Soft Hip Exoskeleton Reduces Physiological Cost and Perceived Exertion In Older Adults During Uphill Walking

Summary

This protocol evaluates the efficacy of a lightweight soft hip exoskeleton in reducing physiological stress during uphill walking in older adults. Using a randomized crossover design, the method demonstrates that the device significantly reduces physiological cost and subjective fatigue in the elderly.

Abstract

Mountain hiking and mountaineering are common activities for active older adults to maintain cardiovascular fitness; however, the high physiological demands of these activities often accelerate the onset of neuromuscular fatigue. This fatigue is a critical risk factor for sports-related injuries, such as falls and musculoskeletal strains. This study aims to investigate whether a lightweight soft hip exoskeleton can mitigate these risks and support safe physical activity for older adults during simulated mountaineering tasks. Twenty healthy older adults (63.45 ± 3.70 years) participated in a randomized crossover trial on a treadmill set to a 15% incline at 3.5 km/h. Participants walked for 15 min under three conditions: No Exoskeleton (NO_EXO), Exoskeleton Active (EXO_ON), and Exoskeleton Passive (EXO_OFF). Outcomes included Physiological Cost Index (PCI), Peak Heart Rate (HR_Peak), and Rating of Perceived Exertion (Borg RPE). Active assistance (EXO_ON) significantly reduced the Physiological Cost Index to 13.80 ± 1.51, compared to NO_EXO (14.68 ± 2.11) and EXO_OFF (15.35 ± 1.83; p = 0.033). Similarly, RPE was significantly lower with active assistance (12.95 ± 1.15) than with unassisted (13.50 ± 1.19) or passive (14.00 ± 1.17) conditions (p = 0.023). Although Peak Heart Rate was lowest in EXO_ON (133.65 ± 7.74 bpm), differences across conditions were not statistically significant (p = 0.267). The device effectively offsets the physiological burden of its own mass, resulting in a net reduction in cardiovascular effort, as evidenced by lower PCI and RPE. These findings suggest that soft hip exoskeletons may improve walking efficiency and reduce perceived exertion during uphill walking in older adults.

Introduction

As the global population ages, maintaining functional mobility in older adults has become a significant public health challenge¹. Regular aerobic exercise plays a clear protective role in preventing cardiovascular disease, musculoskeletal degeneration, and cognitive decline². Among various physical activities, outdoor hiking and mountaineering are widely adopted leisure activities for older adults due to their combination of social engagement and aerobic conditioning benefits². However, compared to level walking, uphill walking imposes a substantial biomechanical load, requiring lower limb joints to generate greater net positive work to elevate the body’s center of mass³.

For older individuals, this increased biomechanical demand is often accompanied by specific compensatory strategies. Research by Franz and Kram indicates that with advancing age, power output during uphill walking exhibits a significant "distal-to-proximal shift"⁴. Specifically, the propulsive capacity of the ankle plantar flexors diminishes in older adults, who consequently rely more heavily on the hip extensors (gluteus maximus and hamstrings) for work production; the proportion of work performed by the hip is significantly higher than in younger controls^4,5. This compensatory mechanism, superimposed on age-related sarcopenia and reduced cardiopulmonary reserve, results in a higher physiological cost during uphill locomotion in older adults, making them highly susceptible to neuromuscular fatigue⁶.

The accumulation of fatigue not only limits exercise tolerance but is also closely linked to an increased risk of falling⁷. Previous studies have shown that lower limb muscle fatigue disrupts gait stability and reduces minimum toe clearance, thereby significantly increasing the probability of tripping and falling^7,8. Therefore, the development of assistive technologies capable of delaying fatigue onset is of great significance for ensuring the safety of older adults during outdoor activities.

Wearable exoskeleton robotics offer a potential pathway to address this issue⁹. While rigid exoskeletons excel at providing high torque output, their substantial mass often restricts the body's natural kinematic degrees of freedom. This can result in a "physiological penalty" where the physiological cost added by wearing the device offsets or even exceeds the benefits derived from the assistance^10,11. Research by Browning et al. confirmed that adding 1 kg to the distal lower extremities significantly increases physiological cost, a non-negligible burden for older adults with limited physical capacity¹⁰. To overcome this limitation, "soft exosuits" based on flexible textiles and Bowden cable transmission have emerged, designed to minimize interference with natural gait by improving human-machine compliance¹².

Although studies involving young healthy cohorts have confirmed that hip assistance can reduce the physiological cost of level walking¹³, research targeting older adults in the high-physiological-demand scenario of inclined walking remains relatively scarce. Given the increased reliance on hip work during uphill walking in older adults^4,14, this study hypothesizes that a lightweight soft hip exoskeleton, by providing precise hip flexion assistive torque, can effectively overcome the physiological penalty of its own mass, thereby significantly reducing the Physiological Cost Index (PCI) and subjective fatigue (RPE) of older adults during a 15% incline walking task. To test this hypothesis, this study follows a within-subject randomized crossover design. The initial subject preparation and device fitting are performed by unblinded staff, while the formal walking assessment is supervised by a blinded assessor.

Protocol

The study protocol was approved by the Clinical Research Ethics Committee of the First Affiliated Hospital, Zhejiang University School of Medicine (Approval No. 2024-Research-022). All experimental procedures were performed in accordance with the principles of the Declaration of Helsinki.

Subject recruitment and screening
For this study, 20 healthy older adults (aged 60–75) were recruited from The First Affiliated Hospital, Zhejiang University School of Medicine¹⁵. Subjects were screened according to the following inclusion criteria: age 60–75 years; capability for independent walking without assistive devices; ability to walk continuously on a treadmill for ≥15 min; and MMSE score ≥24. Subjects with uncontrolled hypertension (resting BP ≥160/100 mmHg); unstable angina, severe arrhythmia, or myocardial infarction within 6 months; stroke, Parkinson’s disease, severe vertigo, or peripheral neuropathy; lower limb fracture or orthopedic surgery within 12 months; severe osteoarthritis or rheumatoid arthritis; open wounds, ulcers, dermatitis, severe varicose veins, or strap allergies at exoskeleton contact areas; BMI >30 kg/m²; or body dimensions outside the device range were excluded. Written informed consent was obtained, and baseline anthropometric metrics, including height, weight, and resting heart rate, were recorded.

Experimental setup
The treadmill was set to 3.5 km/h and 15% incline. A magnetic safety lanyard was attached from the participant’s belt to the treadmill auto-stop switch, and a supervising researcher maintained continuous access to the manual emergency stop. The exoskeleton device was prepared before testing. Battery charge level was verified to be >90%. The drive units were inspected manually to confirm the absence of cable entanglement, mechanical binding, or excessive friction in the transmission mechanisms. The exoskeleton provided terrain-adaptive assistance modes (Climbing, Downhill, Flat) with 5 intensity levels (up to 18 Nm). Climbing Mode incorporated a gait-learning algorithm and an angle-triggered mechanism (0–90° range) to synchronize hip flexion torque with the wearer’s gait phase. Assistance magnitude, predominantly at levels 3 or 4, was personalized based on subjective tolerance during adaptation. Subjects were fitted with a heart rate monitor for continuous data recording. The Borg RPE scale (6–20) and VAS pain scale (0–10) were printed on A4 paper and positioned within the participant’s clear line of sight to ensure immediate accessibility during testing.

Device donning and fitting (Unblinded phase)
An appropriate device size was selected based on participant height (Size L for height >165 cm; Size M for height ≤165 cm). Subjects were assisted in donning the shoulder straps, and the back module height was adjusted to coaxially align the lateral hip motor units with the greater trochanter. The magnetic waist buckle was fastened, and the lateral adjustment straps were tightened above the iliac crest to anchor the device securely to the pelvis and prevent slippage during actuation. Leg connections were secured. Thigh struts were aligned approximately 5–10 cm above the patella. Hooks were secured, and the Boa dial was rotated clockwise to tighten the thigh straps. Strap tension was verified by maintaining approximately 1.0–1.5 cm clearance between the strap and the skin when pulled perpendicular to the skin. Subjects performed squats and high leg lifts to verify the absence of mechanical interference or excessive cable tension (Figure 1). A 15-min adaptation session¹⁶ was conducted. Subjects walked without the device on the treadmill at 3.5 km/h and 15% incline for 5 min. Subjects walked with the unpowered exoskeleton (EXO_OFF) for 5 min. Subjects walked with the powered exoskeleton (EXO_ON) for 5 min while assistance intensity was gradually increased from level 1 to the subject’s comfort level (levels 3 or 4). Subjects then rested in a seated position for 20 min before the formal trials. The 20-min rest period ensured that heart rate and perceived exertion returned to baseline levels¹⁷.

Intervention and blinding
The order of the three experimental conditions was randomized using a computer-generated sequence to minimize learning and fatigue effects. In the NO_EXO condition, subjects walked in standard sportswear without the exoskeleton. In the EXO_ON condition, the device was activated in Climbing Mode with assist intensity set to level 3 or 4. The exact assistance level was determined by verifying a stable gait rhythm without visible postural instability, compensatory movements, or reported resistance to the device torque. All LED indicators (power and mode lights) were covered with black electrical tape in both EXO_ON and EXO_OFF conditions to maintain visual blinding. A 40-min seated rest period was provided between trials to allow heart rate and fatigue levels to return to baseline.

Simulated uphill walking task (Blinded assessment)
Subjects walked on the treadmill for 15 min and were encouraged to maintain natural walking without holding the handrails unless necessary for safety. Heart rate (HR) was continuously recorded. The 180 heart rate data points collected from min 12:00–15:00 at a sampling rate of 1 Hz were averaged to calculate the steady-state heart rate. The absolute highest heart rate recorded during the task was defined as HR_Peak. At min 14, subjects reported their level of exertion using the Borg RPE scale. Immediately after task completion (within 1 min), the VAS scale was used to assess skin integrity and joint pain. Skin contact areas were visually inspected for erythema, blisters, or pressure marks. Joint pain was recorded using a 0–10 cm VAS scale, where 0 indicated no pain, and 10 indicated the worst imaginable pain. Scores ≥4 were considered clinically significant discomfort requiring further evaluation. The Physiological Cost Index (PCI) was calculated using the formula adapted from MacGregor¹⁸: PCI = (Walking Heart Rate − Resting Heart Rate) / Walking Speed, expressed in beats per minute per kilometer per hour (beats/min/km/h).

Statistical analysis

Outcome variables were compared across conditions using ANOVA, with effect sizes reported as partial eta-squared (η²p). Post hoc pairwise comparisons used Bonferroni-corrected paired t-tests, and Cohen's dz with 95% CIs were calculated (Supplementary Table 1). Condition order was counterbalanced, with rest periods between trials to minimize carryover effects. Significance was set at p < 0.05.

Results

Subject characteristics

A total of 20 healthy older adults (10 male, 10 female) completed the study. The subjects had a mean age of 63.45 ± 3.70 years, a height of 1.65 ± 0.07 m, a weight of 60.51 ± 6.59 kg, and a Body Mass Index (BMI) of 22.13 ± 1.58 kg/m² (Table 1). All subjects successfully completed the three walking conditions (NO_EXO, EXO_ON, EXO_OFF) with no reported adverse events.

Physiological cost index

Repeated-measures ANOVA revealed significant differences in PCI across the three conditions (F = 3.61, η²p = 0.112, a medium effect; p = 0.033). The active assistance mode (EXO_ON) demonstrated the lowest physiological cost, with a PCI of 13.80 ± 1.51 beats/min/km/h, significantly lower than both the unassisted condition (NO_EXO: 14.68 ± 2.11, p = 0.017) and the passive condition (EXO_OFF: 15.35 ± 1.83, p = 0.001); NO_EXO was also significantly lower than EXO_OFF (p = 0.024). These results indicate that the exoskeleton, when activated, effectively mitigated the physiological burden of uphill walking (Figure 2).

Peak heart rate (HR_Peak)

The recorded mean peak heart rates were: 136.30 ± 9.23 bpm for the NO_EXO group, 133.65 ± 7.74 bpm for the EXO_ON group, and 137.85 ± 7.42 bpm for the EXO_OFF group. The omnibus ANOVA was non-significant (F = 1.35, η²p = 0.045, p = 0.267). While exploratory paired t-tests indicated nominally lower peak heart rates in EXO_ON compared to NO_EXO (p = 0.023) and EXO_OFF (p = 0.005), these differences warrant cautious interpretation (Figure 3).

Subjective fatigue rating (RPE)

Subjective fatigue, measured via the Borg RPE scale (6–20), showed a significant improvement with active assistance (F = 4.03, η²p = 0.124, a medium effect; p = 0.023). EXO_ON (12.95 ± 1.15) was significantly lower than EXO_OFF (14.00 ± 1.17, p < 0.001), but not NO_EXO (13.50 ± 1.19, p = 0.265) (Figure 4).

Visual analog scale pain score (VAS)

Overall pain scores were negligible across all conditions: 0.25 ± 0.72 for NO_EXO, 0.15 ± 0.49 for EXO_ON, and 0.30 ± 0.73 for EXO_OFF. Statistical analysis showed no significant difference in pain perception (F = 0.27, η²p = 0.009, p = 0.763), indicating that wearing the exoskeleton did not induce any additional physical discomfort.

Data Availability:

All the datasets obtained from the study are included in the article as Figures, Tables, and Supplementary Table.

Figure 1: Demonstration of correct device alignment with the greater trochanter. Please click here to view a larger version of this figure.

Figure 2: Physiological Cost Index (PCI) results. Comparison of physiological efficiency during 15-min uphill walking under NO_EXO, EXO_ON, and EXO_OFF conditions. Box plots (n=20) display the median (line), IQR (box), and full range (whiskers). Definition: PCI = (Walking HR - Resting HR) / Walking Speed. Significant differences are indicated by * = p < 0.05 and ** = p < 0.01. Please click here to view a larger version of this figure.

Figure 3: Peak Heart Rate (HR_Peak) results. Peak cardiovascular stress during uphill walking trials. Box plots (n=20) display the median (line), IQR (box), and full range (whiskers). Definition: HR_Peak is the absolute highest heart rate observed during the task. Exploratory pairwise comparisons are indicated by * = p < 0.05 and ** = p < 0.01. Please click here to view a larger version of this figure.

Figure 4: Rating of Perceived Exertion (RPE) results. Subjective fatigue (Borg Scale 6–20) was reported at min 14. Box plots (n=20) display the median (line), IQR (box), and full range (whiskers). Definition: RPE ranges from 6 (no exertion) to 20 (maximal exertion). Significant differences are indicated by *** = p < 0.001. Please click here to view a larger version of this figure.

Characteristics	Values
Sample size (n)	20
Gender, n (%)
Male	10 (50.0%)
Female	10 (50.0%)
Age (years, Mean ± SD)	63.45 ± 3.70
Height (m, Mean ± SD)	1.65 ± 0.07
Weight (kg, Mean ± SD)	60.51 ± 6.59
Body Mass Index (BMI, kg/m², Mean ± SD)	22.13 ± 1.58
Resting Heart Rate (bpm, Mean ± SD)	76.55 ± 9.02

Table 1. Baseline characteristics of study participants. Summary of demographic and anthropometric data for healthy older adults (n=20). Values are mean ± SD for continuous variables and frequency (%) for gender. Abbreviations: BMI = Body Mass Index; bpm = beats per minute.

Supplementary Table 1: Raw dataset containing participant demographic, physiological, and subjective outcome data collected under NO_EXO, EXO_ON, and EXO_OFF conditions. Variables include age, sex, BMI, heart rate measures, PCI, Borg RPE, VAS pain score, assist level, and condition order.Please click here to download this file.

Discussion

This study evaluated the impact of a lightweight, soft hip exoskeleton on physiological and psychological metrics in older adults during simulated uphill walking, using a randomized, crossover design. The results show that, compared with the baseline condition (NO_EXO) and the passive wear condition (EXO_OFF), the active assistance mode (EXO_ON) significantly reduced PCI and subjective fatigue rating (RPE). This finding supports our hypothesis that targeted hip joint assistance can effectively reduce physiological stress in older adults during loaded walking tasks.

The core of clinical translation for exoskeleton systems lies in balancing "assistive benefit" against "wearable burden"¹¹. In this study, the PCI in the EXO_OFF condition was slightly elevated compared to baseline, a phenomenon consistent with the classical exposition by Browning et al. regarding the increased physiological cost of adding mass to the limbs¹⁰. Due to the relatively lower muscle strength reserve in older adults, their sensitivity to additional load may be higher than that of younger cohorts¹². However, once active assistance was engaged (EXO_ON), the mechanical work provided by the exoskeleton not only successfully offset the physiological penalty caused by the device's self-weight (2.3 kg) but also achieved a net physiological benefit of approximately 6%. This indicates that the system's "assistive-to-weight ratio" has reached a positive yield threshold. This result aligns with findings by Asbeck et al., who suggested that soft exosuits can achieve effective physiological reductions in older populations by reducing antagonist co-contraction and optimizing force transmission efficiency¹². Furthermore, a review by Tang et al. points out that compared to distal (ankle) loading, the physiological penalty induced by proximal (waist/hip) loading is significantly lower because it has a smaller impact on the limb's moment of inertia¹⁹. The device design in this study concentrates the battery and drive units at the waist, strictly adhering to this biomechanical principle to maximize walking economy. Gregorczyk et al. also further confirmed that minimizing distal lower limb burden is key to maintaining normal gait dynamics during loaded walking²⁰.

The benefits observed in this study may be attributed to the high degree of matching between the assistance strategy and the uphill gait characteristics of older adults. Research indicates that during uphill walking, the body adjusts its joint moment distribution strategy, with hip flexion becoming the primary power source for overcoming gravity¹⁴. Due to the decline in ankle power output, older adults are compelled to adopt a "hip-dominant" compensatory strategy^4,5. Our device provides hip flexion torque during the initial swing phase, directly assisting the iliopsoas and rectus femoris. Kim et al. noted that in specific high-power-demand tasks (such as uphill walking), assisting proximal joints (hips) may be more metabolically economical than assisting distal joints (ankles) because it helps reduce the activation of large muscle groups¹³. Our data provides empirical support for the applicability of this theory in the older adult population.

Jin et al., in studies on long-distance and inclined walking, also found that soft hip assistance effectively reduces cumulative muscle load during walking, particularly during the uphill phase, where the work done against gravity increases significantly²¹. Additionally, according to the "Human-in-the-loop" optimization theory by Ding et al., targeted hip flexion assistance can effectively optimize energy flow throughout the gait cycle, enabling subjects to maintain the same walking speed while significantly reducing physiological energy expenditure²². This precise, phase-specific assistance avoids interference with the swing phase, thereby ensuring human-machine kinematic consistency.

Beyond physiological metrics, the significant reduction in subjective fatigue (RPE) (p < 0.05) holds important clinical significance. Older adults often limit physical activity due to the perception of "effort"²³. A reduction in RPE indicates that, under the same physical load, the user's perceived stress is reduced. Furthermore, neuromuscular fatigue is considered an independent risk factor leading to increased gait variability and elevated fall risk in older adults^7,8. Shin et al., in a randomized controlled trial of robot-assisted walking for older adults, reported that reducing the physical demands of walking via external devices can significantly improve gait parameter stability and enhance psychological exercise confidence (Self-efficacy)²⁴. Research by Webster et al. also indicates that fatigue leads to deteriorated biomechanical control upon lower limb landing (e.g., increased knee valgus moment), thereby increasing injury risk²⁵. Thus, the observed reductions in RPE and physiological cost may help mitigate fatigue-related biomechanical risks, though long-term clinical trials are needed to confirm effects on gait stability and injury prevention.

The assist intensity (primarily levels 3–4) was self-selected for comfort rather than systematically varied. This user-driven approach balances safety and physiological efficiency. Low torques (levels 1–2) often fail to overcome the device's mass penalty²⁶, whereas excessive torque risks gait instability and antagonist co-contraction in older adults²⁷. Therefore, self-selected intermediate levels represent an empirical optimal balance point, delivering physiological benefits while maintaining kinematic stability²⁸. Consequently, the exact dose-response relationship between assist intensity and physiological expenditure remains unmeasured.

Although active assistance (EXO_ON) significantly reduced PCI, the difference in Peak Heart Rate (HR_Peak) across conditions was not significant. This lack of significance is likely attributable to the experiment's duration and the cohort's characteristics. The 15-min submaximal protocol allows participants to reach a steady-state cardiovascular response but likely lacks the intensity to drive the heart rate to its absolute physiological limits²⁹. Furthermore, high inter-subject baseline variability among older adults further masks subtle HR_Peak changes. To fully evaluate cardiovascular benefits, future studies should employ longer testing durations and monitor Heart Rate Recovery (HRR)³⁰.

This study has certain general limitations. First, while the treadmill walking environment is controllable, it cannot fully simulate the complexity of real outdoor terrain, which may limit the validity of the results³¹. Second, subjects in this study underwent only short-term adaptation training. Previous longitudinal studies suggest that the nervous system's adaptation to assistive torque is a gradual process²¹. Third, the small sample size (n = 20) and absence of age stratification limit generalizability. Because aging progressively reduces muscle and cardiopulmonary reserves³², assistive efficacy may vary across older subgroups, warranting larger age-stratified trials. Finally, given the heterogeneity of individual gait, future research should aim to develop "Human-in-the-loop" optimization control algorithms to dynamically adjust assistance profiles^31,33, and employ portable metabolic analyzers to precisely measure direct metabolic cost.

In conclusion, this study demonstrates that a lightweight, soft hip exoskeleton can provide effective physiological assistance during uphill walking for older adults. This finding provides a preliminary scientific basis for applying soft robotics technology to elderly mobility assistance and fatigue management.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Borg RPE Scale	N/A	6-20 Scale	Standardized scale for assessing Rating of Perceived Exertion.
GraphPad Prism	GraphPad Software	Version 9.0	Used for data visualization and generating boxplots.
Medical Weight/Height Scale	Meilen	M-C-MSG100	Used for baseline physical characteristic measurements (BMI).
Soft Hip Exoskeleton (GOGO-H)	Hangzhou RoboCT Technology Development Co., Ltd	100189510835	Lightweight soft robotic suit (2.3 kg) providing hip flexion assistance.
SPSS Statistics Software	IBM	Version 26.0	Used for statistical analysis.
Standard Motorized Treadmill	YPOO	M5max	Used for simulated uphill walking at a 15% incline and 3.5 km/h.
Visual Analog Scale (VAS)	N/A	10 cm line	Used for subjective assessment of lower limb pain and discomfort.
Wearable smart band	Xiaomi Communications Co., Ltd.	6932554419790	Wearable smart band used for real-time heart rate and blood oxygen monitoring.