracture, pain, joint instability, or restricted passive range of motion until torque sensing, mechanical range stops, therapist-supervised safety protocols, and formal clinical safety validation are completed.

2.8. Prototype fabrication cost

The overall material and fabrication cost of the developed exoskeleton prototype was approximately USD 155. The major cost components were the Arduino Uno controller (USD 10), ASME-05B servo motor for the hip joint (USD 45), RDS51150 servo motor for the knee joint (USD 40), HC-05 Bluetooth module (USD 5), 1 kg of PLA filament used for additive manufacturing of the mechanical frame (USD 20), and a 12 V Li-ion battery (USD 25). An additional USD 10 was considered for miscellaneous components, including wires, connectors, bolts, and other assembly materials. This cost profile demonstrates that the proposed prototype can be fabricated at a relatively low cost, providing an economically accessible platform for future rehabilitation-oriented research and development.

3. Results and discussion

Two distinct motion conditions were evaluated in this study. First, manual rehabilitation mode was used to characterize the maximum controllable joint excursion and time response of the hip and knee actuators across three preset speed settings. Second, the gait-pattern mode was used to generate predefined sinusoidal reference trajectories for the hip and knee joints to produce coordinated periodic joint motion inspired by the human gait cycle. These two test conditions should not be interpreted as equivalent tasks, since the former evaluates the prototype’s range-of-motion capability, whereas the latter demonstrates the feasibility of generating coordinated gait-inspired joint trajectories under unloaded laboratory conditions.

Figure 13 is showing maximum flexion and extension (a) for hip angles and (b) for knee angles in rehabilitation mode. The physiological range of motion (ROM) of the human knee extends from 0° (full extension) to approximately 135°–150° of flexion. In rehabilitation, the required knee flexion depends on the patient’s condition and stage of recovery. Generally, a knee flexion of about 110°–120° is sufficient for performing daily activities such as walking, climbing stairs, and sit-to-stand movements. In the present study, the maximum knee flexion demonstrated by the prototype was selected to evaluate the mechanical capability and repeatability of the system under laboratory conditions rather than to represent a prescribed rehabilitation protocol for specific patient populations.^38–44 In the present study, the developed exoskeleton prototype was capable of achieving a maximum knee flexion of 140° during bench-top evaluation. This range was selected to assess the mechanical capability and repeatability of the prototype across its available range of motion while remaining within the anatomical range of the human knee. It was not intended to represent a prescribed rehabilitation target for any specific patient group, as the required joint range should be determined according to the individual’s clinical condition and therapist recommendation. The demonstrated range provides a basis for evaluating the prototype under manual rehabilitation and gait-pattern operating modes. During prototype development, the exoskeleton was briefly fitted to a healthy volunteer (Figure 13) (approximately 85 kg, 173 cm) to verify basic mechanical operation. This preliminary verification was qualitative only and no quantitative biomechanical or clinical data were collected. Therefore, these observations are not included as experimental validation. Although the selected servo motors are capable of generating high output torque (37 N·m), the prototype was operated under controlled laboratory conditions with predefined joint-motion limits. Consequently, the available actuator torque was not fully utilized during the present bench-top evaluation. The performance of the exoskeleton was tested by using time vs angle plots in three different speed mode: slow, medium and fast.

Figure 13.

The figure showing maximum flexion and extension (a) for hip angles and (b) for knee angles in rehabilitation mode.

Figure 14 shows the hip-joint time–angle response under three preset speed settings, while Figure 15 shows the corresponding knee-joint response. For each speed mode, the angular displacement of the servo joint was recorded over time. Each test was repeated three times, and the mean angular response was used for plotting. The trial-to-trial variation in commanded servo position remained below ±1° under the tested conditions. The joint-angle measurements were obtained from the commanded servo position reported by the controller rather than from an external motion-capture system or rotary encoder. Similarly, the motion completion time was measured manually using a digital stopwatch. Therefore, the reported repeatability reflects the consistency of the control system during the preliminary prototype evaluation under controlled laboratory conditions. Future studies will employ dedicated angle sensors, high-speed cameras, or motion-capture systems together with automated data acquisition to improve measurement accuracy and reproducibility. A quantitative summary of completion time, angular range, approximate angular velocity, and trial variation is presented in Table 3.

Figure 14.

Graph showing the change in angle with time at different speed setting for hip servo.

Figure 15.

Graph showing the change in angle with time at different speed setting for knee servo.

Table 3.

Summary of joint-level time–angle performance under three preset speed settings.

Joint	Speed mode	Target angular range	Completion time	Approx. angular velocity	Trial variation
Hip	Slow	0°–60°	2.2 s	27.3°/s	< ±1°
Hip	Medium	0°–60°	1.6 s	37.5°/s	< ±1°
Hip	Fast	0°–60°	1.1 s	54.5°/s	< ±1°
Knee	Slow	0°–140°	3.0 s	46.7°/s	< ±1°
Knee	Medium	0°–140°	2.0 s	70.0°/s	< ±1°
Knee	Fast	0°–140°	0.6 s	233.3°/s	< ±1°

As shown in Table 3, increasing the preset speed reduced the completion time for both hip and knee motions. The hip joint completed approximately 60° of motion in 2.2 s, 1.6 s, and 1.1 s under the slow, medium, and fast speed settings, respectively. The knee joint completed approximately 140° of motion in 3.0 s, 2.0 s, and 0.6 s under the corresponding speed settings. The limited variation among repeated trials confirms the repeatability of the proposed Bluetooth-based control architecture and actuator system during preliminary laboratory testing. The three speed settings were included to evaluate the prototype under different operating conditions rather than to define rehabilitation protocols. During actual rehabilitation, joint movement speed should be tailored to the patient’s physical condition and therapeutic requirements. While the fastest knee setting achieved an angular velocity of approximately 233.3°/s, this mode was included only to examine the actuator’s response capability and should not be considered suitable for rehabilitation exercises. Based on the observed motion characteristics, the slow and medium speed settings provided smoother and more controlled joint trajectories and may therefore be more suitable for rehabilitation-oriented applications, subject to further validation through quantitative load-bearing evaluations and comprehensive human-subject studies.

The graph in Figure 16 shows the time-angle response of the hip and knee joints when driven by the predefined sinusoidal reference trajectories. The dashed line represents the hip-joint angle, while the solid line represents the knee-joint angle. The hip joint operates within an angular range of approximately 0°–30°, which was selected to produce a gait-inspired motion profile based on commonly reported hip flexion during normal walking.^38–44 The hip angle gradually increases to its peak and then decreases smoothly before repeating the cycle, producing a continuous sinusoidal trajectory. This motion pattern was selected to approximate the general hip flexion profile reported during normal gait. In contrast, the knee joint exhibits a larger angular excursion, varying from approximately 0° (full extension) to 65° of flexion before returning to the initial position. This knee trajectory was chosen based on commonly reported knee flexion during the gait cycle to generate a gait-inspired joint motion for the prototype.^40,45 The curve is greater than in the hip which indicates the knee bends and straightens faster.

Figure 16.

Predefined sinusoidal reference trajectories of the hip and knee joints used to generate coordinated gait-inspired motion during gait-inspired trajectory-generation mode.

The experimental results show that the proposed prototype achieved repeatable hip and knee joint motion at three preset speeds with a trial-to-trial variation below ±1°. This demonstrates that the chosen servo actuation and Bluetooth-based control architecture is able to provide stable joint level motion for preliminary evaluation of programmed rehabilitation-oriented joint trajectories. However, the present evaluation remains limited to bench-level kinematic testing and simulated sinusoidal trajectories. No human gait trials, load-bearing test, electromyography or clinical outcome measures were included. Therefore, the current findings should be interpreted as preliminary mechanical and control validation rather than evidence of therapeutic effectiveness or independent walking assistance. Future studies should include under-load evaluation, power consumption analysis, mechanical durability study, and experiments for human subjects to test safety, comfort, gait kinematics and rehabilitation results.

A further limitation of the present system is that the gait-inspired trajectory-generation mode is based on predefined trajectories and does not yet incorporate real-time gait-phase detection or user-intent estimation. The integration of inertial sensors, foot-pressure sensors, joint encoders, or electromyographic signals could enable more adaptive assistance-as-needed control in future versions.

4. Conclusion

This paper described the design and initial bench testing of a cost-effective, lightweight, 3D-printed hip-knee exoskeleton prototype for joint training with rehabilitation applications. The system comprised a PLA-based, adjustable mechanical framework, high-torque servomotor actuation, Arduino control, Bluetooth data transfer, and an Android mobile phone application with two modes of operation. The system had an estimated material cost of USD 155 and a weight of 2.1 kg. Time–angle bench testing demonstrated consistent hip and knee actuation at three fixed speed settings, with a ±1° tolerance across trials. The results suggest feasibility in the use of the proposed system as a low-cost engineering platform for future lower-limb rehabilitation studies. But the current study did not conduct under-load testing, human gait trials, safety, comfort and clinical testing. The present work did not include quantitative load-bearing evaluation, fatigue testing of the printed mechanical structure, torque transmission efficiency analysis, or detailed joint-alignment assessment. Although the functional verification was performed on a healthy adult (173 cm, 85 kg) to check mechanical fit and basic commanded motion, no quantitative biomechanical or load-bearing data were collected. Therefore, no conclusions can be drawn regarding rehabilitation effectiveness, gait assistance, long-term safety, user comfort, or clinical outcomes. These aspects require systematic investigation through comprehensive engineering and clinically validated human-subject studies in future work. Future research will address mechanical durability, sensor integration and closed-loop control, power consumption, and ethically cleared human-subject testing to assess comfort, safety, gait assistance and rehabilitation significance.

ORCID iD

Sourav Karmaker https://orcid.org/0009-0008-6394-8452

Data Availability Statement

The time-angle datasets and bill of materials as well as Arduino command logic data of the experiment in this study can be requested by the concerned author. The CAD files and mobile-application interface files can be availed as auxiliary material where necessary.*