Pediatric Lower Limb Rehabilitation Training System with Soft Exosuit and Quantitative Partial Body Weight Support

Abstract

The pediatric period is a crucial window for motor function learning and growth. Individuals with central nervous system injuries like cerebral palsy commonly display severe crouch gait in the lower limbs. Hyperflexion of the knee joints promotes the forward trunk and increases reliance on the handle frame of a walker for support. In this study, we developed a quantitative partial body weight training system integrated with a soft pneumatic exosuit to assist the knee extension during the stance phase of the gait cycle. In the preliminary results for five pediatric cerebral palsy subjects, compared to the baseline condition, excessive knee flexion ameliorated with the assistance of the soft pneumatic exosuit. The peak knee extension and range of motion increased by 19.72° (±3.47°) and 15.46° (±5.06°), respectively. With exosuit assistance, the subjects demonstrated improved gait retraining compared to baseline. They were able to bear significantly more body weight on their affected limb, as evidenced by a 33.3% increase in the fraction of body weight measured by the force plate. Additionally, they relied less on the handrail for support during walking. With more extended knee joints to bear the load over gravity, the pediatric subjects transferred the reliance from external support and upper limbs back to the lower limbs as a more independent status during the loading response to terminal stance.

1. Introduction

The development of pediatric lower limb motor skills is critical during growth. In this crucial window, posture control and musculoskeletal coordination have a rapid development of maturation. However, central nervous system damage has a global high prevalence in early childhood in the form of cerebral palsy (CP) and developmental delay. The birth prevalence of cerebral palsy is approximately 3.4 per 1000 newborns [1,2]. The most common lower limb abnormality of CP is the crouch gait [3,4,5]. The etiology of the crouch gait is multifactorial and varies across developmental stages. From a muscle perspective, hamstring and hip flexor contracture increase the resistance of knee extensions. The weak quadriceps compromise the ability to maintain an extended knee joint while standing against gravity. At the level of neural control, increased co-contraction [6,7], impaired sensation and proprioception, and poor selective motor control exacerbate the tendency toward knee flexion [8,9,10]. Elevated quadriceps loading and metabolic cost reinforce a vicious cycle that is difficult to reverse spontaneously.

Conventional treatments for children with CP include physical therapy, orthotic management, botulinum toxin injections, and surgical interventions. Therapeutic stretching and orthoses can slow the progression of knee contractures [11,12,13,14]. Botulinum toxin injections reduce muscle overactivity and tone for several months [15,16]. Orthopedic soft tissue and bone surgeries are irreversible and may need to be repeated or staged throughout development in childhood [17,18,19]. Although these approaches can mitigate knee deformity and improve function, they are largely passive and do not fundamentally restore neural motor function or active motor control. In practical rehabilitation centers or special schools, therapists commonly assist pediatric CP individuals walking on a treadmill by stabilizing the child’s pelvis or providing posterior manual support on shank. This hands-on facilitation is physically demanding, leading to rapid fatigue and ergonomic load.

A wearable robotic exoskeleton has been incorporated in rehabilitation in conjunction with clinical therapy in several scenarios [20,21,22,23,24]. Actuation is most achieved via electric motors or cable-driven systems. The feasibility of augmented muscle recruitment, stability, and walking efficiency by the wearable exoskeleton has been proved. Nevertheless, participants have predominantly been adolescents or adults, and the limited pediatric cohorts generally include children who are already capable of independent walking. In the population of the pediatric CP, most children have not attained independent standing or walking and exhibited marked functional limitations. For pediatric individuals with CP, the user–device interface is particularly vulnerable: soft tissues and joints are sensitive and delicate. Rigid robotic metal structures with a fixed rotary trajectory are difficult to align with deformed joints caused by CP pathology. Misalignment increases the risk of shear force and contact pressures, thereby leading to discomfort and potential tissue injury in pediatric populations during prolonged use. Soft wearable technology, due to its light weight, low inertia, compliance, and high-power density, is well suited for wearing on both upper and lower limbs [25,26]. The unpowered soft exosuit or actuators produced minimal impedance to preserve pediatric CP natural kinematics. The textile-based material can be flexibly fabricated to follow the required force direction using sewing and heat bonding. Comfortable and breathable anchoring solutions support routine use and training [27,28,29,30]. The material of the exosuit applied in pediatric wearables requires safety, skin compatibility, and manufacturing reliability. TPU was chosen for its optimal balance of airtightness, elasticity, and comfort. Unlike PU, which may deform over time, and PVC, which is heavier and stiffer, TPU provides a lightweight, non-porous barrier that conforms well to body contours and withstands repeated mechanical stress. In this study, TPU provides greater resistance to mechanical stress and deformation with balanced flexibility.

Apart from the device-side factors, the high dependence of CP children on external support and assistive devices is another core challenge [31,32]. During daily activities and training, children with CP frequently use hand-held assistive devices or adopt a forward trunk lean posture to achieve rapid stability through upper limb support (Figure 1) instead of activating lower limb strength, thereby offloading lower limb weight bearing and knee extension demands onto the upper limbs. This easier compensation reinforces an incorrect load path. In the long term, this reliance pattern continually decreases knee extension and push power over gravity, weakens the functions of the quadriceps and plantar flexor, and worsens the crouch gait. To ensure safety, training task specificity, and intensity, a partial body weight support (PBWS) system combined with wearable robotics has become a trend [33,34,35]. At present, clinical trials and research lack quantitative indicators for the external dependence of CP pediatric individuals. The function assessments only document whether a handle frame is used but rarely quantify the body weight percentage borne by the handle frame in the gait cycle. Without quantification and feedback, it is difficult to systematically transfer the load bearing from the upper limbs to the lower limbs. Across different rehabilitation stages, this hampers evaluating the independent walking ability of the lower limbs for a subsequent training protocol in the next step.

To address these challenges, in this study we propose a PBWS monitoring system integrated with a soft pneumatic exosuit for lower limb rehabilitation of pediatric CP individuals. PBWS recorded the percentage body weight from force plate, handle frame, and suspension as the quantification of the external support during walking. The treadmill of the PBWS provided a comfortable walking speed selected by the subjects. For crouch gait, the soft pneumatic exosuit used the textile-based TPU coated Nylon 210D to extend the subjects’ knee joints with compliant, powerful, and comfortable assistance. To accommodate varying impairment severity of crouch gait and bodyweight, we established an “assist-as-torque” model designed according to the required knee extension torque of variable body weight and knee flexion angle. During each gait cycle, the soft pneumatic exosuit inflated to support the knee extension from heel strike to terminal stance. Based on this pediatric lower limb training platform, we proposed two hypotheses. (1) With the assistance of a soft pneumatic exosuit, the excessive knee flexion in the stance phase would be reduced as a more extended status compared to the baseline condition. (2) From the monitoring of the percentage of body weight from force plate, handle frame, and suspension, the load bearing of external support from upper limbs and suspension would transfer back to the lower limb knee extension. Consequently, the load path over gravity during gait would be improved as a more upright posture.

2. Methods and Materials

2.1. Quantitative Partial Body Weight Support Training System

Due to the limited locomotor function of target pediatric subjects, the training system is supposed to provide the following: (1) adjustable body weight support to promote proper postural alignment for safe standing and walking; (2) quantification and feedback on the subjects’ reliance on suspension and handle frames; and (3) a comfortable walking speed during the training. As shown in Figure 2, to ensure the safety of the training, the suspension system connected to both sides of the harness of the pediatric subject afforded a rated loading of 80 kg. The adjustment range of the lifting column was from 187 to 206 cm. The distance between the handles was customized as 40 cm for a comfortable pediatric holding. The treadmill offered a walking speed from 0.5 km/h to 6 km/h, consistent with the routine training protocol of the pediatric centers and special schools. From top to bottom, seven transducers in total were assembled. Two S-type transducers (SBT630D, SIBATUO, Guangzhou, China) were mounted to measure the support from the suspension. One uniaxial force transducer (SBT301, SIBATUO, Guangzhou, China) was installed beneath the handle frame to quantify the support from the handrail. Four transducers (SBT673, SIBATUO, Guangzhou, China) were positioned on the corners of the force plate to capture the vertical ground reaction force. The data of the transducers were captured by data acquisition DAQ (USB6211, NI, Austin, TX, USA). The external support forces were normalized to the percentage of the subject’s body weight. The distribution of the load path on these three components was monitored during the subject’s walking training. This approach enabled a quantitative evaluation of the walking independence and assistance intensity of the lower limb rehabilitation. The entire partial bodyweight support training system had compact spatial dimensions of 115 cm × 77 cm × 200 cm. The system is highly suitable for deployment in rehabilitation centers, special schools, and even at home using flexibility for different training scenarios.

2.2. Soft Pneumatic Lower Limb Assistive Exosuit

As the most common pathology of CP lower limbs, prolonged crouch gait generated abnormal stress on the knee joints. Sufficient, fast-response, and comfortable extension support were the challenges in daily rehabilitation training. Hence, we proposed a soft pneumatic exosuit combined with walking training based on the PBWS system to assist knee extensions during the gait cycle. The material of the pneumatic padding was TPU coated Nylon 210D used to form the structure connected to the shank and thigh. The anchoring of the exosuit employed skin-friendly materials.

For pediatric subjects, ensuring comfort and biomechanical alignment is critical due to the sensitivity of the knee joint—particularly the popliteal area, which is prone to discomfort under compression. Young children typically have more fatty tissue around the posterior knee, making them more susceptible to pressure-related irritation. Moreover, the knee joint exhibits a polycentric motion rather than a single-axis rotation, meaning that a uniaxial electric motor can easily misalign with the joint center during walking, leading to instability and discomfort.

Traditional I-shaped or bellow-type pneumatic exosuits often expand at the center, generating uneven pressure and stress concentrations at contact areas, especially around the popliteal region. To address this, we developed a novel W-shaped exosuit design that spans across the joint without compressing the sensitive posterior knee. This structure anchors more effectively to the thigh and shank, distributing assistive force more uniformly and maintaining alignment throughout gait cycles.

Finite Element Analysis (FEA) using Abaqus (SIMULIA Abaqus FEA 2023, Dassault Systèmes, Vélizy-Villacoublay, France) was conducted to compare the I-shape and W-shape designs under equal volume conditions in Figure 3. Both models used 3D shell elements (S4R) with a thickness of 0.3 mm and material properties of Nylon TPU 210D (Young’s modulus: 120 MPa, Poisson’s ratio: 0.35, density: 1.16 × 10⁻³ g/mm³). Contact interfaces between the thigh and shank were modeled with quadrilateral elements (R3D4), simulating joint angles from 0° to 90°. Results showed that the W-shaped design significantly reduced stress in the popliteal area and maintained better conformity with the joint’s polycentric motion.

As detailed in Figure 4c, the extensible strap with sewn Velcro delivers easy and adjustable wearing around the knee joint of the subjects. The non-slip silicon layer strengthens the fixation of the brace. The breathable fabric at the skin interface ensures comfort in long-time use. The extension torque transmitted on both sides during the inflation of the exosuit in Figure 4b. The unpowered exosuit for the natural swing phase of the gait cycle is presented in Figure 4a.

2.3. Testing Bench

To test the performance of the soft pneumatic exosuit, we set up an experiment bench to record the data of pressure, time, and angle. The experiment consisted of a pressure regulator (ITV2050, SMC, Tokyo, Japan), stepper-based positioning unit (SST42D2121, Shinano Kenshi, Ueda, Japan), torque transducer (SBT850A, SIBATUO, Guangzhou, China) and control unit (STM32F407, STMicroelectronics, Geneva, Switzerland) in Figure 5. The air pressure transmitted to the exosuit as the command from the PC to the control unit. For standardized production of the padding, we engaged the manufacturer (Tian Cheng Plastic Products Co., Ltd., Dongguan, China) to use high-frequency welding for mass production. The welding width of the padding was 5 mm. The diameter of the tube connected to the air inlet was 8 mm. For standardized production of the padding, we relied on the manufacturer’s experience in this material. According to the manufacturer’s guidance, the padding can withstand the maximum pressure at 300 kPa. Considering the joint safety and comfort for pediatric subjects with CP, we set the upper operating pressure limit at 100 kPa and it can generate knee torque at 90 Nm. Five subjects in this study only used less than 30 Nm for standing assistance.

2.4. Assist-As-Torque Model

The assistance principle is based on the knee joint extension torque required by children with cerebral palsy of varying body weights during ambulation. The design was for 2–5 year olds with a weight under 30 kg [36]. Lerner et al. [37] used the electronic motor of the exoskeleton to facilitate CP children and adolescents with walking, demonstrated that the required torque, with a maximum of 0.38 Nm/kg, varied according to bodyweight. With reference to them, the required torque for a 30 kg subject was calculated as 11.4 Nm. The internal volume is defined by a cross section comprising two semicircles of radius $\left(r\right)$ and a central rectangle of width $\left(w\right)$ and height $\left(2r\right)$. In our study, five recruited subjects had a thigh circumference of 34.8 ± 7.47 cm, and the shank circumference was 26.8 ± 7.72 cm. The maximum width of the exosuit padding was under 80 mm. When the pressure was under 100 kPa, the padding with $30 \mathrm{m}\mathrm{m}<w<\mathrm{m}\mathrm{m}$ and $15 \mathrm{m}\mathrm{m}<r<25 \mathrm{m}\mathrm{m}$ complied with the desired torque. Considering the anatomical dimensions of the subjects, the required torque, and the pneumatic pressure range, the final design dimensions were set to $w=30 \mathrm{m}\mathrm{m}$ and $r=25 \mathrm{m}\mathrm{m}$.

In the pneumatic model of the exosuit, in a previous study of Christopher and colleagues [38], they proposed the pneumatic work $\left(W\right)$ performed by the pressure $\left(P\right)$, volume $\left(V\right)$, angle $\left(\theta \right)$, and torque $\left(\tau \right)$ relationship as Equations (1) and (2). We further refined the relationship between the required input air pressure and the knee flexion angles corresponding to different baseline crouch gait patterns based on the cross section of our exosuit. Figure 6 illustrates the parameters of the padding of the exosuit for modeling. The cross sectional area could be parameterized with radius $\left(r\right)$ and width $(w$). The torque of the padding $\left({\tau }_{padding}\right)$ arised from the work performed by the pressure $\left(P\right)$ under the volume $\left(V\right)$ changing in the folding domain [38]. During the donning of the exosuit, the knee flexion angle was denoted by$\theta$. The $\alpha$ was the supplementary angle of the padding folding intersection with interface. The assist-as-torque model defined the required torque ${\tau }_{required}$ via the variation in the knee flexion angle and compressed pressure.

As shown in Figure 7, compared to the theoretical prediction, the error was 2.13 Nm as an average group level. In the most common crouch gait with a knee flexion angle of 30°, the error was 1.66 Nm. This model serves as a guide for the pneumatic exosuit to provide targeted assistance, enabling contracted knee joints to achieve more natural extension of the stance phase during the gait cycle.

$$W=\int PdV=\int \tau d\theta$$

(1)

$$\tau =P \left(V\left(\theta \right)\right)· {\displaystyle \frac{dV\left(\theta \right)}{d\theta }}$$

(2)

$${\tau }_{required}=k_{required}{\tau }_{padding}$$

(3)

$${\tau }_{padding}=P{V}^{\prime }\left(\theta \right)=P{\displaystyle \frac{\left(\pi r^2+2wr\right)2r}{{\cos }^2\left[\frac{\left(\frac{\theta }{2}+\alpha \right)}{2}\right]}}$$

(4)

$${\tau }_{required}=k_{required}P{\displaystyle \frac{\left(\pi r^2+2wr\right)2r}{{\cos }^2\left[\frac{\left(\frac{\theta }{2}+\alpha \right)}{2}\right]}}$$

(5)

To validate the model, we measured the torque output on a test bench under varying air pressures and knee joint angles and compared the experimental results with the theoretical values calculated from the proposed model. The torque measurements were conducted across a range of knee flexion angles from 0° to 70° covering the crouch knee flexion angles of different severities and input air pressures from 0 to 100 kPa for practical safety. As shown in Figure 7, compared to the theoretical prediction, the error was 2.13 Nm as an average group level. In the most common crouch gait with knee flexion angle of 30°, the error was 1.66 Nm. The results demonstrated a strong correlation between the experimental data and the model predictions, which provided a basis for pressure selection under different conditions.

2.5. Response Time

In practical walking training, the self-selected walking speed was changed by the subjects with different Gross Motor Function levels. Alignment from the assistance and gait phase of the impaired function pediatric subjects was the key in improving the quality, safety, and comfort of walking. The related study adopted the response of inflation and deflation to map the gait event. Based on the velocity, cadence, and stride length in the research by Kim et al. (2014) [39], the gait cycle could be calculated around 1.34 s. To test the different pressure responses, we varied the command pressure of the exosuit from 0 to 100 kPa.

In Figure 8a, the command pressure (red dotted line) and the pressure measured in the experiment (blue line) are plotted. At a setpoint of 40 kPa, the measured pressure increased from 0 to 40 kPa within 0.16 s (inflation time) and subsequently decreased from 40 to 0 kPa within 0.18 s (deflation time). Figure 8b shows the response time from 20 kPa to 100 kPa. The inflation/deflation time were marked above each pressure. From the experiment result, the mean inflation time was 0.22 s, and the corresponding deflation time was 0.2 s with an error of 0.45 kPa. The inflation time was 16.4% of the gait cycle of 1.34 s, which was sufficient in the loading response and pre-swing phase of the gait cycle.

2.6. Control Principle

At the start of the experiment, each subject’s body weight was measured to determine the required pneumatic pressure based on a previously established torque-to-body weight model. Two wireless IMUs (Hi221, SEA LAND, Taiwan, China) sampled from 50 to 100 Hz were worn on the thighs and shanks of the subjects to calculate the kinematic knee flexion angle. Force-sensitive resistors (A401, Tekscan, Boston, MA, USA) were placed on the heel and forefoot of the insole. After donning the pneumatic exosuit, the pressure and corresponding FSR threshold were recorded when the subject fully extended their knee joint as the calibration.

In the assistance mode, the FSRs detected the gait event of the gait cycle in order to initiate inflation and deflation from the air compressor (Jun-Air, A/S, Norresundby, Denmark) to the exosuit. As shown in Figure 9, inflation lasted from the heel strike to the terminal stance as a knee extension support. When the loading force was lower than the FSR threshold, the exosuit released the air immediately without the impedance in the swing phase. The pressure regulator adjusted the pressure for each subject as the setting in the calibration session. The real-time signal of FSR and air pressure were recorded with DAQ (USB6211, NI, Austin, TX, USA) as 1000 Hz. The synchronization to IMU was implemented by Python (version 3.12.1) script.

3. Pilot Testing Design

3.1. Pediatric Cerebral Palsy Subjects

Five local CP subjects were recruited in this study, and the inclusion criteria were as follows: (1) diagnosed with a type of spastic diplegic cerebral palsy; (2) aged between 2 and 5 years; (3) GMFCS level was less than IV; and (4) possessing the ability to follow instructions. Subjects would be excluded if (1) they underwent treatment with botulinum toxin or surgery within the past six months and (2) any other neuromuscular disorder affected their motor patterns. The demographics including age, body weight, height, and Gross Motor Function level of subjects are shown in

Table 1. Subjects’ demographics.

Title 1	Mean ± Standard Deviation	Min–Max
Age (year)	4 ± 1.1	2–5
Body Weight (kg)	16.38 ± 3.43	10.8–20.5
Height (cm)	105.06 ±13.36	90.2–122.3
Thigh Circumference (cm)	34.8 ± 7.47	25–35
Shank Circumference (cm)	26.8 ± 7.72	18–31.5
GMFCS	2.60 ± 0.49	II–III
Exosuit Pressure (kPa)	56 ± 8	40–60

. All subjects and their guardians were provided clinical trial consent approved by the Joint Chinese University of Hong Kong—New Territories East Cluster Clinical Research Ethics Committee (The Joint CUHK-NTEC CREC, Ref. No. CREC 2021.171). The clinical trial was registered on ClinicalTrials.gov (ID NCT05580497). All subjects and their guardians received a detailed explanation of the potential risks of the experiments and signed informed consent before participating in this study.

3.2. Experiment Procedures

Following functional screening, we measured each recruited subject’s body weight and lower limb anthropometrics to fabricate and adjust the individualized brace of the soft pneumatic exosuit. During the first lab visit, all five pediatric CP subjects completed a standardized baseline and calibration protocol. To determine the required assistance pressure for full knee extension, each subject performed a sit-to-stand task using the exosuit. Body weight was measured to estimate the necessary pneumatic torque based on a validated assist-as-torque model. With the exosuit donned, calibration was performed at full knee extension, recording both the assistance pressure and force-sensitive resistor (FSR) threshold for gait phase detection. The actual measured pressure was from 40 to 60 kPa. Subjects then completed a 1 min-warm-up walk to familiarize themselves with the system and select a comfortable treadmill speed. A baseline trial without exosuit assistance was recorded during 2 min of walking. Immediately thereafter, the subjects performed a 5 min-assisted-walking trial with the exosuit powered (pressurized air inflation) at the calibrated settings and the selected speed. Under both baseline and powered conditions, the knee kinematic joint flexion angle and the percentage body weight (PBW) support by the treadmill force plates, the handle frame, and the suspension were recorded.

4. Results

4.1. Kinematic Improvement of Knee Joints

We implemented pilot testing under baseline and powered conditions on a total of five subjects with cerebral palsy (CP) and crouch gait. With assistance from a soft pneumatic exosuit, participants demonstrated a clear reduction in excessive knee flexion during a stance without hyperextension or atypical reverse-flexion patterns (Figure 10a–e). Ensemble knee-angle trajectories showed greater extension during weight acceptance and mid-stance and an expanded range over the gait cycle. Quantitatively, relative to baseline, Subject 4 showed the greatest improvement: the peak stance-phase knee extension (i.e., lowest knee flexion) increased by 25.05°, and knee joint range of motion (ROM) increased by 22.43°. Across all participants (N = 5; see Figure 10f), peak stance-phase knee extension increased from 26.15° (±5.32°) to 6.88° (±4.79°), and ROM increased from 38.19° (±3.78°) to 53.65° (±4.08°).

4.2. Partial Body Weight Support Analysis

With the knee extension support from the soft pneumatic exosuit during walking, dependence on external supports decreased, reflected by a redistribution of load toward more independent weight-bearing in the vertical direction. Figure 11a–e depicted the peak percentage of bodyweight on the force plate, handle, and suspension under baseline and powered soft pneumatic exosuit conditions. At the individual level, especially for Subject 3, the peak percentage of body weight from force plate increased from 51% to 84%. Meanwhile, the peak percentage of body weight from handle frame decreased from 23% to 8% and peak PBW from suspension decreased from 18% to 7%. At the group level, the peak percentage of body weight from the force plate increased from 57.6% (±4.9%) to 76.8% (±5.6%), peak percentage of body weight from the handle frame decreased from 23% (±4.8%) to 15% (±5.2%); and peak percentage of body weight from the suspension decreased from 19.8% (±5.0%) to 12.2% (±2.9%).

5. Discussion

In this study, we developed a partial body weight support monitoring system integrated with soft pneumatic exosuit assistance for pediatric lower limb training. To test the feasibility, our validation progressed from the laboratory benchtop to a clinical evaluation involving five pediatric subjects with cerebral palsy. According to the subjects’ feedback, the proposed soft pneumatic exosuit combined ease of donning, compatibility with the skin, and a robust assistance force. From the pilot testing results, assistance during the stance phase from the soft pneumatic exosuit promoted the knee extension of the subjects and alleviated their crouch gait. Thereby, it effectively shifted the load from the external support of the handle frame and suspension, reducing dependence on the support. The quantification of the support from the training system indicated the translational potential for pediatric rehabilitation.

Regarding user–system interaction, test results demonstrated that the torque capacity and response time met the requirements for extension support during critical phases of the pediatric gait cycle. The soft, breathable anchoring brace improved wearing comfort, accommodated dimensional changes due to misalignment and growth, and reduced the risk of skin scratches and compression on the joints. These properties were crucial for the pediatric subjects with fragile soft tissues, small joint surface areas, and compliance that is highly sensitive to discomfort.

Pediatric individuals with crouch gait are commonly associated with weak quadriceps and an overactive gastrocnemius of the lower limb muscle. We proved our hypothesis that appropriate augmentation of the extension of the knee during the stance phase of the gait cycle contributes to the stabilization of the knee joint from loading response to mid-stance. In addition, the loading response prolonged 2.2% of the gait cycle overall under the inflation of the soft pneumatic exosuit. The longer extension duration reduced the compensation from upper limb support from the handle frame or trunk support from suspension. Instead of the “impose-trajectory” of a rigid mechanical exoskeleton or passive orthoses, the compliance of the soft pneumatic exosuit used “assist-as-torque” while preserving pediatric individuals’ natural joint movement. Aimed at the age stage of the target population, maintaining sensory feedback and locomotor freedom were necessary to relearn and restore motor functions. Walking training also encouraged the pediatric subjects to use their paretic limbs more with sufficient intensity. This was consistent with the principle of strengthening neural plasticity in pediatric rehabilitation.

A partial body weight support monitoring system was crucial in explaining the pediatric subjects’ independence. Quantifying the reliance on external support in real training scenarios using conventional gait laboratory indicators is difficult. We simultaneously recorded the partial body weight support ratios from the force plate, handle frame, and suspension, which were capable of reflecting the differences between baseline and powered assistance conditions. As the extension assistance of the soft pneumatic exosuit provided as the torque the body weight scale, independent lower limb loading increased significantly. Compared to the baseline, both loading on the handle frame and harness were decreased. The subjects transferred the lean-forward posture with an upper-limb-dominated strategy to an upright position during the stance phase. This “assist-as-torque” intervention strategy maintained safety while providing a biomechanically appropriate alignment of the knee joints. The proposed training system gave feedback to the therapists regarding the threshold for balancing the support with the handle frame and harness. In long-term rehabilitation, the training could gradually decrease the external support as planned behavior to further enhance the motor function to achieve independent standing and walking.

This study targets crouch gait—one of the most common gait abnormalities in children with cerebral palsy—by focusing on the knee joint. While ankle–foot orthoses (AFOs) can correct ankle alignment and partially assist with walking, knee dysfunction in crouch gait remains difficult to address. The knee is a polycentric, multi-axis joint, and conventional single-axis electric motors can cause discomfort and poor kinematic coordination when used alone. To overcome these limitations, we developed a novel soft actuation knee mechanism that delivers assistance through two adjacent limb segments directly across the knee joint, which also provides a more compliant interaction with the wearer than conventional electric actuators. Our device was specifically designed to reduce crouch posture and improve weight bearing and gait symmetry in pediatric CP patients. Future work will explore extending this soft actuator approach to coordinated multi-joint assistance (hip, knee, and ankle) to provide integrated, comfortable support for children with complex gait disorders. In this preliminary study, we prioritized safety and comfort, so we implemented a simple fixed threshold inflation/deflation controller to validate integration of the PBWS monitoring system with the soft pneumatic exosuit. The simple fixed threshold control also makes the system more predictable and easier for children to understand and anticipate, which helps them prepare mentally and feel more confident when walking with the device during early training. In future work, we will develop an adaptive control strategy that dynamically adjusts assistive torque to each child’s real-time gait.

Furthermore, in this study, we focused on quantifying vertical force distribution by monitoring how much of the load was supported by the patient’s affected limb versus the handrail during gait. While shear force was not included in the current set up, incorporating shear-force data would provide a more comprehensive assessment of lower limb loading and gait stability. In future work, we plan to integrate a three-axis load cell into the system to capture shear components alongside vertical forces, enabling a more detailed analysis of the joint loading and rehabilitation progress.

Additionally, due to the strict inclusion criteria, only a small sample size of five CP subjects with GMFCS levels from II to III were recruited. Studies with a similar sample size for pediatric training showed the feasibility of extrapolation to a larger population. Lerner et al. depicted that the knee extension assistance improved the CP subjects’ lower-extremity gait kinematics and muscle activity [40,41]. Elena et al. used the ATLAS 2030 gait exoskeleton to enhance the strength and ROM of subjects with spinal muscular atrophy and CP [42,43]. However, a larger sample size with long-term rehabilitation training will yield a more comprehensive evaluation of the system’s impact. Our preliminary study involved a limited number of participants, and the observed improvements in gait and load distribution were consistently aligned with the intended effects of the proposed system. To better evaluate the effectiveness of the proposed system, future studies will incorporate randomized controlled trials comparing outcomes with a matched control group receiving conventional therapy. For the next phase, we will conduct a power analysis to determine the appropriate sample size for a randomized controlled trial comparing the proposed system with conventional therapy. In addition, we will apply appropriate statistical tests to evaluate the significance of motor function improvements and ensure robust, quantitative validation of our findings.

With the quantitative PBWS system and “assist-as-torque” strategy, the diversity of the motor function level will be augmented to recruit subjects with severe- and mild-level disorders. According to the previous studies, we will implement the complete walking rehabilitation training trial with 12–20 sessions [44,45,46]. Each session will include at least 30 min of walking training with a quantitative PBWS monitoring system and pneumatic exosuit. The assistance training strategies will vary according to impairment levels. The preliminary results of GMFCS levels from II to III will be a reference of max assistance limits for GMFCS Level I and safety, starting the range for GMFCS Level IV.

To validate the intervention outcomes, we will implement pre- and post-assessment using standardized clinical tools. For the motor function, we will adopt the Gross Motor Function Measure (GMFM88). A 10 Meter Walking Test (10MWT) and a 6 Minute Walking Test (6MWT) will be used for scoring speed and endurance. The Modified Ashworth Scale (MAS) of lower limbs will evaluate the muscle tone of the joints of hip, knee, and ankle joints. The kinematic and neuromuscular data will be combined to comprehensively analyze motor function improvement. In addition, we will apply appropriate statistical tests to evaluate the significance of motor function improvements and ensure robust, quantitative validation of our findings.

The current system is relatively bulky and optimized for controlled clinical settings. To support home-based rehabilitation for children, future work will focus on miniaturizing key components—such as the pneumatic power unit and control electronics—and improving overall portability. We plan to explore lightweight materials, compact actuator designs, and wearable-friendly layouts to enable safe and effective use in home environments. To enhance the durability of the exosuit, increasing the welding width of padding will improve pressure maintenance. On this basis, we will develop a multi-chamber padding to provide more flexible and uniform support and reduce local stress concentration [47]. Finally, subjects with similar pathologies, such as developmental delays and pediatric stroke, could also benefit from the proposed system. By recruiting the different types of pediatric subjects with central nervous system damage and a typical developmental control group, the system’s generalization ability in clinical application will be improved.

6. Conclusions

In conclusion, we developed a PBWS monitoring system integrated with a soft pneumatic exosuit in rehabilitation to enhance the knee extension and walking independence of pediatric CP subjects. The “assist-as-torque” model was established according to the body weight of the subjects to support the knee extension during the gait cycle. The preliminary results of five CP subjects first demonstrated that the peak knee extension increased by 19.72° (±3.47°) under the assistance of the soft pneumatic exosuit. The range of motion improved by 15.46° (± 5.06°) on average at group level. With the more extended knee joint from loading response to mid-stance during the walking gait, the subjects were capable of transferring the reliance on upper limbs and the trunk from the handle frame and suspension to the lower limb. Quantitively, a 33.3% increase in the fraction of body weight was measured by the force plate. The subjects were able to bear significantly more body weight on their affected limb. Additionally, they relied less on the handrail for support during walking. These results highlighted the potential of integrating soft exosuits to improve the motor function and independence of pediatric subjects with impairments such as CP. Quantitation of partial body weight support provided the guidance to implement more personalized and effective assistance in different stages of rehabilitation.