{"paper_id":"4bd2b42f-dfae-430b-b4a9-62837cdce5e1","body_text":"Biomimetic Self-Powered Smart Insole with AI-Enhanced Mechanodiagnosis for Continuous Gait Monitoring | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Biomimetic Self-Powered Smart Insole with AI-Enhanced Mechanodiagnosis for Continuous Gait Monitoring Feng Xu, Yingchun Li, Yarong Ding, Yuze Zhang, Xing Guo, Kaixin Lei, and 15 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6314292/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 07 Jan, 2026 Read the published version in Research → Version 1 posted You are reading this latest preprint version Abstract Continuous gait analysis is essential for early detection and management of neuromuscular disorders, yet current wearable technologies face limitations in sensing capacity, energy autonomy, and real-time diagnostic capabilities, restricting their clinical adoption. Here, we present a biomimetic smart insole that synergizes nature-inspired sensing, self-sustaining energy harvesting, and artificial intelligence (AI) to enable continuous, clinically actionable gait monitoring. Mimicking the mechanosensory architecture of mantis legs, our dual-microstructure capacitive sensor achieves a sensitivity of 0.602 kPa ⁻ ¹, a detection limit of 0.10 Pa, and a broad sensing range (0.10 Pa–1.40 MPa) with exceptional durability (>12,000 cycles), outperforming state-of-the-art wearable sensors. A custom-designed flexible circuit wirelessly streams 16-channel pressure data to a companion APP, providing real-time visualization of dynamic force fields through chromatic mapping. The system’s energy autonomy is ensured by a hybrid perovskite solar cell/lithium-sulfur battery, enabling continuous operation across diverse environments. An embedded AI framework combines a random forest classifier (96% accuracy in foot arch abnormality detection) with a convolutional neural network (97.6% accuracy in classifying 12 pathological gait patterns), translating raw sensor data into clinical insights. This platform bridges the gap between wearable sensing and precision diagnostics, offering transformative potential for early disease detection, personalized rehabilitation, and telemedicine, and thus establishing a paradigm for next-generation intelligent wearables in global healthcare. Physical sciences/Engineering/Electrical and electronic engineering Physical sciences/Materials science Physical sciences/Energy science and technology/Renewable energy/Solar energy Health sciences/Medical research Physical sciences/Engineering/Biomedical engineering Flexible pressure sensors Closed-loop power supply Intelligent insole Gait monitoring AI-assisted mechanodiagnosis Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction The growing prevalence of age-related mobility decline, chronic musculoskeletal disorders (e.g., Parkinsonian gait, diabetic neuropathy, osteoarthritis), and structural foot abnormalities (e.g., flatfoot, high-arched foot) underscores the urgent need for wearable systems capable of continuous, biomechanical monitoring of pathological gait patterns (staggering gait, asymmetric strides) 1,2 . Gait analysis serves as a vital biomarker for neuromuscular health, offering insights into disease progression and rehabilitation efficacy 3,4 . However, clinical gait assessment remains tethered to laboratory environments due to the reliance on optical motion-capture systems and force plates, which are cost-prohibitive, spatially constrained, and incapable of capturing real-world walking dynamics 5,6 . Wearable pressure-sensing insoles could decentralize this process, yet existing solutions fail to meet three interconnected requirements of clinical viability, i.e. , sensors with durability and resolution to span the full biomechanical force spectrum, energy autonomy for uninterrupted operation, and real-time interpretation of pathological gait signatures 7-9 . Current flexible pressure sensors face fundamental trade-offs in performance. The plantar interface subjects devices to extreme mechanical demands: transient forces exceeding 1 MPa during athletic activity, subtle shifts (<1 Pa) during postural adjustments, and cyclic loading over thousands of daily steps, a regime that rapidly degrades conventional sensors with homogeneous or single-scale microstructures 10,11 . While bioinspired designs, such as those mimicking gecko feet (hierarchical stress distribution) or spider slit sensilla (vibration sensitivity), have improved dynamic responsiveness 12,13 , these systems often sacrifice the mechanical robustness (>10,000 cycles) or biocompatibility for prolonged use. Energy autonomy poses a parallel challenge. High-density sensor arrays and wireless data transmission demand significant power, yet bulky batteries or frequent recharging disrupt continuous monitoring, a prerequisite for capturing episodic gait abnormalities 14,15 . Although emerging solutions like biofuel cells and triboelectric nanogenerators (TENGs) have been explored for battery-free wearables, biofuel cells degrade with sweat exposure, while TENGs require vigorous motion and generate high-voltage, low-current outputs incompatible with modern wearable electronics 16-19 . Recent advances in perovskite solar cells (PSCs), which achieve high power conversion efficiency under indoor lighting while maintaining flexibility, offer promise for self-sustaining wearables 20-22 . Pairing PSCs with high-energy-density storage units like lithium-sulfur (Li-S) batteries could enable closed-loop power systems resilient to environmental fluctuations 23 . However, no prior work has successfully integrated these technologies to deliver seamless operation across real-world environments, from low-light indoor settings to variable outdoor conditions. Beyond hardware limitations, translating raw sensor data into actionable clinical insights remains a major challenge. Although machine learning algorithms achieve high accuracy in classifying gait abnormalities under controlled conditions 24 , most wearable systems lack real-time processing, intuitive visualization, and adaptive feedback, which are essential for patient adherence and timely clinical decision-making 25 . A transformative platform must not only capture dynamic plantar pressure maps but also interpret them through AI models trained on heterogeneous pathological gaits, all of which should operate autonomously at the point of care. Here, we introduce a biomimetic smart insole that addresses these challenges through three synergistic strategies ( Fig. 1 ). First, inspired by the hierarchical mechanosensory architecture of mantis legs, we developed a dual-microstructure capacitive pressure sensor combining polydimethylsiloxane (PDMS) microstructures with compressible foam. This design achieves an ultralow detection limit (0.10 Pa), high sensitivity (0.602 kPa⁻¹), and a broad dynamic range (0.10 Pa–1.40 MPa), with >12,000-cycle durability surpassing existing wearable sensors. Second, a hybrid perovskite solar cell/lithium-sulfur (PSC/Li-S) energy system ensures uninterrupted operation across diverse environments, eliminating dependency on external power. Third, an embedded AI framework synergizes a random forest classifier for foot arch abnormality detection (96% accuracy) and a one-dimensional convolutional neural network (1D-CNN) for classifying 12 pathological gait patterns (97.6% accuracy). A companion mobile APP visualizes dynamic force fields through chromatic mapping, empowering clinicians with real-time, interpretable diagnostics. By integrating high-fidelity biomechanical sensing, self-sufficient energy harvesting, and edge-computing intelligence, this platform bridges a critical gap between wearable technology and clinical practice. Its ability to provide continuous, AI-augmented gait analysis outside laboratory settings advances preventive care paradigms, from reducing fall risks in elderly populations to enabling personalized rehabilitation in telemedicine frameworks. Dual microstructure design achieves clinically relevant sensitivity across biomechanical pressure ranges. To enable continuous detection of gait-related pressures from subtle postural shifts to high-impact forces, we engineered a capacitive pressure sensor as inspired by the hierarchical elasticity of mantis legs ( Fig. 1 ). The sensor integrates thermoplastic polyurethane (TPU) foam doped with poly(3,4-ethylenedioxythiophene) (PEDOT) via vapor-phase polymerization (VPP) and PDMS pyramid microstructures on both surfaces ( Fig. S1 ). Energy-dispersive spectroscopy (EDS), Fourier-transform infrared (FTIR) spectroscopy, and Raman spectroscopy confirm homogeneous PEDOT distribution within the TPU matrix ( Fig. S2, Fig. S3 ). This doping strategy increases the foam’s Young’s modulus by 38% and dielectric constant by 4.2× compared to undoped TPU ( Supplementary Note 1 ), establishing a mechanically robust foundation for pressure sensing. To optimize stress distribution and dielectric modulation, we patterned PDMS pyramids (14 µm width, 9 µm height, 7 µm spacing) onto the PEDOT-doped foam ( Fig. 2a ). Finite element simulations reveal that dual microstructures amplify localized deformation by 72% under load compared to flat designs, enhancing the air gap effect and dielectric polarization ( Fig. 2b , Supplementary Note 2, Video S1 ). Experimentally, sensors with double-sided pyramids achieve a sensitivity of 0.602 kPa⁻¹, which is 2.8× higher than non-patterned sensors (0.213 kPa⁻¹) and 1.9× higher than single-sided designs (0.318 kPa⁻¹) ( Fig. 2c, Fig. S7 ). The working range extends to 1.4 MPa, exceeding the peak plantar pressures observed during walking (1.2 MPa) 26 . These results demonstrate that our biomimetic microstructures synergistically enhance sensitivity and biomechanical compatibility. To address the clinical need for detecting subtle pressure variations and transient forces, we quantified sensor response times under physiologically relevant loads. The device resolves pressure changes as low as 0.10 Pa ( Fig. 2d ) and exhibits rapid response/recovery times of 135 ms (0.5 kPa), 167 ms (1 kPa), and 192 ms (5 kPa) ( Fig. 2e , Fig. S9 ), outperforming existing wearable sensors 27-37 . Cyclic loading tests (100 kPa at 2 Hz) confirm stable performance over 12,000 cycles with <3% signal drift ( Fig. 2f ), meeting the durability requirements for multi-day gait monitoring. We benchmark our sensor against 10 state-of-the-art flexible pressure sensors 27-37 across four metrics critical for clinical use, including sensitivity (0.602 kPa⁻¹), detection limit (0.10 Pa), response time (135 ms), and working range (0.10 Pa–1.40 MPa) ( Fig. 2g , Table S1 ). The device achieves top-tier performance in combined sensitivity and dynamic range, while ensuring biocompatibility (ISO 10993-5 certified) and maintaining mechanical flexibility. This performance profile enables continuous monitoring of pathological gait patterns, such as diabetic foot ulcer precursors and osteoarthritis-related joint overloading. By combining biomimetic microstructure engineering with conductive polymer doping, we developed a capacitive pressure sensor that overcomes key limitations in wearable gait analysis. The dual-microstructure design achieves clinically relevant sensitivity and durability, while the 0.10 Pa detection limit identifies subclinical biomechanical anomalies. Dual-microstructure sensors withstand extreme mechanical stress and gradient pressure. To evaluate sensor durability under real-world mechanical stress, we affixed a 1 cm × 1 cm sensor to a car tire tread, subjecting it to 300 kPa compressive and 6 kPa shear forces during continuous driving ( Fig. 3a ). We observed stable wireless capacitance signals during prolonged continuous testing, with no signal drift or anomalies ( Video S2 ). Post-test scanning electron microscopy (SEM) reveals intact TPU/PEDOT foam pores and undamaged PDMS pyramids ( Fig. 3b ), confirming the dual-microstructure design resists delamination and deformation under extreme loads. These results demonstrate the sensor’s robustness for wearable applications requiring prolonged exposure to harsh biomechanical forces, such as athletic monitoring or industrial exoskeletons. To demonstrate gradient pressure sensing capability and clinical-grade sensitivity, we further tested the sensor’s response to subtle pressures ( e.g. , acoustic waves, airflow), clinically relevant middle pressures ( e.g. , pulse), and angular-dependent pressures from finger bending ( Fig. 3c, Supplementary Note 3 ). The sensor demonstrates excellent responsiveness to a 100 dB acoustic wave and airflow at speeds below the human skin’s detection threshold (≈0.50 m s⁻¹). The system also resolves carotid pulse harmonics (0.50–20 Hz) and distinguishes ankle pulse amplitude variations (ΔC: 0.80–1.20 pF), which is linked to peripheral artery disease 38 . During gesture recognition, capacitance shifts enable precise differentiation of finger flexion angles (30°–120°), with signal output showing a strong linear correlation (R² = 0.98). These results indicate the sensor’s ability to detect subclinical biomechanical anomalies, such as early-stage neuropathy or joint stiffness, with precision rivaling laboratory-grade instruments. To ensure compatibility with long-term wearable use, we performed cytotoxicity assays using L929 fibroblast cells and skin irritation tests per ISO 10993-5 standards ( Fig. S10 ). We observed no reduction in cell viability (>95% survival) or erythema in epidermal models after 72-hour exposure. These findings validate the sensor’s safety for uninterrupted skin contact, addressing a critical barrier to clinical adoption of wearable electronics. To validate spatial resolution for medical diagnostics, we fabricated a 4 × 4 array and map pressure distributions from lightweight objects (1–10 Pa). The array accurately identified the spatial pressure profiles of origami cranes and leaf positioned at distinct locations ( Fig. 3d,e ) This spatial recognition capability highlights the platform’s potential for adaptive prosthetics, pressure ulcer prevention, and robotic tactile sensing. By unifying mechanical resilience, gradient pressure sensing capability, and biocompatibility, our dual-microstructure sensors overcome key limitations in current wearable biomechanical monitors. The 300 kPa durability ensures reliable operation under strenuous activities, while 0.10 Pa sensitivity enables early detection of neuromuscular degradation, a critical advance for preventive healthcare. Integrated photo-rechargeable system sustains continuous operation of wearable biomedical devices. To enable autonomous, long-term operation of wearable health monitors, we developed a closed-loop energy system that synergizes PSCs with Li–S batteries ( Fig. 4a ). The power management system dynamically adapts operational modes through an adaptive energy allocation strategy responsive to environmental illumination. Under sufficient outdoor solar irradiance, PSCs simultaneously charge Li-S batteries and power sensor-integrated circuits; when sunlight diminishes below operational threshold, photovoltaic supply prioritizes sensor circuit demands while preserving Li-S energy reserves. Indoor operation transitions to Li-S discharge mode for sustaining power supply of continuous sensing. This tri-mode regulation ensures uninterrupted cross-environment sensing. The PSCs, fabricated on indium tin oxide (ITO)-coated glass or PET substrates with silver electrodes, achieve a power conversion efficiency (PCE) of 21.3% under standard solar illumination (AM1.5G, 100 mW cm⁻²) through optimized charge transport layers of SnO₂ and spiro-OMeTAD ( Fig. 4b,c ). This configuration delivers a short-circuit current density ( J SC ) of 23.10 mA cm⁻² and an open-circuit voltage ( V OC ) of 1.15 V, providing sufficient energy density to power continuous gait monitoring. To assess the energy transfer for wearable applications, we connected four PSCs in series (3.88 V output) to Li–S batteries with sulfur-infused carbon electrodes ( Supplementary Note 4 ). Galvanostatic charge–discharge cycles reveal nearly identical voltage profiles between PSC-charged and DC-powered systems ( Fig. 4d ), confirming efficient solar energy conversion. The PSC-charged batteries retain 99.7% of their initial discharge capacity (170.1 mA h g⁻¹) after five cycles, matching conventional charging methods ( Fig. 4e ). After repeated photocharge–discharge cycles, the Li–S batteries maintain >95% capacity retention and the PSCs show <5% PCE degradation ( Fig. 4f , Supplementary Note 4 ). These results confirm the system’s resilience to fluctuating environmental conditions, ensuring uninterrupted power for sensors, wireless modules, and edge-computing units in smart insoles. The self-charging system achieves an average energy storage efficiency of 72.15% and an average overall photocharging conversion efficiency of 11.14% ( Fig. 4g ). The stability of PSCs is important for the self-charging system. We exposed the unencapsulated PSC to air with a relative humidity of 40 % at room temperature. With 94% efficiency retention over 1000-hour aging ( Supplementary Note 4 ), the PSCs enable reliable energy autonomy for wearable systems, achieving a key milestone toward clinical-grade health monitoring. To demonstrate practical utility, we drived 14 commercial LEDs using eight series-connected PSCs, achieving stable illumination under simulated sunlight ( Fig. 4h , Video S3 ). By integrating high-efficiency PSCs with durable Li–S batteries, we overcome the energy autonomy challenge in wearable biomechanical monitoring. The system’s high photoelectric conversion efficiency and long-term runtime enables continuous gait analysis across diverse real-world settings, from low-light clinical environments to outdoor activities. This self-sustaining platform eliminates reliance on external charging, advancing telehealth applications such as remote diabetic foot ulcer prevention and Parkinson’s disease progression tracking. Integrated smart insole system enables real-time gait monitoring for clinical applications. To address the unmet need for continuous, clinically actionable gait analysis, we integrated 16 high-performance capacitive pressure sensors into a self-powered smart insole ( Fig. 5a ). The system combines a sustainable energy module (PSCs/Li–S batteries), wireless data transmission, and a mobile interface for real-time feedback, achieving seamless operation indoor and outdoor, with no interruptions in data acquisition. These results indicate the platform’s readiness for clinical deployment in rehabilitation and remote patient monitoring. To ensure precise detection of gait abnormalities, we positioned sensors at biomechanically critical regions: toes (2 sensors), forefoot (6 sensors), midfoot (4 sensors), and heel (4 sensors) ( Fig. 5b ). Each 1 cm² sensor, separated by a TPU/PEDOT/PDMS dielectric layer, resolves pressure gradients as low as 0.10 Pa. During walking trials, the array captures heel-strike forces (800–1200 kPa) and midfoot pressure shifts (<10 kPa) with high correlation to laboratory force plates. These findings validate the sensor layout’s capacity to diagnose conditions like diabetic foot ulcers and osteoarthritis through localized pressure anomalies. To enable real-time diagnostics, we connected the sensor array to a flexible printed circuit board (FPCB) with Bluetooth 5.0 ( Fig. 5c ). The system wirelessly streams data to a mobile app, which generates dynamic heatmaps that visually encode pressure variations through color gradients ( Fig. 5d , Video S4 ). The platform has potential to replace intermittent clinic visits with continuous, telemedicine-enabled care. To eliminate reliance on external power, we affixed PSCs and Li–S batteries to the shoe exterior, with the control circuit and APP additionally powered by solar energy for system testing purposes ( Fig. 5e,f ). The PSCs maintain 21.3% efficiency under standard solar illumination, charging the batteries to 170 mA h g⁻¹ ( Fig. 4e ). These metrics confirm the energy module’s suitability for real-world use, including rural or low-resource settings where charging infrastructure is limited. To validate clinical utility, we tested the insole during yoga and simulated falls. During yoga, it distinguishes balance shifts between poses ( e.g. , warrior II and tree) with high accuracy, matching commercial pressure pads ( Fig. 5g) . The system employs real-time pressure distribution analysis to detect and issue directional fall warnings for forward, backward, leftward, and rightward movements ( Fig. 5h ). These results demonstrate performance comparable to commercial pressure mats, underscoring the system's dual functionality in both fall prevention and personalized rehabilitation. To assess performance under dynamic loads, we analyzed stair climbing and marching in place ( Videos S5 , S6 ). The insole tracks pressure evolution across 32 foots regions, with capacitive signal heatmaps and real-time curves closely matching actual pressure patterns ( Supplementary Note 5 ). Stable and precise signal outputs under high-impact and rapid movements demonstrate its reliability for sports science, injury prevention, and athletic training applications. By unifying biomimetic sensing, self-sufficient energy harvesting, and AI-driven analytics, this smart insole advances wearable biomechanics beyond episodic measurements. The system’s correlation with lab-grade tools and long-term autonomy positions it as a transformative tool for preventive care, enabling early intervention in diabetic neuropathy, reducing fall risks in elderly populations, and optimizing athletic training protocols. AI-assisted mechanodiagnosis in static foot arch abnormality detection and dynamic gait pattern recognition using smart insole data To enable automated, clinical-grade analysis of plantar pressure distribution, we integrated a machine learning framework with a 16-sensor smart insole system. A random forest model was trained on 500 static pressure profiles (35% high arch, 35% flatfoot, 30% normal) collected during upright standing, achieving 96% classification accuracy on an independent test set. The model identifies characteristic pressure deviations in flatfoot and high-arched feet, showing strong correlation with clinical diagnoses from podiatrists and laboratory force plates ( Fig. 6a ). To demonstrate the analytical effects of AI, a color-coded confusion matrix ( Fig. 6b ) depicts high diagnostic precision, with pixel intensity representing classification confidence. We investigated the impact of readout layer size on recognition accuracy, revealing that reducing network size decreases accuracy, yet a ∼95% accuracy is maintained even at a network size of 65 ( Fig. 6c ). t-SNE visualization of multidimensional pressure data ( Fig. 6d ) revealed three distinct clusters corresponding to arch types, validating the model's capacity to resolve subclinical biomechanical signatures for early intervention in structural foot abnormalities and stress fracture prevention. To resolve gait time-dependent recognition issues, we engineered the 1D-CNN to analyze time-series capacitance data from 12 gait types, including slow walk (SW), fast walk (FW), marching in place (MP), Left limp (LL), Right limp (RL), Shuffling gait (SG), Dragging walk (DW), kicking walk (KW), toe walking (TW), heel walking (HW), foot eversion (FE), foot inversion (FI). Figure 6e presents the capacitive signal outputs acquired by the smart insole during the identification of distinct gait patterns, alongside a schematic illustration of the 1D-CNN framework used for gait classification. The gait recognition system based on 1D-CNN algorithm exhibits a classification accuracy of up to 97.6%，demonstrating excellent performance in distinguishing 12 pathological and physiological gait patterns ( Fig. 6f ). From the model training process, it can be observed that the model converges rapidly, achieving an accuracy of ~97% after around 45 epochs (Fig. 6g) , indicating both efficient training and strong classification capability. Fig. 6h further demonstrates the t-SNE dimensionality reduction outcomes, with well-defined clustering patterns for the 12 gait types, offering promising applications in the early diagnosis of diabetic neuropathy, osteoarthritis, and stroke-related gait asymmetry. This AI-enhanced platform transforms raw plantar pressure data into clinical-grade diagnostics, detecting structural foot arch abnormalities and dynamic gait anomalies through biomimetic sensing-adaptive machine learning integration. The dual-modality system achieves 96% accuracy for static arch classification and 97.6% accuracy for 12 gait pattern recognition, surpassing the accuracies of other algorithms ( Fig. 6f, Supplementary Note 6 ), which demonstrates capabilities in early subclinical gait deviation detection (e.g., prodromal Parkinsonian gait) and personalized rehabilitation monitoring (e.g., asymmetric gait quantification). By bridging laboratory biomechanics with clinical workflows, it establishes a precision mechanomedicine framework for intelligent wearables, addressing preventive care gaps in neuromuscular disorders through telemedicine-enabled interventions like real-time fall-risk alerts. Discussion This study establishes a closed-loop wearable platform that redefines the paradigm of continuous gait analysis by integrating biomimetic sensing, energy autonomy, and clinically actionable AI. Our work advances three critical frontiers in wearable biomedical engineering, i.e. , achieving laboratory-grade diagnostic precision in real-world settings, enabling self-sustaining operation across diverse environments, and bridging the gap between raw biomechanical data and precision rehabilitation, previously unrealized in existing systems. Bioinspired dual-microstructure sensors achieve clinical-grade sensitivity in real-world settings. The dual-microstructure capacitive sensor represents a breakthrough in wearable mechanodiagnostics. Current wearable sensors for gait analysis lack the sensitivity and durability required to detect early-stage neuromuscular or musculoskeletal pathologies. By mimicking the hierarchical stress distribution of mantis legs, which are evolved to withstand dynamic loads while retaining sensitivity 39 , we resolve a fundamental trade-off in flexible sensors: achieving sub-0.1 Pa resolution while maintaining robustness under extreme pressures (> 1.40 MPa). The synergy of PEDOT-doped TPU foam (enhancing dielectric constant from 3.38 to 5.08) and PDMS pyramid arrays (amplifying air-gap effects) enables a sensitivity of 0.602 kPa⁻¹ and a dynamic range up to 1.40 MPa (Fig. 2 c). Critically, these sensors retain > 95% accuracy after 12,000 loading cycles (Fig. 2 f), making them combine clinical-grade resolution with long-term reliability for continuous monitoring. This breakthrough addresses a key barrier in mechanomedicine, i.e. , detecting subclinical gait deviations in disorders like diabetic neuropathy or Parkinson’s disease before irreversible damage occurs 40 . Hybrid energy autonomy enables uninterrupted monitoring across diverse environments. Our hybrid PSC/Li–S system overcomes the energy bottleneck that has long constrained wearable diagnostics, particularly for elderly or mobility-impaired users 41 , 42 . Our closed-loop PSC/Li–S system overcomes this by delivering 21.3% PCE (Fig. 4 c) and 72.15% average energy storage efficiency for long-term operation in darkness, which surpass TENG and biofuel-cell alternatives which remain impractical for chronic use 43 , 44 . The PSCs maintain 21.3% efficiency under standard solar illumination (100 mW cm⁻²), while the Li–S batteries provide stable 3.88 V output, ensuring continuous data acquisition during daily activities (Fig. 4 f). By enabling continuous operation in sunlight-limited indoor settings and harsh outdoor environments, this platform overcomes a critical limitation of conventional energy harvesting systems (Fig. 5 a). This energy autonomy is transformative for chronic care 45 – 47 , aligns with the WHO’s Sustainable Development Goals by providing a low-cost (< $ 50 projected at scale), maintenance-free solution for resource-limited settings. Explainable AI transforms biomechanical data into actionable clinical decisions. The dual-model architecture, random forest for static foot arch analysis (96% accuracy) and 1D-CNN for dynamic gait classification (97.6% accuracy), surpasses the accuracies of other algorithms (Fig. 6 f, Supplementary Note 6 ) 48 , 49 . The algorithm’s high accuracy and lightweight architecture (~ 1.6 MB) helps with data locally on smartphones, enabling real-time fall-risk alerts. By correlating plantar pressure maps with known pathological signatures (e.g., Parkinsonian gait, Diabetic plantar ulcer), the system has potential to provides explainable risk scores for falls or ulceration, enabling proactive care. Future iterations could integrate inertial sensors to correlate pressure maps with joint kinematics, further personalizing interventions 50 – 52 . Toward scalable mechanomedicine: clinical integration and global health equity. Our platform’s impact extends beyond individual diagnostics. By wirelessly transmitting HIPAA-compliant reports to telehealth platforms, this system is expected to enable clinicians to remotely monitor more patients without compromising care quality, addressing a critical need in aging populations 42 , 53 . Pilot deployments in rural clinics will reduce fall-related hospitalizations, validating its socioeconomic impact. However, three challenges must be addressed to realize global scalability: (1) expanding sensor coverage to kinetic chains (e.g., hip-knee-ankle dynamics), (2) optimizing PSC/Li–S stability under tropical humidity (> 80% RH), and (3) validating AI generalizability across diverse populations. Collaborative trials with orthopedic centers are underway to integrate the insole with EHRs, enabling predictive analytics for conditions like osteoarthritis progression. Besides, ongoing industry partnerships are advancing scalable manufacturing (< $ 0.12/sensor) and regulatory approval pathways. Conclusions Overall, this work establishes a new paradigm for wearable biomedical devices by integrating biomimetic design, sustainable energy systems, and clinical-grade AI. The platform’s ability to detect subclinical gait pathologies, empower remote monitoring, and reduce hospitalization costs positions it as a transformative tool for global precision medicine. As healthcare pivots toward preventive and personalized models, our closed-loop approach provides a scalable blueprint for transforming wearables into essential medical devices. By unifying real-time biomechanical data acquisition, chromatic force-field visualization, and explainable AI analytics, this work transitions wearable technology from passive monitoring to actionable clinical decision-making. Its energy-independent design and scalability address critical barriers to adoption in underserved populations, where gait disorders are prevalent but underdiagnosed. Future efforts will focus on large-scale longitudinal validation of fall-risk prediction and rehabilitation tracking, as well as integration with digital health ecosystems for predictive care. This study pioneers a paradigm shift in precision healthcare, demonstrating how intelligent, autonomous wearables can democratize access to clinical-grade biomechanical diagnostics and advance global health equity. Methods Material . TPU foam was obtained from Suzhou Shensai New Materials Co., Ltd. Ferric chloride (FeCl 3 ) and ethanol were purchased from Tianjin Damao Chemical Reagent Co., Ltd. PDMS was obtained from Dow Corning Co., Ltd. 3,4-ethylene dioxythiophene (EDOT) and Methanol were purchased from Sigma-Aldrich Chemical Co., Ltd. Preparation of PEDOT/TPU foam . To prepare the oxidant solution, 0.2 g of FeCl 3 was mixed with 4 mL of methanol and stirred. The TPU foam was cleaned with anhydrous ethanol and deionized water, then soaked in the oxidant solution for 15 min. After removing and heating at 70°C for 15 min, a device for vapor phase polymerization was set up, fixing the TPU foam on a petri dish lid and dripping EDOT monomer onto the bottom filter paper. The device was heated at 75°C for 2 h, flipping the foam after 1 h for even polymerization. The TPU/PEDOT foam was then cleaned with anhydrous ethanol and deionized water to remove excess monomers, oxidant solution and other impurities. Preparation of the dielectric layer with micropyramid array . PDMS precursor and curing agent were mixed 10:1 and spin-coated on a photolithography silicon template with a micropyramid pattern at 1000 rpm for 60 s. TPU/PEDOT foam was adhered to the PDMS and cured at 80°C for 3 h. Single-sided and double-sided micropyramid array PEDOT/TPU foams were prepared in turn using the aforementioned method. Electrodes and a dielectric layer were assembled using copper foil and the micropyramid array PEDOT/TPU foam, encapsulated with breathable 3M Tegaderm tape. Characterization and measurement of the capacitive pressure sensor . Fourier Transform Infrared Spectroscopy (FTIR, NEXUS870, Thermo Nicolet) and Raman Spectrometer (Xplora plus, HORIBA) were used to analyze chemical bonds and chemical structures of the composite materials. Scanning electron microscopy (SEM, S-570, Hitachi) was applied for characterizing the morphology of the materials. The mechanical and electrical properties of the sensors were analyzed simultaneously by a universal testing machine (3365, Instron) and a precision LCR meter (TruEbox-01RC, LinkZill). This LCR meter was miniaturization, portable, and featured a wireless connection. Cell viability and staining assay . Cell viability was assessed using CCK-8 cell viability assay kit (Beyotime, Shanghai, China) with L-929 cells seeded in 96-well plates at a density of 2×10 3 cells per well and incubated with material extract. Optical density was measured at 450 nm using a microplate reader (Biotek, Burlington, VT, USA) after 1, 2, and 3 days. AO and EB stained live and dead cells, respectively, in L-929 cells (a density of 2×10 5 cells per well) cultivated in 6-well plates with material extract for 1, 2, and 3 days, observed under a Nikon fluorescence microscope (Japan, Ti-S). Solar cell fabrication and characterization . The PSCs were deposited on the indium tin oxide (ITO) glass or PET substrates (2×2.5 cm 2 , 10 Ω per square). The cleaned ITO substrates were treated by UV-ozone for 20 min. Then the SnO 2 film was deposited on the ITO substrate by spin-coating the SnO 2 precursor (5%) at 4000 rpm for 30 s and annealed on a hot plate at 150°C for 30 min in air. After cooling down to room temperature, the substrate was treated with UV-ozone for 10 min before transferring to a nitrogen-filled glove box. For the perovskite film preparation, 691.5 mg PbI 2 and 19.5 mg CsI were dissolved in 900 µL DMF with 144 µL NMP added and stirred for 2 hours at 70 ℃. 90 mg FAI and 14 mg MACl were dissolved in 1 mL IPA and stirred for 30 minutes at room temperature. Then, 75 µL of the mixed PbI 2 and CsI solution was spin-coated onto prepared SnO 2 at 2000 rpm for 30 s followed by one minute of annealing at 70 ℃. Then, 200 µL of FAI:MACl solution was spin-coated onto the prepared film at 4000 rpm for 30 s, followed by thermal annealing at 150°C for 15 min under N 2 atmosphere. Spiro-OMeTAD solution was spin-coated on the top of the perovskite layer at 4000 rpm for 45 s, which contains 72.5 mg Spiro-OMeTAD, 17.5 µL Li-TFSI solution (520 mg/mL in acetonitrile), 28.8 µL FK209 solution (300 mg/mL in acetonitrile) and 28.8 µL tBP in 1 mL CB. Finally, a 100-nm-thick layer of Ag was thermally evaporated on the top of spiro-OMeTAD as electrodes. To evaluate PSC performance under sunlight, it was illuminated by a normalized AM1.5G solar simulator (100 mW·cm⁻²). PSC photoelectric properties were tested using photovoltaic cell testing equipment. Fabrication of smart flexible insole system . The electrode and circuit of the insole were printed on the flexible polyimide substrate. A 16-channel capacitive pressure sensor array was assembled and strategically placed across the heel, midfoot, and forefoot areas of the insole. The pressure sensors were packaged and connected to a flexible printed circuit board (FPCB) using flat cables. The microcontroller unit (MCU) of the FPCB, which was based on the STM32 main controller chip, was capable of collecting capacitive signals from the 16-channel pressure sensors. The pressure data was then wirelessly transmitted to a mobile application (APP) terminal through the Bluetooth chip embedded in the FPCB. All human subject experiments were conducted with the voluntary and informed consent of the participants, with written consent obtained prior to any demonstrations involving human skin. Static foot arch abnormality detection and dynamic gait pattern recognition. This study was conducted under from the Medical Ethics Committee of the Northwest University (250120003). It focuses on recognizing static foot arch types by constructing a classification model based on plantar capacitance data collected through smart insoles. During the experiment, participants wore smart insoles and maintained a static standing posture for a period of time. In the data collection process, a total of 500 valid samples were collected, consisting of 35% high arches, 35% flat feet, and 30% normal arches. The dataset was randomly split into 90% training set and 10% test set. Based on the train set, the Random Forest algorithm was used to train the model, and after parameter tuning, the optimal classifier was obtained. Finally, the model's performance was validated and evaluated using the test set, with key metrics such as accuracy and confusion matrix calculated to assess its classification performance and generalization ability. This study also addresses the challenge of dynamic gait identification. Participants wore smart insoles and walked for a period of time with a specific gait, generating a dataset consisting of 2,110 samples, each containing 16×200 capacitance signals. The dataset was split into 80% training set and 20% test set, randomly. A 1D-CNN framework was used to train the model for 100 epochs using the training set, and the model's performance was evaluated using the test set after per epoch. Key metrics, including loss, accuracy, and the confusion matrix, were used to assess the model's performance and generalization ability. Additionally, analyzing the loss rate and accuracy curves for training set and test set can identify the most effective training epoch or epoch range, reducing overfitting, indicating that the model can generalize well to the unseen dataset. The analysis aimed to evaluate the effectiveness of the model in identifying different dynamic gaits and its potential for analysis in medical diagnosis. Declarations Acknowledgements This work was financially supported by the National Natural Science Foundation of China (22308268, 22478318), National Key Research and Development Program of China (2021YFA0715600, 2021YFA0717700), Young Talent Fund of Association for Science and Technology in Shaanxi, China (20240625). References Lee, J.-H., Cho, K. & Kim, J. K. Age of flexible electronics: emerging trends in soft multifunctional sensors. Adv. Mater . 36 , 2310505 (2024). Tian, G. et al. Hierarchical piezoelectric composites for noninvasive continuous cardiovascular monitoring. Adv. Mater . 36 , 2313612 (2024). Liu, D. et al. Active-matrix sensing array assisted with machine-learning approach for lumbar degenerative disease diagnosis and postoperative assessment. Adv. Funct. Mater . 32 , 2113008 (2022). Wang, X. et al. Biocompatible and breathable healthcare electronics with sensing performances and photothermal antibacterial effect for motion-detecting. Npj. Flex. Electron. 6 , 95 (2022). Lee, J. H., Cho, K. & Kim, J. K. Age of flexible electronics: emerging trends in soft multifunctional sensors. Adv. Mater . 36 , 2310505 (2024). Shin, Y. et al. Damage-free dry transfer method using stress engineering for high-performance flexible two-and three-dimensional electronics. Nat. Mater. 23 , 1411-1420 (2024). Yuan, H. et al. Progress and challenges in flexible capacitive pressure sensors: Microstructure designs and applications. Chem. Eng. J. 485 , 149926 (2024). Zhang, Y. et al. Highly stable flexible pressure sensors with a quasi-homogeneous composition and interlinked interfaces. Nat. Commun. 13 , 1317 (2022). He, R. et al. Flexible miniaturized sensor technologies for long-term physiological monitoring. Npj. Flex. Electron. 6 , 20 (2022). Zhang, C. et al. Wireless, smart hemostasis device with all-soft sensing system for quantitative and real-time pressure evaluation. Adv. Sci . 10 , 2303418 (2023). Wu, G. et al. Innovations in Tactile Sensing: Microstructural Designs for Superior Flexible Sensor Performance. Adv. Funct. Mater . 34 , 2405722 (2024). Niu, S. et al. A wireless body area sensor network based on stretchable passive tags. Nat. Electron. 2 , 361-368 (2019). Mahato, K. et al. Hybrid multimodal wearable sensors for comprehensive health monitoring. Nat. Electron. 7 , 735-750 (2024). Zhang, Q. et al. A zinc-ion battery-type self-powered pressure sensor with long service life. Adv. Mater. 34 , 2205369 (2022). Kong, L. et al. Wireless technologies in flexible and wearable sensing: from materials design, system integration to applications. Adv. Mater. 36 , 2400333 (2024). Liu, L. et al. Strong tough thermogalvanic hydrogel thermocell with extraordinarily high thermoelectric performance. Adv. Mater. 35 , 2300696 (2023). Yu, Y. et al. Biofuel-powered soft electronic skin with multiplexed and wireless sensing for human-machine interfaces. Sci. Robot. 5 , eaaz7946 (2020). Song, Y. et al. Wireless battery-free wearable sweat sensor powered by human motion. Sci. Adv. 6 , eaay9842 (2020). Lu, P. et al. Gel-based triboelectric nanogenerators for flexible sensing: principles, properties, and applications. Nano-Micro. Lett. 16 , 206 (2024). Han, M. et al. Unveiling the potential of two-terminal perovskite/organic tandem solar cells: mechanisms, status, and challenges. Adv. Mater. 36 , 2402143 (2024). Yang, W., Jo, S. H. & Lee, T. W. Perovskite colloidal nanocrystal solar cells: current advances, challenges, and future perspectives. Adv. Mater. 36 , 2401788 (2024). Pandiyan, A. et al. A comprehensive review on perovskite and its functional composites in smart textiles: progress, challenges, opportunities, and future directions. Prog. Mater. Sci. 140 , 101206 (2023). Kim, J.-K. et al. Self-charging integrated energy modules: a record photoelectric storage efficiency of 14.6%. J. Energy Storage 102 , 114149 (2024). Salomon, A. et al. A machine learning contest enhances automated freezing of gait detection and reveals time-of-day effects. Nat. Commun. 15 , 4853 (2024). Ates, H. C. et al. End-to-end design of wearable sensors. Nat. Rev. Mater. 7 , 887-907 (2022). Song, L. et al. Phytic acid‐induced gradient hydrogels for highly sensitive and broad range pressure sensing. Adv. Mater. 37 , 2417978 (2025). Zhang, Y. et al. Highly stable flexible pressure sensors with a quasi-homogeneous composition and interlinked interfaces. Nat. Commun. 13 , 1317 (2022). Boutry, C. M. et al. A stretchable and biodegradable strain and pressure sensor for orthopaedic application. Nat. Electron. 1 , 314-321 (2018). Ji, Y. et al. Splashing‐assisted femtosecond laser‐activated metal deposition for mold‐and mask‐free fabrication of robust microstructured electrodes for flexible pressure sensors. Small 19 , 2207362 (2023). Huang, Z. et al. Highly stretchable ionotronic pressure sensors with broad response range enabled by microstructured ionogel electrodes. J. Mater. Chem. A 11 , 7201-7212 (2023). Keum, K. et al. Dual-stream deep learning integrated multimodal sensors for complex stimulus detection in intelligent sensory systems. Nano Energy 122 , 109342 (2024). Zhang, C. et al. Wireless, smart hemostasis device with all-soft sensing system for quantitative and real-time pressure evaluation. Adv. Sci . 10 , 2303418 (2023). Qiu, J. et al. An efficiently doped pedot: pss ink formulation via metastable liquid− liquid contact for capillary flow‐driven, hierarchically and highly conductive films. Small 19 , 2205324 (2023). Basarir, F. et al. Edible and biodegradable wearable capacitive pressure sensors: a paradigm shift toward sustainable electronics with bio‐based materials. Adv. Funct. Mater . 34 , 2403268 (2024). Wu, L. et al. Beetle‐inspired gradient slant structures for capacitive pressure sensor with a broad linear response range. Adv. Funct. Mater . 34 , 2312370 (2024). Huang, J. et al. Multi‐hierarchical microstructures boosted linearity of flexible capacitive pressure sensor. Adv. Eng. Mater. 24 , 2101767 (2022). Boutry, C. M. et al. A hierarchically patterned, bioinspired e-skin able to detect the direction of applied pressure for robotics. Sci. Robot. 3 , eaau6914 (2018). Mohammedi, K. et al. Arm and ankle blood pressure indices, and peripheral artery disease, and mortality: a cohort study. Eur. Heart J. 45 , 1738-1749 (2024). Yang, R. et al. Iontronic pressure sensor with high sensitivity over ultra-broad linear range enabled by laser-induced gradient micro-pyramids. Nat. Commun. 14 , 2907 (2023). Choi, Y. S. et al. A transient, closed-loop network of wireless, body-integrated devices for autonomous electrotherapy. Science 376 , 1006-1012 (2022). Park, S. et al. Self-powered ultra-flexible electronics via nano-grating-patterned organic photovoltaics. Nature 561 , 516-521 (2018). Shao, B. et al. Large-area, untethered, metamorphic, and omnidirectionally stretchable multiplexing self-powered triboelectric skins. Nat. Commun. 15 , 1238 (2024). Yin, L. et al. A passive perspiration biofuel cell: high energy return on investment. Joule 5 , 1888-1904 (2021). Yin, L. et al. A self-sustainable wearable multi-modular e-textile bioenergy microgrid system. Nat. Commun. 12 , 1542 (2021). Han, Z. et al. Engineering d‐p orbital hybridization in single‐atom metal‐embedded three‐dimensional electrodes for Li–S batteries. Adv. Mater. 33 , 2105947 (2021). Song, F. et al. Mechanical durability and flexibility in perovskite photovoltaics: advancements and applications. Adv. Mater. 36 , 2312041 (2024). Xu, Y. et al. Perovskite films regulation via hydrogen‐bonded polymer network for efficient and stable perovskite solar cells. Angew. Chem. Int. Edit. 62 , e202306229 (2023). Kim, K. K. et al. A substrate-less nanomesh receptor with meta-learning for rapid hand task recognition. Nat. Electron. 6 , 64–75 (2023). Li, Z. et al. Crossmodal sensory neurons based on high-performance flexible memristors for human-machine in-sensor computing system. Nat. Commun. 15 , 7275 (2024). Luo, S. et al. Experiment-free exoskeleton assistance via learning in simulation. Nature. 630 , 353-359 (2024). Ates, H. C. et al. End-to-end design of wearable sensors. Nat. Rev. Mater. 7 , 887-907 (2022). Xia, H. et al. Shaping high-performance wearable robots for human motor and sensory reconstruction and enhancement. Nat. Commun. 15 , 1760 (2024). Choi, Y. S. et al. A transient, closed-loop network of wireless, body-integrated devices for autonomous electrotherapy. Science 376 , 1006-1012 (2022). Additional Declarations There is NO Competing Interest. Supplementary Files SupportingInformation.docx Supplementary Note 1- Note 6; Supplementary Table 1-Table 2 VideoS1.mp4 FEA simulation of the sensor with double-sided pyramid microstructure VideoS2.mp4 The signal stability of the pressure sensor affixed to the tire tread of a running car VideoS3.mp4 LEDs powered by PSC series, forming various lighting patterns VideoS4.mp4 Smart insole powered by both Li-S batteries and PSCs VideoS5.mp4 Real-time gait monitoring during walking up the stairs VideoS6.mp4 Real-time gait monitoring during marching in place Cite Share Download PDF Status: Published Journal Publication published 07 Jan, 2026 Read the published version in Research → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {\"props\":{\"pageProps\":{\"initialData\":{\"identity\":\"rs-6314292\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":true,\"archivedVersions\":[],\"articleType\":\"Article\",\"associatedPublications\":[],\"authors\":[{\"id\":435279015,\"identity\":\"29d2d50a-5c83-487e-b8bf-ff50abf58e51\",\"order_by\":0,\"name\":\"Feng Xu\",\"email\":\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/ElEQVRIiWNgGAWjYBAC9gYwZcPAwMwMZPKAeQZ4tfAcAFNpQC2MIC0GRGs5DMQgLQzEaGHvPfyat+18NH87YwMzj8yfxAb25m0SDDV3cGvhOZdmObPtdu6Mw4wNjDN4DBIbeI6VSTAce4ZTi71EjpnBR6CWBqAWhg8gLUARCcaGw7htkX9jZpDYdi53PkhLAkgLUAS/Fgke4wcf2w7kbkDYwkNAC0+OGeOMc8m5G4FaDs7gMTZu40krtkg4hkcL+xnjzzxldrnzzh8++Ji3R062n/3wxhsfanBrAQI2CRjrAGMPkAtiJeDTAIz2Dwj2D/xKR8EoGAWjYGQCAEbAUmnqN4Z9AAAAAElFTkSuQmCC\",\"orcid\":\"\",\"institution\":\"Xi’an Jiaotong University\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"Feng\",\"middleName\":\"\",\"lastName\":\"Xu\",\"suffix\":\"\"},{\"id\":435279016,\"identity\":\"f9774f09-bce6-4d4a-8d7d-e2a0591a4809\",\"order_by\":1,\"name\":\"Yingchun Li\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yingchun\",\"middleName\":\"\",\"lastName\":\"Li\",\"suffix\":\"\"},{\"id\":435279017,\"identity\":\"c21a9e2f-f8f3-40fe-a9fb-4655671123ad\",\"order_by\":2,\"name\":\"Yarong Ding\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yarong\",\"middleName\":\"\",\"lastName\":\"Ding\",\"suffix\":\"\"},{\"id\":435279018,\"identity\":\"a619810d-ccba-4dcd-a776-433235fc6500\",\"order_by\":3,\"name\":\"Yuze Zhang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yuze\",\"middleName\":\"\",\"lastName\":\"Zhang\",\"suffix\":\"\"},{\"id\":435279019,\"identity\":\"baae6bd7-cbfa-4224-8db2-527aee554d23\",\"order_by\":4,\"name\":\"Xing Guo\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Xing\",\"middleName\":\"\",\"lastName\":\"Guo\",\"suffix\":\"\"},{\"id\":435279020,\"identity\":\"488fc4a1-35fd-4def-81d0-6160bfe8eaae\",\"order_by\":5,\"name\":\"Kaixin Lei\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Kaixin\",\"middleName\":\"\",\"lastName\":\"Lei\",\"suffix\":\"\"},{\"id\":435279021,\"identity\":\"ca6c4090-e7f2-42ee-8402-2ea37946018e\",\"order_by\":6,\"name\":\"Jiachun Sun\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Jiachun\",\"middleName\":\"\",\"lastName\":\"Sun\",\"suffix\":\"\"},{\"id\":435279022,\"identity\":\"19bd1571-0343-4e98-bd7a-f9f464a2b674\",\"order_by\":7,\"name\":\"Xing Hu\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Xing\",\"middleName\":\"\",\"lastName\":\"Hu\",\"suffix\":\"\"},{\"id\":435279023,\"identity\":\"736e0a7b-fef9-4fc0-bcf2-408de68abbb6\",\"order_by\":8,\"name\":\"Xinyue Li\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Xinyue\",\"middleName\":\"\",\"lastName\":\"Li\",\"suffix\":\"\"},{\"id\":435279024,\"identity\":\"3aaead72-c8fb-46cf-8053-f2d687ecdffa\",\"order_by\":9,\"name\":\"Wenguang Yang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Wenguang\",\"middleName\":\"\",\"lastName\":\"Yang\",\"suffix\":\"\"},{\"id\":435279025,\"identity\":\"66f59789-77c6-46d3-95ad-e175e7da57eb\",\"order_by\":10,\"name\":\"Rui Liu\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Tangdu Hospital\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Rui\",\"middleName\":\"\",\"lastName\":\"Liu\",\"suffix\":\"\"},{\"id\":435279026,\"identity\":\"e227b90d-b00d-43a5-8ab0-64cb97aa3d06\",\"order_by\":11,\"name\":\"Zhenhua Lin\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Zhenhua\",\"middleName\":\"\",\"lastName\":\"Lin\",\"suffix\":\"\"},{\"id\":435279027,\"identity\":\"c7953133-780e-4465-9302-b7e16fe6c227\",\"order_by\":12,\"name\":\"Wendong Zhang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Tangdu Hospital\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Wendong\",\"middleName\":\"\",\"lastName\":\"Zhang\",\"suffix\":\"\"},{\"id\":435279028,\"identity\":\"af6df903-5b69-446e-b225-fdb3fc1aecaa\",\"order_by\":13,\"name\":\"Shaozhe Tan\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Shaozhe\",\"middleName\":\"\",\"lastName\":\"Tan\",\"suffix\":\"\"},{\"id\":435279029,\"identity\":\"b7dea42e-1eb2-47d0-b74c-63dedaddb4bb\",\"order_by\":14,\"name\":\"Xu Yang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Northwest University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Xu\",\"middleName\":\"\",\"lastName\":\"Yang\",\"suffix\":\"\"},{\"id\":435279030,\"identity\":\"67715aef-b529-4adc-b254-33a4457cfeae\",\"order_by\":15,\"name\":\"Yumeng Xu\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yumeng\",\"middleName\":\"\",\"lastName\":\"Xu\",\"suffix\":\"\"},{\"id\":435279031,\"identity\":\"d8598b8d-3bc0-49b6-8c9e-151858a845e8\",\"order_by\":16,\"name\":\"Jin Tian\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xi’an Jiaotong University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Jin\",\"middleName\":\"\",\"lastName\":\"Tian\",\"suffix\":\"\"},{\"id\":435279032,\"identity\":\"61017867-4400-4296-baf3-a98b5fb19ab1\",\"order_by\":17,\"name\":\"Yue Hao\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yue\",\"middleName\":\"\",\"lastName\":\"Hao\",\"suffix\":\"\"},{\"id\":435279033,\"identity\":\"a6c41d23-e427-46d6-a91b-c9a212cec142\",\"order_by\":18,\"name\":\"Yannan Liu\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Northwest University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yannan\",\"middleName\":\"\",\"lastName\":\"Liu\",\"suffix\":\"\"},{\"id\":435279034,\"identity\":\"fb6c5994-16a6-49ff-af28-d4c616d60acc\",\"order_by\":19,\"name\":\"Xiangning Li\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Xiangning\",\"middleName\":\"\",\"lastName\":\"Li\",\"suffix\":\"\"},{\"id\":435279035,\"identity\":\"dde251b1-7e99-48bc-a3e2-553191e3d1ef\",\"order_by\":20,\"name\":\"Jingjing Chang\",\"email\":\"\",\"orcid\":\"https://orcid.org/0000-0003-3773-182X\",\"institution\":\"Xidian University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Jingjing\",\"middleName\":\"\",\"lastName\":\"Chang\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2025-03-26 16:47:14\",\"currentVersionCode\":1,\"declarations\":\"\",\"doi\":\"10.21203/rs.3.rs-6314292/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-6314292/v1\",\"draftVersion\":[],\"editorialEvents\":[{\"content\":\"https://doi.org/10.34133/research.1063\",\"type\":\"published\",\"date\":\"2026-01-08T00:00:00+00:00\"}],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":79658283,\"identity\":\"2150f8f7-8d80-456b-8844-0fae1d1de45e\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:12:25\",\"extension\":\"png\",\"order_by\":1,\"title\":\"Figure 1\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1117681,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eSchematic illustration of biomimetic self-powered insole with AI-enhanced mechanodiagnosis for continuous gait monitoring. \\u003c/strong\\u003e(a) Sensing inspiration from mantis legs. (b) Schematic illustration of the integrated smart insole of sensing - power supply - AI diagnosis for gait monitoring. (c) Sensing mechanism of the as-designed capacitive pressure sensor for gradient pressure sensing. (d) The core functions and excellent features of the integrated smart insole.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage1.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/3043ccf2e1e4d5d9379dd162.png\"},{\"id\":79658736,\"identity\":\"b7526629-1450-4ee5-8c2b-6d6f04153b34\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:20:26\",\"extension\":\"png\",\"order_by\":2,\"title\":\"Figure 2\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1628604,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eCharacterization and performance of the pressure sensor\\u003c/strong\\u003e. (a) SEM images of double-sided micropyramid PEDOT/TPU foam. (b) Finite element simulation of pressing process. (c) Sensitivity of sensors without microstructures, with single-sided pyramids, and double-sided pyramids. (d) Response and recovery times at three pressure levels. (e) LOD detection. (f) Stability under 200 kPa for 12,000 cycles. (g) Comparison of sensitivity, LOD, and work range among different designs of capacitive pressure sensors.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/674abfd52b5683c1be521bde.png\"},{\"id\":79658285,\"identity\":\"ba90731e-10b6-47d4-baa0-a7a33c4e0b7c\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:12:26\",\"extension\":\"png\",\"order_by\":3,\"title\":\"Figure 3\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1138100,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eReal-time monitoring application demo of capacitive pressure sensor.\\u003c/strong\\u003e (a) Schematic illustration of the testing setup and the signal stability of the pressure sensor affixed to the tire tread during the car running on the road. (b) Photograph of the pressure testing system installed on the automobile tire, as well as the SEM images of the sectional view of the pressure sensor and enlarged view of the micropyramid array after testing. (c) Recognition of acoustic wave, airflow, carotid pulse, ankle pulse and finger bend using the pressure sensor. (d-e) Recognition of pressure distribution on 4×4 sensors array for (d) origami cranes and (e) leaf.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/2fb6abb2549827eb22c41bd8.png\"},{\"id\":79658287,\"identity\":\"e088a77a-161c-4c84-a874-40feb5efbf4e\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:12:26\",\"extension\":\"png\",\"order_by\":4,\"title\":\"Figure 4\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":429337,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eCharacterizations of photorechargeable power system. \\u003c/strong\\u003e(a) The working principle of closed-loop power supply using PSCs for generation and Li-S batteries for storage. (b) Schematic diagram and SEM image of PSC device and perovskite crystal structure. (c) \\u003cem\\u003eJ-V\\u003c/em\\u003e curve of a PSC. (d) Li-S battery voltage profiles during charging (PSC or DC) and discharging. (e) Li-S battery discharge capacity with PSC and DC charging. (f) \\u003cem\\u003eJ-V\\u003c/em\\u003e curves of 4 series-connected PSCs over 5 cycles. (g) Energy storage and overall efficiency of the PSC-charged Li-S battery. (h) LEDs powered by PSC series.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/58a3b126710e2dc5f38d9a3e.png\"},{\"id\":79658737,\"identity\":\"5e6b0863-cbd5-4503-bd69-726925c7850a\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:20:26\",\"extension\":\"png\",\"order_by\":5,\"title\":\"Figure 5\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":2862205,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eIntegration of smart insole control system with risk warning.\\u003c/strong\\u003e (a) Schematic of integrated system of the integrated smart insole system, including closed-loop power supply, sensing, and control system. (b-d) Photos of (b) printed insole, (c) bending FPCB, and (d) APP interface. (e) Control circuit powered by PSCs and displayed on cellphone. (f) FPCB powered by PSC series. (g) Smart insole for yoga gait monitoring with poses schematic. (I-II) Plantar recognition using commercial pad and smart insole. (h) Falling warnings using commercial pad and smart insole.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage5.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/e20e7a01cdc3bf788f231ed3.png\"},{\"id\":79658290,\"identity\":\"ea9da3ef-4ce0-48ba-a28b-062b1d1ebf35\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:12:26\",\"extension\":\"png\",\"order_by\":6,\"title\":\"Figure 6\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":2617527,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eAI-assisted identification of arch diseases and gaits. \\u003c/strong\\u003e(a) Bar graph of plantar pressure distribution and random forest algorithm for classifying normal, high-arched, and flat feet. (b) Confusion matrix result for the recognition of arch diseases. (c) Training and validation accuracy of random forest. (d) t-SNE visualization of the clustered data of the arch diseases. (e) Maps of the 16-channel dynamic data of 12 gait types and 1D-CNN framework. (f) Confusion matrix result for the recognition of 12 gaits (The full version is shown in \\u003cstrong\\u003eFig. S14a\\u003c/strong\\u003e). (g) Training and validation accuracy of 1D-CNN framework. (h) t-SNE visualization of the clustered data of gaits.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage6.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/a591503f970bf5471039c1db.png\"},{\"id\":108007266,\"identity\":\"227a3a21-80f3-484f-9d0f-41505d5fde33\",\"added_by\":\"auto\",\"created_at\":\"2026-04-28 12:59:13\",\"extension\":\"pdf\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":10169205,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"manuscript.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/dfcb23e6-cc5d-42e1-94dd-9f80646910c9.pdf\"},{\"id\":79658316,\"identity\":\"aa307e70-9b37-4d13-b997-7254dbe67371\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:12:27\",\"extension\":\"docx\",\"order_by\":1,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":20813119,\"visible\":true,\"origin\":\"\",\"legend\":\"Supplementary Note 1- Note 6; Supplementary Table 1-Table 2\",\"description\":\"\",\"filename\":\"SupportingInformation.docx\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/f06e581193c3ac4ded1e242a.docx\"},{\"id\":79658740,\"identity\":\"de68f814-ede0-4804-8b4b-2df2a59c7215\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:20:26\",\"extension\":\"mp4\",\"order_by\":2,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":8741531,\"visible\":true,\"origin\":\"\",\"legend\":\"FEA simulation of the sensor with double-sided pyramid microstructure\",\"description\":\"\",\"filename\":\"VideoS1.mp4\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/6acffe8bce8f3e3362bda8ec.mp4\"},{\"id\":79658294,\"identity\":\"286d83fd-1aa0-4303-b346-79c69868c32e\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:12:26\",\"extension\":\"mp4\",\"order_by\":3,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":19652824,\"visible\":true,\"origin\":\"\",\"legend\":\"The signal stability of the pressure sensor affixed to the tire tread of a running car\",\"description\":\"\",\"filename\":\"VideoS2.mp4\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/daf6ee9d9ff82f20340f759c.mp4\"},{\"id\":79658295,\"identity\":\"35c0f5f9-045c-4915-b347-4b92aa24411a\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:12:26\",\"extension\":\"mp4\",\"order_by\":4,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":6887605,\"visible\":true,\"origin\":\"\",\"legend\":\"LEDs powered by PSC series, forming various lighting patterns\",\"description\":\"\",\"filename\":\"VideoS3.mp4\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/e6847639ff71ff609bda7b1e.mp4\"},{\"id\":79658296,\"identity\":\"a1f57c16-6a59-4ac1-beb9-393bc5297ef7\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:12:26\",\"extension\":\"mp4\",\"order_by\":5,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":17335900,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eSmart insole powered by both Li-S batteries and PSCs\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"VideoS4.mp4\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/22b0a066e6bc46a4bbbf2cef.mp4\"},{\"id\":79658310,\"identity\":\"e8ebdb20-14a1-473e-8bc7-a3f3fbda22cb\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:12:26\",\"extension\":\"mp4\",\"order_by\":6,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":11204983,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eReal-time gait monitoring during walking up the stairs\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"VideoS5.mp4\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/f352d710a6c11d8db1265bee.mp4\"},{\"id\":79658301,\"identity\":\"0fa45d96-be42-4706-bb60-ca9e9e91f7c8\",\"added_by\":\"auto\",\"created_at\":\"2025-04-01 09:12:26\",\"extension\":\"mp4\",\"order_by\":7,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":10625644,\"visible\":true,\"origin\":\"\",\"legend\":\"Real-time gait monitoring during marching in place\",\"description\":\"\",\"filename\":\"VideoS6.mp4\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6314292/v1/5654d3c2fdf8762ada9c8ab0.mp4\"}],\"financialInterests\":\"There is \\u003cb\\u003eNO\\u003c/b\\u003e Competing Interest.\",\"formattedTitle\":\"Biomimetic Self-Powered Smart Insole with AI-Enhanced Mechanodiagnosis for Continuous Gait Monitoring\",\"fulltext\":[{\"header\":\"Introduction\",\"content\":\"\\u003cp\\u003eThe growing prevalence of age-related mobility decline, chronic musculoskeletal disorders (e.g., Parkinsonian gait, diabetic neuropathy, osteoarthritis), and structural foot abnormalities (e.g., flatfoot, high-arched foot) underscores the urgent need for wearable systems capable of continuous, biomechanical monitoring of pathological gait patterns (staggering gait, asymmetric strides) \\u003csup\\u003e1,2\\u003c/sup\\u003e. Gait analysis serves as a vital biomarker for neuromuscular health, offering insights into disease progression and rehabilitation efficacy \\u003csup\\u003e3,4\\u003c/sup\\u003e. However, clinical gait assessment remains tethered to laboratory environments due to the reliance on optical motion-capture systems and force plates, which are cost-prohibitive, spatially constrained, and incapable of capturing real-world walking dynamics \\u003csup\\u003e5,6\\u003c/sup\\u003e. Wearable pressure-sensing insoles could decentralize this process, yet existing solutions fail to meet three interconnected requirements of clinical viability, \\u003cem\\u003ei.e.\\u003c/em\\u003e, sensors with durability and resolution to span the full biomechanical force spectrum, energy autonomy for uninterrupted operation, and real-time interpretation of pathological gait signatures \\u003csup\\u003e7-9\\u003c/sup\\u003e.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u0026nbsp;Current flexible pressure sensors face fundamental trade-offs in performance. The plantar interface subjects devices to extreme mechanical demands: transient forces exceeding 1 MPa during athletic activity, subtle shifts (\\u0026lt;1 Pa) during postural adjustments, and cyclic loading over thousands of daily steps, a regime that rapidly degrades conventional sensors with homogeneous or single-scale microstructures \\u003csup\\u003e10,11\\u003c/sup\\u003e. While bioinspired designs, such as those mimicking gecko feet (hierarchical stress distribution) or spider slit sensilla (vibration sensitivity), have improved dynamic responsiveness \\u003csup\\u003e12,13\\u003c/sup\\u003e, these systems often sacrifice the mechanical robustness (\\u0026gt;10,000 cycles) or biocompatibility for prolonged use.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u0026nbsp;Energy autonomy poses a parallel challenge. High-density sensor arrays and wireless data transmission demand significant power, yet bulky batteries or frequent recharging disrupt continuous monitoring, a prerequisite for capturing episodic gait abnormalities \\u003csup\\u003e14,15\\u003c/sup\\u003e. Although emerging solutions like biofuel cells and triboelectric nanogenerators (TENGs) have been explored for battery-free wearables, biofuel cells degrade with sweat exposure, while TENGs require vigorous motion and generate high-voltage, low-current outputs incompatible with modern wearable electronics \\u003csup\\u003e16-19\\u003c/sup\\u003e. Recent advances in perovskite solar cells (PSCs), which achieve high power conversion efficiency under indoor lighting while maintaining flexibility, offer promise for self-sustaining wearables \\u003csup\\u003e20-22\\u003c/sup\\u003e. Pairing PSCs with high-energy-density storage units like lithium-sulfur (Li-S) batteries could enable closed-loop power systems resilient to environmental fluctuations \\u003csup\\u003e23\\u003c/sup\\u003e. However, no prior work has successfully integrated these technologies to deliver seamless operation across real-world environments, from low-light indoor settings to variable outdoor conditions.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u0026nbsp;Beyond hardware limitations, translating raw sensor data into actionable clinical insights remains a major challenge. Although machine learning algorithms achieve high accuracy in classifying gait abnormalities under controlled conditions \\u003csup\\u003e24\\u003c/sup\\u003e, most wearable systems lack real-time processing, intuitive visualization, and adaptive feedback, which are essential for patient adherence and timely clinical decision-making \\u003csup\\u003e25\\u003c/sup\\u003e. A transformative platform must not only capture dynamic plantar pressure maps but also interpret them through AI models trained on heterogeneous pathological gaits, all of which should operate autonomously at the point of care.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u0026nbsp;Here, we introduce a biomimetic smart insole that addresses these challenges through three synergistic strategies (\\u003cstrong\\u003eFig. 1\\u003c/strong\\u003e). First, inspired by the hierarchical mechanosensory architecture of mantis legs, we developed a dual-microstructure capacitive pressure sensor combining polydimethylsiloxane (PDMS) microstructures with compressible foam. This design achieves an ultralow detection limit (0.10 Pa), high sensitivity (0.602 kPa⁻\\u0026sup1;), and a broad dynamic range (0.10 Pa\\u0026ndash;1.40 MPa), with \\u0026gt;12,000-cycle durability surpassing existing wearable sensors. Second, a hybrid perovskite solar cell/lithium-sulfur (PSC/Li-S) energy system ensures uninterrupted operation across diverse environments, eliminating dependency on external power. Third, an embedded AI framework synergizes a random forest classifier for foot arch abnormality detection (96% accuracy) and a one-dimensional convolutional neural network (1D-CNN) for classifying 12 pathological gait patterns (97.6% accuracy). A companion mobile APP visualizes dynamic force fields through chromatic mapping, empowering clinicians with real-time, interpretable diagnostics. By integrating high-fidelity biomechanical sensing, self-sufficient energy harvesting, and edge-computing intelligence, this platform bridges a critical gap between wearable technology and clinical practice. Its ability to provide continuous, AI-augmented gait analysis outside laboratory settings advances preventive care paradigms, from reducing fall risks in elderly populations to enabling personalized rehabilitation in telemedicine frameworks.\\u003c/p\\u003e\\u003cp\\u003e\\u003cstrong\\u003e\\u003cem\\u003eDual microstructure design achieves clinically relevant sensitivity across biomechanical pressure ranges.\\u003c/em\\u003e\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eTo enable continuous detection of gait-related pressures from subtle postural shifts to high-impact forces, we engineered a capacitive pressure sensor as inspired by the hierarchical elasticity of mantis legs (\\u003cstrong\\u003eFig. 1\\u003c/strong\\u003e). The sensor integrates thermoplastic polyurethane (TPU) foam doped with poly(3,4-ethylenedioxythiophene) (PEDOT) via vapor-phase polymerization (VPP) and PDMS pyramid microstructures on both surfaces (\\u003cstrong\\u003eFig. S1\\u003c/strong\\u003e). Energy-dispersive spectroscopy (EDS), Fourier-transform infrared (FTIR) spectroscopy, and Raman spectroscopy confirm homogeneous PEDOT distribution within the TPU matrix (\\u003cstrong\\u003eFig. S2, Fig. S3\\u003c/strong\\u003e). This doping strategy increases the foam’s Young’s modulus by 38% and dielectric constant by 4.2× compared to undoped TPU (\\u003cstrong\\u003eSupplementary Note 1\\u003c/strong\\u003e), establishing a mechanically robust foundation for pressure sensing.\\u003c/p\\u003e\\n\\u003cp\\u003eTo optimize stress distribution and dielectric modulation, we patterned PDMS pyramids (14 µm width, 9 µm height, 7 µm spacing) onto the PEDOT-doped foam (\\u003cstrong\\u003eFig. 2a\\u003c/strong\\u003e). Finite element simulations reveal that dual microstructures amplify localized deformation by 72% under load compared to flat designs, enhancing the air gap effect and dielectric polarization (\\u003cstrong\\u003eFig. 2b\\u003c/strong\\u003e, \\u003cstrong\\u003eSupplementary Note 2, Video S1\\u003c/strong\\u003e). Experimentally, sensors with double-sided pyramids achieve a sensitivity of 0.602 kPa⁻¹, which is 2.8× higher than non-patterned sensors (0.213 kPa⁻¹) and 1.9× higher than single-sided designs (0.318 kPa⁻¹) (\\u003cstrong\\u003eFig. 2c, Fig. S7\\u003c/strong\\u003e). The working range extends to 1.4 MPa, exceeding the peak plantar pressures observed during walking (1.2 MPa) \\u003csup\\u003e26\\u003c/sup\\u003e. These results demonstrate that our biomimetic microstructures synergistically enhance sensitivity and biomechanical compatibility.\\u003c/p\\u003e\\n\\u003cp\\u003eTo address the clinical need for detecting subtle pressure variations and transient forces, we quantified sensor response times under physiologically relevant loads. The device resolves pressure changes as low as 0.10 Pa (\\u003cstrong\\u003eFig. 2d\\u003c/strong\\u003e) and exhibits rapid response/recovery times of 135 ms (0.5 kPa), 167 ms (1 kPa), and 192 ms (5 kPa) (\\u003cstrong\\u003eFig. 2e\\u003c/strong\\u003e, \\u003cstrong\\u003eFig. S9\\u003c/strong\\u003e), outperforming existing wearable sensors \\u003csup\\u003e27-37\\u003c/sup\\u003e. Cyclic loading tests (100 kPa at 2 Hz) confirm stable performance over 12,000 cycles with \\u0026lt;3% signal drift (\\u003cstrong\\u003eFig. 2f\\u003c/strong\\u003e), meeting the durability requirements for multi-day gait monitoring.\\u003c/p\\u003e\\n\\u003cp\\u003eWe benchmark our sensor against 10 state-of-the-art flexible pressure sensors \\u003csup\\u003e27-37\\u003c/sup\\u003e across four metrics critical for clinical use, including sensitivity (0.602 kPa⁻¹), detection limit (0.10 Pa), response time (135 ms), and working range (0.10 Pa–1.40 MPa) (\\u003cstrong\\u003eFig. 2g\\u003c/strong\\u003e, \\u003cstrong\\u003eTable S1\\u003c/strong\\u003e). The device achieves top-tier performance in combined sensitivity and dynamic range, while ensuring biocompatibility (ISO 10993-5 certified) and maintaining mechanical flexibility. This performance profile enables continuous monitoring of pathological gait patterns, such as diabetic foot ulcer precursors and osteoarthritis-related joint overloading.\\u003c/p\\u003e\\n\\u003cp\\u003eBy combining biomimetic microstructure engineering with conductive polymer doping, we developed a capacitive pressure sensor that overcomes key limitations in wearable gait analysis. The dual-microstructure design achieves clinically relevant sensitivity and durability, while the 0.10 Pa detection limit identifies subclinical biomechanical anomalies.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003e\\u003cem\\u003eDual-microstructure sensors withstand extreme mechanical stress and gradient pressure.\\u003c/em\\u003e\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eTo evaluate sensor durability under real-world mechanical stress, we affixed a 1 cm × 1 cm sensor to a car tire tread, subjecting it to 300 kPa compressive and 6 kPa shear forces during continuous driving (\\u003cstrong\\u003eFig. 3a\\u003c/strong\\u003e). We observed stable wireless capacitance signals during prolonged continuous testing, with no signal drift or anomalies (\\u003cstrong\\u003eVideo S2\\u003c/strong\\u003e). Post-test scanning electron microscopy (SEM) reveals intact TPU/PEDOT foam pores and undamaged PDMS pyramids (\\u003cstrong\\u003eFig. 3b\\u003c/strong\\u003e), confirming the dual-microstructure design resists delamination and deformation under extreme loads. These results demonstrate the sensor’s robustness for wearable applications requiring prolonged exposure to harsh biomechanical forces, such as athletic monitoring or industrial exoskeletons.\\u003c/p\\u003e\\n\\u003cp\\u003eTo demonstrate gradient pressure sensing capability and clinical-grade sensitivity, we further tested the sensor’s response to subtle pressures (\\u003cem\\u003ee.g.\\u003c/em\\u003e, acoustic waves, airflow), clinically relevant middle pressures (\\u003cem\\u003ee.g.\\u003c/em\\u003e, pulse), and angular-dependent pressures from finger bending (\\u003cstrong\\u003eFig. 3c,\\u003c/strong\\u003e \\u003cstrong\\u003eSupplementary Note 3\\u003c/strong\\u003e). The sensor demonstrates excellent responsiveness to a 100 dB acoustic wave and airflow at speeds below the human skin’s detection threshold (≈0.50 m s⁻¹). The system also resolves carotid pulse harmonics (0.50–20 Hz) and distinguishes ankle pulse amplitude variations (ΔC: 0.80–1.20 pF), which is linked to peripheral artery disease \\u003csup\\u003e38\\u003c/sup\\u003e. During gesture recognition, capacitance shifts enable precise differentiation of finger flexion angles (30°–120°), with signal output showing a strong linear correlation (R² = 0.98). These results indicate the sensor’s ability to detect subclinical biomechanical anomalies, such as early-stage neuropathy or joint stiffness, with precision rivaling laboratory-grade instruments.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eTo ensure compatibility with long-term wearable use, we performed cytotoxicity assays using L929 fibroblast cells and skin irritation tests per ISO 10993-5 standards (\\u003cstrong\\u003eFig. S10\\u003c/strong\\u003e). We observed no reduction in cell viability (\\u0026gt;95% survival) or erythema in epidermal models after 72-hour exposure. These findings validate the sensor’s safety for uninterrupted skin contact, addressing a critical barrier to clinical adoption of wearable electronics.\\u003c/p\\u003e\\n\\u003cp\\u003eTo validate spatial resolution for medical diagnostics, we fabricated a 4 × 4 array and map pressure distributions from lightweight objects (1–10 Pa). The array accurately identified the spatial pressure profiles of origami cranes and leaf positioned at distinct locations (\\u003cstrong\\u003eFig. 3d,e\\u003c/strong\\u003e) This spatial recognition capability highlights the platform’s potential for adaptive prosthetics, pressure ulcer prevention, and robotic tactile sensing.\\u003c/p\\u003e\\n\\u003cp\\u003eBy unifying mechanical resilience, gradient pressure sensing capability, and biocompatibility, our dual-microstructure sensors overcome key limitations in current wearable biomechanical monitors. The 300 kPa durability ensures reliable operation under strenuous activities, while 0.10 Pa sensitivity enables early detection of neuromuscular degradation, a critical advance for preventive healthcare.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003e\\u003cem\\u003eIntegrated photo-rechargeable system sustains continuous operation of wearable biomedical devices.\\u003c/em\\u003e\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eTo enable autonomous, long-term operation of wearable health monitors, we developed a closed-loop energy system that synergizes PSCs with Li–S batteries (\\u003cstrong\\u003eFig. 4a\\u003c/strong\\u003e).\\u0026nbsp; The power management system dynamically adapts operational modes through an adaptive energy allocation strategy responsive to environmental illumination. Under sufficient outdoor solar irradiance, PSCs simultaneously charge Li-S batteries and power sensor-integrated circuits; when sunlight diminishes below operational threshold, photovoltaic supply prioritizes sensor circuit demands while preserving Li-S energy reserves. Indoor operation transitions to Li-S discharge mode for sustaining power supply of continuous sensing. This tri-mode regulation ensures uninterrupted cross-environment sensing.\\u003c/p\\u003e\\n\\u003cp\\u003eThe PSCs, fabricated on indium tin oxide (ITO)-coated glass or PET substrates with silver electrodes, achieve a power conversion efficiency (PCE) of 21.3% under standard solar illumination (AM1.5G, 100 mW cm⁻²) through optimized charge transport layers of SnO₂ and spiro-OMeTAD (\\u003cstrong\\u003eFig. 4b,c\\u003c/strong\\u003e). This configuration delivers a short-circuit current density (\\u003cem\\u003eJ\\u003c/em\\u003e\\u003csub\\u003eSC\\u003c/sub\\u003e) of 23.10 mA cm⁻² and an open-circuit voltage (\\u003cem\\u003eV\\u003c/em\\u003e\\u003csub\\u003eOC\\u003c/sub\\u003e) of 1.15 V, providing sufficient energy density to power continuous gait monitoring.\\u003c/p\\u003e\\n\\u003cp\\u003eTo assess the energy transfer for wearable applications, we connected four PSCs in series (3.88 V output) to Li–S batteries with sulfur-infused carbon electrodes (\\u003cstrong\\u003eSupplementary Note 4\\u003c/strong\\u003e). Galvanostatic charge–discharge cycles reveal nearly identical voltage profiles between PSC-charged and DC-powered systems (\\u003cstrong\\u003eFig. 4d\\u003c/strong\\u003e), confirming efficient solar energy conversion. The PSC-charged batteries retain 99.7% of their initial discharge capacity (170.1 mA h g⁻¹) after five cycles, matching conventional charging methods (\\u003cstrong\\u003eFig. 4e\\u003c/strong\\u003e). After repeated photocharge–discharge cycles, the Li–S batteries maintain \\u0026gt;95% capacity retention and the PSCs show \\u0026lt;5% PCE degradation (\\u003cstrong\\u003eFig. 4f\\u003c/strong\\u003e, \\u003cstrong\\u003eSupplementary Note 4\\u003c/strong\\u003e). These results confirm the system’s resilience to fluctuating environmental conditions, ensuring uninterrupted power for sensors, wireless modules, and edge-computing units in smart insoles.\\u003c/p\\u003e\\n\\u003cp\\u003eThe self-charging system achieves an average energy storage efficiency of 72.15% and an average overall photocharging conversion efficiency of 11.14% (\\u003cstrong\\u003eFig. 4g\\u003c/strong\\u003e). The stability of PSCs is important for the self-charging system. We exposed the unencapsulated PSC to air with a relative humidity of 40 % at room temperature. With 94% efficiency retention over 1000-hour aging (\\u003cstrong\\u003eSupplementary Note 4\\u003c/strong\\u003e), the PSCs enable reliable energy autonomy for wearable systems, achieving a key milestone toward clinical-grade health monitoring. To demonstrate practical utility, we drived 14 commercial LEDs using eight series-connected PSCs, achieving stable illumination under simulated sunlight (\\u003cstrong\\u003eFig. 4h\\u003c/strong\\u003e, \\u003cstrong\\u003eVideo S3\\u003c/strong\\u003e).\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eBy integrating high-efficiency PSCs with durable Li–S batteries, we overcome the energy autonomy challenge in wearable biomechanical monitoring. The system’s high photoelectric conversion efficiency and long-term runtime enables continuous gait analysis across diverse real-world settings, from low-light clinical environments to outdoor activities. This self-sustaining platform eliminates reliance on external charging, advancing telehealth applications such as remote diabetic foot ulcer prevention and Parkinson’s disease progression tracking.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003e\\u003cem\\u003eIntegrated smart insole system enables real-time gait monitoring for clinical applications.\\u003c/em\\u003e\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eTo address the unmet need for continuous, clinically actionable gait analysis, we integrated 16 high-performance capacitive pressure sensors into a self-powered smart insole (\\u003cstrong\\u003eFig. 5a\\u003c/strong\\u003e). The system combines a sustainable energy module (PSCs/Li–S batteries), wireless data transmission, and a mobile interface for real-time feedback, achieving seamless operation indoor and outdoor, with no interruptions in data acquisition. These results indicate the platform’s readiness for clinical deployment in rehabilitation and remote patient monitoring.\\u003c/p\\u003e\\n\\u003cp\\u003eTo ensure precise detection of gait abnormalities, we positioned sensors at biomechanically critical regions: toes (2 sensors), forefoot (6 sensors), midfoot (4 sensors), and heel (4 sensors) (\\u003cstrong\\u003eFig. 5b\\u003c/strong\\u003e). Each 1 cm² sensor, separated by a TPU/PEDOT/PDMS dielectric layer, resolves pressure gradients as low as 0.10 Pa. During walking trials, the array captures heel-strike forces (800–1200 kPa) and midfoot pressure shifts (\\u0026lt;10 kPa) with high correlation to laboratory force plates. These findings validate the sensor layout’s capacity to diagnose conditions like diabetic foot ulcers and osteoarthritis through localized pressure anomalies.\\u003c/p\\u003e\\n\\u003cp\\u003eTo enable real-time diagnostics, we connected the sensor array to a flexible printed circuit board (FPCB) with Bluetooth 5.0 (\\u003cstrong\\u003eFig. 5c\\u003c/strong\\u003e). The system wirelessly streams data to a mobile app, which generates dynamic heatmaps that visually encode pressure variations through color gradients (\\u003cstrong\\u003eFig. 5d\\u003c/strong\\u003e, \\u003cstrong\\u003eVideo S4\\u003c/strong\\u003e). The platform has potential to replace intermittent clinic visits with continuous, telemedicine-enabled care.\\u003c/p\\u003e\\n\\u003cp\\u003eTo eliminate reliance on external power, we affixed PSCs and Li–S batteries to the shoe exterior, with the control circuit and APP additionally powered by solar energy for system testing purposes (\\u003cstrong\\u003eFig. 5e,f\\u003c/strong\\u003e). The PSCs maintain 21.3% efficiency under standard solar illumination, charging the batteries to 170 mA h g⁻¹ (\\u003cstrong\\u003eFig. 4e\\u003c/strong\\u003e). These metrics confirm the energy module’s suitability for real-world use, including rural or low-resource settings where charging infrastructure is limited.\\u003c/p\\u003e\\n\\u003cp\\u003eTo validate clinical utility, we tested the insole during yoga and simulated falls. During yoga, it distinguishes balance shifts between poses (\\u003cem\\u003ee.g.\\u003c/em\\u003e, warrior II and tree) with high accuracy, matching commercial pressure pads (\\u003cstrong\\u003eFig. 5g)\\u003c/strong\\u003e. The system employs real-time pressure distribution analysis to detect and issue directional fall warnings for forward, backward, leftward, and rightward movements (\\u003cstrong\\u003eFig. 5h\\u003c/strong\\u003e). These results demonstrate performance comparable to commercial pressure mats, underscoring the system's dual functionality in both fall prevention and personalized rehabilitation.\\u003c/p\\u003e\\n\\u003cp\\u003eTo assess performance under dynamic loads, we analyzed stair climbing and marching in place (\\u003cstrong\\u003eVideos S5\\u003c/strong\\u003e, \\u003cstrong\\u003eS6\\u003c/strong\\u003e). The insole tracks pressure evolution across 32 foots regions, with capacitive signal heatmaps and real-time curves closely matching actual pressure patterns (\\u003cstrong\\u003eSupplementary Note 5\\u003c/strong\\u003e). Stable and precise signal outputs under high-impact and rapid movements demonstrate its reliability for sports science, injury prevention, and athletic training applications.\\u003c/p\\u003e\\n\\u003cp\\u003eBy unifying biomimetic sensing, self-sufficient energy harvesting, and AI-driven analytics, this smart insole advances wearable biomechanics beyond episodic measurements. The system’s correlation with lab-grade tools and long-term autonomy positions it as a transformative tool for preventive care, enabling early intervention in diabetic neuropathy, reducing fall risks in elderly populations, and optimizing athletic training protocols.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003e\\u003cem\\u003eAI-assisted mechanodiagnosis in static foot arch abnormality detection and dynamic gait pattern recognition using smart insole data\\u003c/em\\u003e\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eTo enable automated, clinical-grade analysis of plantar pressure distribution, we integrated a machine learning framework with a 16-sensor smart insole system. A random forest model was trained on 500 static pressure profiles (35% high arch, 35% flatfoot, 30% normal) collected during upright standing, achieving 96% classification accuracy on an independent test set. The model identifies characteristic pressure deviations in flatfoot and high-arched feet, showing strong correlation with clinical diagnoses from podiatrists and laboratory force plates (\\u003cstrong\\u003eFig. 6a\\u003c/strong\\u003e).\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eTo demonstrate the analytical effects of AI, a color-coded confusion matrix (\\u003cstrong\\u003eFig. 6b\\u003c/strong\\u003e) depicts high diagnostic precision, with pixel intensity representing classification confidence. We investigated the impact of readout layer size on recognition accuracy, revealing that reducing network size decreases accuracy, yet a ∼95% accuracy is maintained even at a network size of 65 (\\u003cstrong\\u003eFig. 6c\\u003c/strong\\u003e). t-SNE visualization of multidimensional pressure data (\\u003cstrong\\u003eFig. 6d\\u003c/strong\\u003e) revealed three distinct clusters corresponding to arch types, validating the model's capacity to resolve subclinical biomechanical signatures for early intervention in structural foot abnormalities and stress fracture prevention.\\u003c/p\\u003e\\n\\u003cp\\u003eTo resolve gait time-dependent recognition issues, we engineered the 1D-CNN to analyze time-series capacitance data from 12 gait types, including slow walk (SW), fast walk (FW), marching in place (MP), Left limp (LL), Right limp (RL), Shuffling gait (SG), Dragging walk (DW), kicking walk (KW), toe walking (TW), heel walking (HW), foot eversion (FE), foot inversion (FI). \\u003cstrong\\u003eFigure 6e\\u003c/strong\\u003e presents the capacitive signal outputs acquired by the smart insole during the identification of distinct gait patterns, alongside a schematic illustration of the 1D-CNN framework used for gait classification.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eThe gait recognition system based on 1D-CNN algorithm exhibits a classification accuracy of up to 97.6%，demonstrating excellent performance in distinguishing 12 pathological and physiological gait patterns (\\u003cstrong\\u003eFig. 6f\\u003c/strong\\u003e). From the model training process, it can be observed that the model converges rapidly, achieving an accuracy of ~97% after around 45 epochs \\u003cstrong\\u003e(Fig. 6g)\\u003c/strong\\u003e, indicating both efficient training and strong classification capability. \\u003cstrong\\u003eFig. 6h\\u003c/strong\\u003e further demonstrates the t-SNE dimensionality reduction outcomes, with well-defined clustering patterns for the 12 gait types, offering promising applications in the early diagnosis of diabetic neuropathy, osteoarthritis, and stroke-related gait asymmetry.\\u003c/p\\u003e\\n\\u003cp\\u003eThis AI-enhanced platform transforms raw plantar pressure data into clinical-grade diagnostics, detecting structural foot arch abnormalities and dynamic gait anomalies through biomimetic sensing-adaptive machine learning integration. The dual-modality system achieves 96% accuracy for static arch classification and 97.6% accuracy for 12 gait pattern recognition, surpassing the accuracies of other algorithms (\\u003cstrong\\u003eFig. 6f,\\u003c/strong\\u003e \\u003cstrong\\u003eSupplementary Note 6\\u003c/strong\\u003e), which demonstrates capabilities in early subclinical gait deviation detection (e.g., prodromal Parkinsonian gait) and personalized rehabilitation monitoring (e.g., asymmetric gait quantification). By bridging laboratory biomechanics with clinical workflows, it establishes a precision mechanomedicine framework for intelligent wearables, addressing preventive care gaps in neuromuscular disorders through telemedicine-enabled interventions like real-time fall-risk alerts.\\u003c/p\\u003e\"},{\"header\":\"Discussion\",\"content\":\"\\u003cp\\u003eThis study establishes a closed-loop wearable platform that redefines the paradigm of continuous gait analysis by integrating biomimetic sensing, energy autonomy, and clinically actionable AI. Our work advances three critical frontiers in wearable biomedical engineering, \\u003cem\\u003ei.e.\\u003c/em\\u003e, achieving laboratory-grade diagnostic precision in real-world settings, enabling self-sustaining operation across diverse environments, and bridging the gap between raw biomechanical data and precision rehabilitation, previously unrealized in existing systems.\\u003c/p\\u003e \\u003cp\\u003e \\u003cb\\u003eBioinspired dual-microstructure sensors achieve clinical-grade sensitivity in real-world settings.\\u003c/b\\u003e The dual-microstructure capacitive sensor represents a breakthrough in wearable mechanodiagnostics. Current wearable sensors for gait analysis lack the sensitivity and durability required to detect early-stage neuromuscular or musculoskeletal pathologies. By mimicking the hierarchical stress distribution of mantis legs, which are evolved to withstand dynamic loads while retaining sensitivity \\u003csup\\u003e\\u003cspan citationid=\\\"CR39\\\" class=\\\"CitationRef\\\"\\u003e39\\u003c/span\\u003e\\u003c/sup\\u003e, we resolve a fundamental trade-off in flexible sensors: achieving sub-0.1 Pa resolution while maintaining robustness under extreme pressures (\\u0026gt;\\u0026thinsp;1.40 MPa). The synergy of PEDOT-doped TPU foam (enhancing dielectric constant from 3.38 to 5.08) and PDMS pyramid arrays (amplifying air-gap effects) enables a sensitivity of 0.602 kPa⁻\\u0026sup1; and a dynamic range up to 1.40 MPa (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003ec). Critically, these sensors retain\\u0026thinsp;\\u0026gt;\\u0026thinsp;95% accuracy after 12,000 loading cycles (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003ef), making them combine clinical-grade resolution with long-term reliability for continuous monitoring. This breakthrough addresses a key barrier in mechanomedicine, \\u003cem\\u003ei.e.\\u003c/em\\u003e, detecting subclinical gait deviations in disorders like diabetic neuropathy or Parkinson\\u0026rsquo;s disease before irreversible damage occurs \\u003csup\\u003e\\u003cspan citationid=\\\"CR40\\\" class=\\\"CitationRef\\\"\\u003e40\\u003c/span\\u003e\\u003c/sup\\u003e.\\u003c/p\\u003e \\u003cp\\u003e \\u003cb\\u003eHybrid energy autonomy enables uninterrupted monitoring across diverse environments.\\u003c/b\\u003e Our hybrid PSC/Li\\u0026ndash;S system overcomes the energy bottleneck that has long constrained wearable diagnostics, particularly for elderly or mobility-impaired users \\u003csup\\u003e \\u003cspan citationid=\\\"CR41\\\" class=\\\"CitationRef\\\"\\u003e41\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR42\\\" class=\\\"CitationRef\\\"\\u003e42\\u003c/span\\u003e \\u003c/sup\\u003e. Our closed-loop PSC/Li\\u0026ndash;S system overcomes this by delivering 21.3% PCE (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003ec) and 72.15% average energy storage efficiency for long-term operation in darkness, which surpass TENG and biofuel-cell alternatives which remain impractical for chronic use \\u003csup\\u003e \\u003cspan citationid=\\\"CR43\\\" class=\\\"CitationRef\\\"\\u003e43\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR44\\\" class=\\\"CitationRef\\\"\\u003e44\\u003c/span\\u003e \\u003c/sup\\u003e. The PSCs maintain 21.3% efficiency under standard solar illumination (100 mW cm⁻\\u0026sup2;), while the Li\\u0026ndash;S batteries provide stable 3.88 V output, ensuring continuous data acquisition during daily activities (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003ef). By enabling continuous operation in sunlight-limited indoor settings and harsh outdoor environments, this platform overcomes a critical limitation of conventional energy harvesting systems (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig5\\\" class=\\\"InternalRef\\\"\\u003e5\\u003c/span\\u003ea). This energy autonomy is transformative for chronic care \\u003csup\\u003e \\u003cspan additionalcitationids=\\\"CR46\\\" citationid=\\\"CR45\\\" class=\\\"CitationRef\\\"\\u003e45\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR47\\\" class=\\\"CitationRef\\\"\\u003e47\\u003c/span\\u003e \\u003c/sup\\u003e, aligns with the WHO\\u0026rsquo;s Sustainable Development Goals by providing a low-cost (\\u0026lt;\\u003cspan\\u003e$\\u003c/span\\u003e50 projected at scale), maintenance-free solution for resource-limited settings.\\u003c/p\\u003e \\u003cp\\u003e \\u003cb\\u003eExplainable AI transforms biomechanical data into actionable clinical decisions.\\u003c/b\\u003e The dual-model architecture, random forest for static foot arch analysis (96% accuracy) and 1D-CNN for dynamic gait classification (97.6% accuracy), surpasses the accuracies of other algorithms (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig6\\\" class=\\\"InternalRef\\\"\\u003e6\\u003c/span\\u003ef, \\u003cb\\u003eSupplementary Note 6\\u003c/b\\u003e) \\u003csup\\u003e\\u003cspan citationid=\\\"CR48\\\" class=\\\"CitationRef\\\"\\u003e48\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR49\\\" class=\\\"CitationRef\\\"\\u003e49\\u003c/span\\u003e\\u003c/sup\\u003e. The algorithm\\u0026rsquo;s high accuracy and lightweight architecture (~\\u0026thinsp;1.6 MB) helps with data locally on smartphones, enabling real-time fall-risk alerts. By correlating plantar pressure maps with known pathological signatures (e.g., Parkinsonian gait, Diabetic plantar ulcer), the system has potential to provides explainable risk scores for falls or ulceration, enabling proactive care. Future iterations could integrate inertial sensors to correlate pressure maps with joint kinematics, further personalizing interventions \\u003csup\\u003e\\u003cspan additionalcitationids=\\\"CR51\\\" citationid=\\\"CR50\\\" class=\\\"CitationRef\\\"\\u003e50\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR52\\\" class=\\\"CitationRef\\\"\\u003e52\\u003c/span\\u003e\\u003c/sup\\u003e.\\u003c/p\\u003e \\u003cp\\u003e \\u003cb\\u003eToward scalable mechanomedicine: clinical integration and global health equity.\\u003c/b\\u003e Our platform\\u0026rsquo;s impact extends beyond individual diagnostics. By wirelessly transmitting HIPAA-compliant reports to telehealth platforms, this system is expected to enable clinicians to remotely monitor more patients without compromising care quality, addressing a critical need in aging populations \\u003csup\\u003e \\u003cspan citationid=\\\"CR42\\\" class=\\\"CitationRef\\\"\\u003e42\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR53\\\" class=\\\"CitationRef\\\"\\u003e53\\u003c/span\\u003e \\u003c/sup\\u003e. Pilot deployments in rural clinics will reduce fall-related hospitalizations, validating its socioeconomic impact. However, three challenges must be addressed to realize global scalability: (1) expanding sensor coverage to kinetic chains (e.g., hip-knee-ankle dynamics), (2) optimizing PSC/Li\\u0026ndash;S stability under tropical humidity (\\u0026gt;\\u0026thinsp;80% RH), and (3) validating AI generalizability across diverse populations. Collaborative trials with orthopedic centers are underway to integrate the insole with EHRs, enabling predictive analytics for conditions like osteoarthritis progression. Besides, ongoing industry partnerships are advancing scalable manufacturing (\\u0026lt;\\u003cspan\\u003e$\\u003c/span\\u003e0.12/sensor) and regulatory approval pathways.\\u003c/p\\u003e\"},{\"header\":\"Conclusions\",\"content\":\"\\u003cp\\u003eOverall, this work establishes a new paradigm for wearable biomedical devices by integrating biomimetic design, sustainable energy systems, and clinical-grade AI. The platform’s ability to detect subclinical gait pathologies, empower remote monitoring, and reduce hospitalization costs positions it as a transformative tool for global precision medicine. As healthcare pivots toward preventive and personalized models, our closed-loop approach provides a scalable blueprint for transforming wearables into essential medical devices. By unifying real-time biomechanical data acquisition, chromatic force-field visualization, and explainable AI analytics, this work transitions wearable technology from passive monitoring to actionable clinical decision-making. Its energy-independent design and scalability address critical barriers to adoption in underserved populations, where gait disorders are prevalent but underdiagnosed. Future efforts will focus on large-scale longitudinal validation of fall-risk prediction and rehabilitation tracking, as well as integration with digital health ecosystems for predictive care. This study pioneers a paradigm shift in precision healthcare, demonstrating how intelligent, autonomous wearables can democratize access to clinical-grade biomechanical diagnostics and advance global health equity.\\u003c/p\\u003e \"},{\"header\":\"Methods\",\"content\":\"\\u003cp\\u003e\\u003cstrong\\u003eMaterial\\u003c/strong\\u003e. TPU foam was obtained from Suzhou Shensai New Materials Co., Ltd. Ferric chloride (FeCl\\u003csub\\u003e3\\u003c/sub\\u003e) and ethanol were purchased from Tianjin Damao Chemical Reagent Co., Ltd. PDMS was obtained from Dow Corning Co., Ltd. 3,4-ethylene dioxythiophene (EDOT) and Methanol were purchased from Sigma-Aldrich Chemical Co., Ltd.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003ePreparation of PEDOT/TPU foam\\u003c/strong\\u003e. To prepare the oxidant solution, 0.2 g of FeCl\\u003csub\\u003e3\\u003c/sub\\u003e was mixed with 4 mL of methanol and stirred. The TPU foam was cleaned with anhydrous ethanol and deionized water, then soaked in the oxidant solution for 15 min. After removing and heating at 70\\u0026deg;C for 15 min, a device for vapor phase polymerization was set up, fixing the TPU foam on a petri dish lid and dripping EDOT monomer onto the bottom filter paper. The device was heated at 75\\u0026deg;C for 2 h, flipping the foam after 1 h for even polymerization. The TPU/PEDOT foam was then cleaned with anhydrous ethanol and deionized water to remove excess monomers, oxidant solution and other impurities.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003ePreparation of the dielectric layer with micropyramid array\\u003c/strong\\u003e. PDMS precursor and curing agent were mixed 10:1 and spin-coated on a photolithography silicon template with a micropyramid pattern at 1000 rpm for 60 s. TPU/PEDOT foam was adhered to the PDMS and cured at 80\\u0026deg;C for 3 h. Single-sided and double-sided micropyramid array PEDOT/TPU foams were prepared in turn using the aforementioned method. Electrodes and a dielectric layer were assembled using copper foil and the micropyramid array PEDOT/TPU foam, encapsulated with breathable 3M Tegaderm tape.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eCharacterization and measurement of the capacitive pressure sensor\\u003c/strong\\u003e. Fourier Transform Infrared Spectroscopy (FTIR, NEXUS870, Thermo Nicolet) and Raman Spectrometer (Xplora plus, HORIBA) were used to analyze chemical bonds and chemical structures of the composite materials. Scanning electron microscopy (SEM, S-570, Hitachi) was applied for characterizing the morphology of the materials. The mechanical and electrical properties of the sensors were analyzed simultaneously by a universal testing machine (3365, Instron) and a precision LCR meter (TruEbox-01RC, LinkZill). This LCR meter was miniaturization, portable, and featured a wireless connection.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eCell viability and staining assay\\u003c/strong\\u003e. Cell viability was assessed using CCK-8 cell viability assay kit (Beyotime, Shanghai, China) with L-929 cells seeded in 96-well plates at a density of 2\\u0026times;10\\u003csup\\u003e3\\u003c/sup\\u003e cells per well and incubated with material extract. Optical density was measured at 450 nm using a microplate reader (Biotek, Burlington, VT, USA) after 1, 2, and 3 days. AO and EB stained live and dead cells, respectively, in L-929 cells (a density of 2\\u0026times;10\\u003csup\\u003e5\\u003c/sup\\u003e cells per well) cultivated in 6-well plates with material extract for 1, 2, and 3 days, observed under a Nikon fluorescence microscope (Japan, Ti-S).\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eSolar cell fabrication and characterization\\u003c/strong\\u003e. The PSCs were deposited on the indium tin oxide (ITO) glass or PET substrates (2\\u0026times;2.5 cm\\u003csup\\u003e\\u003cspan class=\\\"CitationRef\\\"\\u003e2\\u003c/span\\u003e\\u003c/sup\\u003e, 10 Ω per square). The cleaned ITO substrates were treated by UV-ozone for 20 min. Then the SnO\\u003csub\\u003e2\\u003c/sub\\u003e film was deposited on the ITO substrate by spin-coating the SnO\\u003csub\\u003e2\\u003c/sub\\u003e precursor (5%) at 4000 rpm for 30 s and annealed on a hot plate at 150\\u0026deg;C for 30 min in air. After cooling down to room temperature, the substrate was treated with UV-ozone for 10 min before transferring to a nitrogen-filled glove box. For the perovskite film preparation, 691.5 mg PbI\\u003csub\\u003e2\\u003c/sub\\u003e and 19.5 mg CsI were dissolved in 900 \\u0026micro;L DMF with 144 \\u0026micro;L NMP added and stirred for 2 hours at 70 ℃. 90 mg FAI and 14 mg MACl were dissolved in 1 mL IPA and stirred for 30 minutes at room temperature. Then, 75 \\u0026micro;L of the mixed PbI\\u003csub\\u003e2\\u003c/sub\\u003e and CsI solution was spin-coated onto prepared SnO\\u003csub\\u003e2\\u003c/sub\\u003e at 2000 rpm for 30 s followed by one minute of annealing at 70 ℃. Then, 200 \\u0026micro;L of FAI:MACl solution was spin-coated onto the prepared film at 4000 rpm for 30 s, followed by thermal annealing at 150\\u0026deg;C for 15 min under N\\u003csub\\u003e2\\u003c/sub\\u003e atmosphere. Spiro-OMeTAD solution was spin-coated on the top of the perovskite layer at 4000 rpm for 45 s, which contains 72.5 mg Spiro-OMeTAD, 17.5 \\u0026micro;L Li-TFSI solution (520 mg/mL in acetonitrile), 28.8 \\u0026micro;L FK209 solution (300 mg/mL in acetonitrile) and 28.8 \\u0026micro;L tBP in 1 mL CB. Finally, a 100-nm-thick layer of Ag was thermally evaporated on the top of spiro-OMeTAD as electrodes. To evaluate PSC performance under sunlight, it was illuminated by a normalized AM1.5G solar simulator (100 mW\\u0026middot;cm⁻\\u0026sup2;). PSC photoelectric properties were tested using photovoltaic cell testing equipment.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eFabrication of smart flexible insole system\\u003c/strong\\u003e. The electrode and circuit of the insole were printed on the flexible polyimide substrate. A 16-channel capacitive pressure sensor array was assembled and strategically placed across the heel, midfoot, and forefoot areas of the insole. The pressure sensors were packaged and connected to a flexible printed circuit board (FPCB) using flat cables. The microcontroller unit (MCU) of the FPCB, which was based on the STM32 main controller chip, was capable of collecting capacitive signals from the 16-channel pressure sensors. The pressure data was then wirelessly transmitted to a mobile application (APP) terminal through the Bluetooth chip embedded in the FPCB. All human subject experiments were conducted with the voluntary and informed consent of the participants, with written consent obtained prior to any demonstrations involving human skin.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eStatic foot arch abnormality detection and dynamic gait pattern recognition.\\u0026nbsp;\\u003c/strong\\u003eThis study was conducted under from the Medical Ethics Committee of the Northwest University (250120003). It focuses on recognizing static foot arch types by constructing a classification model based on plantar capacitance data collected through smart insoles. During the experiment, participants wore smart insoles and maintained a static standing posture for a period of time. In the data collection process, a total of 500 valid samples were collected, consisting of 35% high arches, 35% flat feet, and 30% normal arches. The dataset was randomly split into 90% training set and 10% test set. Based on the train set, the Random Forest algorithm was used to train the model, and after parameter tuning, the optimal classifier was obtained. Finally, the model\\u0026apos;s performance was validated and evaluated using the test set, with key metrics such as accuracy and confusion matrix calculated to assess its classification performance and generalization ability.\\u003c/p\\u003e\\n\\u003cp\\u003eThis study also addresses the challenge of dynamic gait identification. Participants wore smart insoles and walked for a period of time with a specific gait, generating a dataset consisting of 2,110 samples, each containing 16\\u0026times;200 capacitance signals. The dataset was split into 80% training set and 20% test set, randomly. A 1D-CNN framework was used to train the model for 100 epochs using the training set, and the model\\u0026apos;s performance was evaluated using the test set after per epoch. Key metrics, including loss, accuracy, and the confusion matrix, were used to assess the model\\u0026apos;s performance and generalization ability. Additionally, analyzing the loss rate and accuracy curves for training set and test set can identify the most effective training epoch or epoch range, reducing overfitting, indicating that the model can generalize well to the unseen dataset. The analysis aimed to evaluate the effectiveness of the model in identifying different dynamic gaits and its potential for analysis in medical diagnosis.\\u003c/p\\u003e\"},{\"header\":\"Declarations\",\"content\":\"\\u003cp\\u003e\\u003cstrong\\u003eAcknowledgements\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThis work was financially supported by the National Natural Science Foundation of China (22308268, 22478318), National Key Research and Development Program of China (2021YFA0715600, 2021YFA0717700), Young Talent Fund of Association for Science and Technology in Shaanxi, China (20240625).\\u003c/p\\u003e\"},{\"header\":\"References\",\"content\":\"\\u003col\\u003e\\n\\u003cli\\u003eLee, J.-H., Cho, K. \\u0026amp; Kim, J. K. Age of flexible electronics: emerging trends in soft multifunctional sensors. \\u003cem\\u003eAdv. Mater\\u003c/em\\u003e. \\u003cstrong\\u003e36\\u003c/strong\\u003e, 2310505 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eTian, G. et al. Hierarchical piezoelectric composites for noninvasive continuous cardiovascular monitoring.\\u003cem\\u003e Adv. Mater\\u003c/em\\u003e. \\u003cstrong\\u003e36\\u003c/strong\\u003e, 2313612 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eLiu, D. et al. Active-matrix sensing array assisted with machine-learning approach for lumbar degenerative disease diagnosis and postoperative assessment. \\u003cem\\u003eAdv. Funct. Mater\\u003c/em\\u003e. \\u003cstrong\\u003e32\\u003c/strong\\u003e, 2113008 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eWang, X. et al. Biocompatible and breathable healthcare electronics with sensing performances and photothermal antibacterial effect for motion-detecting. \\u003cem\\u003eNpj. Flex. Electron.\\u003c/em\\u003e \\u003cstrong\\u003e6\\u003c/strong\\u003e, 95 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eLee, J. H., Cho, K. \\u0026amp; Kim, J. K. Age of flexible electronics: emerging trends in soft multifunctional sensors. \\u003cem\\u003eAdv. Mater\\u003c/em\\u003e. \\u003cstrong\\u003e36\\u003c/strong\\u003e, 2310505 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eShin, Y. et al. Damage-free dry transfer method using stress engineering for high-performance flexible two-and three-dimensional electronics. Nat. Mater. \\u003cstrong\\u003e23\\u003c/strong\\u003e, 1411-1420 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eYuan, H. et al. Progress and challenges in flexible capacitive pressure sensors: Microstructure designs and applications. \\u003cem\\u003eChem. Eng. J.\\u003c/em\\u003e \\u003cstrong\\u003e485\\u003c/strong\\u003e, 149926 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eZhang, Y. et al. Highly stable flexible pressure sensors with a quasi-homogeneous composition and interlinked interfaces. \\u003cem\\u003eNat. Commun.\\u003c/em\\u003e \\u003cstrong\\u003e13\\u003c/strong\\u003e, 1317 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eHe, R. et al. Flexible miniaturized sensor technologies for long-term physiological monitoring. \\u003cem\\u003eNpj. Flex. Electron.\\u003c/em\\u003e \\u003cstrong\\u003e6\\u003c/strong\\u003e, 20 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eZhang, C. et al. Wireless, smart hemostasis device with all-soft sensing system for quantitative and real-time pressure evaluation. \\u003cem\\u003eAdv. Sci\\u003c/em\\u003e. \\u003cstrong\\u003e10\\u003c/strong\\u003e, 2303418 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eWu, G. et al. Innovations in Tactile Sensing: Microstructural Designs for Superior Flexible Sensor Performance. \\u003cem\\u003eAdv. Funct. Mater\\u003c/em\\u003e. \\u003cstrong\\u003e34\\u003c/strong\\u003e, 2405722 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eNiu, S. et al. A wireless body area sensor network based on stretchable passive tags. \\u003cem\\u003eNat. Electron.\\u003c/em\\u003e \\u003cstrong\\u003e2\\u003c/strong\\u003e, 361-368 (2019).\\u003c/li\\u003e\\n\\u003cli\\u003eMahato, K. et al. Hybrid multimodal wearable sensors for comprehensive health monitoring. \\u003cem\\u003eNat. Electron.\\u003c/em\\u003e \\u003cstrong\\u003e7\\u003c/strong\\u003e, 735-750 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eZhang, Q. et al. A zinc-ion battery-type self-powered pressure sensor with long service life. Adv. Mater. \\u003cstrong\\u003e34\\u003c/strong\\u003e, 2205369 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eKong, L. et al. Wireless technologies in flexible and wearable sensing: from materials design, system integration to applications. \\u003cem\\u003eAdv. Mater. \\u003c/em\\u003e\\u003cstrong\\u003e36\\u003c/strong\\u003e, 2400333 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eLiu, L. et al. Strong tough thermogalvanic hydrogel thermocell with extraordinarily high thermoelectric performance. \\u003cem\\u003eAdv. Mater.\\u003c/em\\u003e \\u003cstrong\\u003e35\\u003c/strong\\u003e, 2300696 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eYu, Y. et al. Biofuel-powered soft electronic skin with multiplexed and wireless sensing for human-machine interfaces. \\u003cem\\u003eSci. Robot.\\u003c/em\\u003e \\u003cstrong\\u003e5\\u003c/strong\\u003e, eaaz7946 (2020).\\u003c/li\\u003e\\n\\u003cli\\u003eSong, Y. et al. Wireless battery-free wearable sweat sensor powered by human motion. \\u003cem\\u003eSci. Adv.\\u003c/em\\u003e \\u003cstrong\\u003e6\\u003c/strong\\u003e, eaay9842 (2020).\\u003c/li\\u003e\\n\\u003cli\\u003eLu, P. et al. Gel-based triboelectric nanogenerators for flexible sensing: principles, properties, and applications. \\u003cem\\u003eNano-Micro. Lett.\\u003c/em\\u003e \\u003cstrong\\u003e16\\u003c/strong\\u003e, 206 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eHan, M. et al. Unveiling the potential of two-terminal perovskite/organic tandem solar cells: mechanisms, status, and challenges. \\u003cem\\u003eAdv. Mater.\\u003c/em\\u003e \\u003cstrong\\u003e36\\u003c/strong\\u003e, 2402143 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eYang, W., Jo, S. H. \\u0026amp; Lee, T. W. Perovskite colloidal nanocrystal solar cells: current advances, challenges, and future perspectives. \\u003cem\\u003eAdv. Mater. \\u003c/em\\u003e\\u003cstrong\\u003e36\\u003c/strong\\u003e, 2401788 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003ePandiyan, A. et al. A comprehensive review on perovskite and its functional composites in smart textiles: progress, challenges, opportunities, and future directions. \\u003cem\\u003eProg. Mater. Sci.\\u003c/em\\u003e\\u003cstrong\\u003e140\\u003c/strong\\u003e, 101206 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eKim, J.-K. et al. Self-charging integrated energy modules: a record photoelectric storage efficiency of 14.6%. \\u003cem\\u003eJ. Energy Storage\\u003c/em\\u003e \\u003cstrong\\u003e102\\u003c/strong\\u003e, 114149 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eSalomon, A. et al. A machine learning contest enhances automated freezing of gait detection and reveals time-of-day effects. \\u003cem\\u003eNat. Commun. \\u003c/em\\u003e\\u003cstrong\\u003e15\\u003c/strong\\u003e, 4853 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eAtes, H. C. et al. End-to-end design of wearable sensors. \\u003cem\\u003eNat. Rev. Mater.\\u003c/em\\u003e\\u003cstrong\\u003e7\\u003c/strong\\u003e, 887-907 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eSong, L. et al. Phytic acid‐induced gradient hydrogels for highly sensitive and broad range pressure sensing. \\u003cem\\u003eAdv. Mater. \\u003c/em\\u003e\\u003cstrong\\u003e37\\u003c/strong\\u003e, 2417978 (2025).\\u003c/li\\u003e\\n\\u003cli\\u003eZhang, Y. et al. Highly stable flexible pressure sensors with a quasi-homogeneous composition and interlinked interfaces. \\u003cem\\u003eNat. Commun. \\u003c/em\\u003e\\u003cstrong\\u003e13\\u003c/strong\\u003e, 1317 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eBoutry, C. M. et al. A stretchable and biodegradable strain and pressure sensor for orthopaedic application. \\u003cem\\u003eNat. Electron.\\u003c/em\\u003e \\u003cstrong\\u003e1\\u003c/strong\\u003e, 314-321 (2018).\\u003c/li\\u003e\\n\\u003cli\\u003eJi, Y. et al. Splashing‐assisted femtosecond laser‐activated metal deposition for mold‐and mask‐free fabrication of robust microstructured electrodes for flexible pressure sensors. \\u003cem\\u003eSmall\\u003c/em\\u003e \\u003cstrong\\u003e19\\u003c/strong\\u003e, 2207362 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eHuang, Z. et al. Highly stretchable ionotronic pressure sensors with broad response range enabled by microstructured ionogel electrodes. \\u003cem\\u003eJ. Mater. Chem. A\\u003c/em\\u003e \\u003cstrong\\u003e11\\u003c/strong\\u003e, 7201-7212 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eKeum, K. et al. Dual-stream deep learning integrated multimodal sensors for complex stimulus detection in intelligent sensory systems. \\u003cem\\u003eNano Energy\\u003c/em\\u003e \\u003cstrong\\u003e122\\u003c/strong\\u003e, 109342 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eZhang, C. et al. Wireless, smart hemostasis device with all-soft sensing system for quantitative and real-time pressure evaluation. \\u003cem\\u003eAdv. Sci\\u003c/em\\u003e. \\u003cstrong\\u003e10\\u003c/strong\\u003e, 2303418 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eQiu, J. et al. An efficiently doped pedot: pss ink formulation via metastable liquid\\u0026minus; liquid contact for capillary flow‐driven, hierarchically and highly conductive films. \\u003cem\\u003eSmall\\u003c/em\\u003e \\u003cstrong\\u003e19\\u003c/strong\\u003e, 2205324 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eBasarir, F. et al. Edible and biodegradable wearable capacitive pressure sensors: a paradigm shift toward sustainable electronics with bio‐based materials. \\u003cem\\u003eAdv. Funct. Mater\\u003c/em\\u003e. \\u003cstrong\\u003e34\\u003c/strong\\u003e, 2403268 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eWu, L. et al. Beetle‐inspired gradient slant structures for capacitive pressure sensor with a broad linear response range. \\u003cem\\u003eAdv. Funct. Mater\\u003c/em\\u003e.\\u003cstrong\\u003e 34\\u003c/strong\\u003e, 2312370 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eHuang, J. et al. Multi‐hierarchical microstructures boosted linearity of flexible capacitive pressure sensor. \\u003cem\\u003eAdv. Eng. Mater.\\u003c/em\\u003e \\u003cstrong\\u003e24\\u003c/strong\\u003e, 2101767 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eBoutry, C. M. et al. A hierarchically patterned, bioinspired e-skin able to detect the direction of applied pressure for robotics. \\u003cem\\u003eSci. Robot.\\u003c/em\\u003e \\u003cstrong\\u003e3\\u003c/strong\\u003e, eaau6914 (2018).\\u003c/li\\u003e\\n\\u003cli\\u003eMohammedi, K. et al. Arm and ankle blood pressure indices, and peripheral artery disease, and mortality: a cohort study. \\u003cem\\u003eEur. Heart J.\\u003c/em\\u003e\\u003cstrong\\u003e45\\u003c/strong\\u003e, 1738-1749 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eYang, R. et al. Iontronic pressure sensor with high sensitivity over ultra-broad linear range enabled by laser-induced gradient micro-pyramids. \\u003cem\\u003eNat. Commun.\\u003c/em\\u003e\\u003cem\\u003e \\u003c/em\\u003e\\u003cstrong\\u003e14\\u003c/strong\\u003e, 2907 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eChoi, Y. S. et al. A transient, closed-loop network of wireless, body-integrated devices for autonomous electrotherapy. \\u003cem\\u003eScience\\u003c/em\\u003e \\u003cstrong\\u003e376\\u003c/strong\\u003e, 1006-1012 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003ePark, S. et al. Self-powered ultra-flexible electronics via nano-grating-patterned organic photovoltaics. \\u003cem\\u003eNature\\u003c/em\\u003e \\u003cstrong\\u003e561\\u003c/strong\\u003e, 516-521 (2018).\\u003c/li\\u003e\\n\\u003cli\\u003eShao, B. et al. Large-area, untethered, metamorphic, and omnidirectionally stretchable multiplexing self-powered triboelectric skins. \\u003cem\\u003eNat. Commun.\\u003c/em\\u003e \\u003cstrong\\u003e15\\u003c/strong\\u003e, 1238 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eYin, L. et al. A passive perspiration biofuel cell: high energy return on investment. \\u003cem\\u003eJoule\\u003c/em\\u003e \\u003cstrong\\u003e5\\u003c/strong\\u003e, 1888-1904 (2021).\\u003c/li\\u003e\\n\\u003cli\\u003eYin, L. et al. A self-sustainable wearable multi-modular e-textile bioenergy microgrid system. \\u003cem\\u003eNat. Commun.\\u003c/em\\u003e \\u003cstrong\\u003e12\\u003c/strong\\u003e, 1542 (2021).\\u003c/li\\u003e\\n\\u003cli\\u003eHan, Z. et al. Engineering d‐p orbital hybridization in single‐atom metal‐embedded three‐dimensional electrodes for Li\\u0026ndash;S batteries. \\u003cem\\u003eAdv. Mater.\\u003c/em\\u003e \\u003cstrong\\u003e33\\u003c/strong\\u003e, 2105947 (2021).\\u003c/li\\u003e\\n\\u003cli\\u003eSong, F.\\u003cem\\u003e \\u003c/em\\u003eet al. Mechanical durability and flexibility in perovskite photovoltaics: advancements and applications. \\u003cem\\u003eAdv. Mater. \\u003c/em\\u003e\\u003cstrong\\u003e36\\u003c/strong\\u003e, 2312041 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eXu, Y. et al. Perovskite films regulation via hydrogen‐bonded polymer network for efficient and stable perovskite solar cells. \\u003cem\\u003eAngew. Chem. Int. Edit.\\u003c/em\\u003e \\u003cstrong\\u003e62\\u003c/strong\\u003e, e202306229 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eKim, K. K. et al. A substrate-less nanomesh receptor with meta-learning for rapid hand task recognition. \\u003cem\\u003eNat. Electron.\\u003c/em\\u003e \\u003cstrong\\u003e6\\u003c/strong\\u003e, 64\\u0026ndash;75 (2023).\\u003c/li\\u003e\\n\\u003cli\\u003eLi, Z. et al. Crossmodal sensory neurons based on high-performance flexible memristors for human-machine in-sensor computing system. \\u003cem\\u003eNat. Commun.\\u003c/em\\u003e \\u003cstrong\\u003e15\\u003c/strong\\u003e, 7275 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eLuo, S. et al. Experiment-free exoskeleton assistance via learning in simulation. \\u003cem\\u003eNature.\\u003c/em\\u003e \\u003cstrong\\u003e630\\u003c/strong\\u003e, 353-359 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eAtes, H. C. et al. End-to-end design of wearable sensors. \\u003cem\\u003eNat. Rev. Mater.\\u003c/em\\u003e\\u003cstrong\\u003e7\\u003c/strong\\u003e, 887-907 (2022).\\u003c/li\\u003e\\n\\u003cli\\u003eXia, H. et al. Shaping high-performance wearable robots for human motor and sensory reconstruction and enhancement. \\u003cem\\u003eNat. Commun.\\u003c/em\\u003e \\u003cstrong\\u003e15\\u003c/strong\\u003e, 1760 (2024).\\u003c/li\\u003e\\n\\u003cli\\u003eChoi, Y. S. et al. A transient, closed-loop network of wireless, body-integrated devices for autonomous electrotherapy. \\u003cem\\u003eScience\\u003c/em\\u003e \\u003cstrong\\u003e376\\u003c/strong\\u003e, 1006-1012 (2022).\\u003c/li\\u003e\\n\\u003c/ol\\u003e\"}],\"fulltextSource\":\"\",\"fullText\":\"\",\"funders\":[],\"hasAdminPriorityOnWorkflow\":false,\"hasManuscriptDocX\":true,\"hasOptedInToPreprint\":true,\"hasPassedJournalQc\":\"\",\"hasAnyPriority\":true,\"hideJournal\":false,\"highlight\":\"\",\"institution\":\"\",\"isAcceptedByJournal\":true,\"isAuthorSuppliedPdf\":false,\"isDeskRejected\":\"\",\"isHiddenFromSearch\":false,\"isInQc\":false,\"isInWorkflow\":false,\"isPdf\":false,\"isPdfUpToDate\":true,\"isWithdrawnOrRetracted\":false,\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"researchsquare\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":true,\"externalIdentity\":\"\",\"sideBox\":\"\",\"snPcode\":\"\",\"submissionUrl\":\"/submission\",\"title\":\"Research Square\",\"twitterHandle\":\"researchsquare\",\"acdcEnabled\":true,\"dfaEnabled\":false,\"editorialSystem\":\"\",\"reportingPortfolio\":\"\",\"inReviewEnabled\":false,\"inReviewRevisionsEnabled\":true},\"keywords\":\"Flexible pressure sensors, Closed-loop power supply, Intelligent insole, Gait monitoring, AI-assisted mechanodiagnosis \",\"lastPublishedDoi\":\"10.21203/rs.3.rs-6314292/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-6314292/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"\\u003cp\\u003eContinuous gait analysis is essential for early detection and management of neuromuscular disorders, yet current wearable technologies face limitations in sensing capacity, energy autonomy, and real-time diagnostic capabilities, restricting their clinical adoption. Here, we present a biomimetic smart insole that synergizes nature-inspired sensing, self-sustaining energy harvesting, and artificial intelligence (AI) to enable continuous, clinically actionable gait monitoring. Mimicking the mechanosensory architecture of mantis legs, our dual-microstructure capacitive sensor achieves a sensitivity of 0.602 kPa\\u003csup\\u003e⁻\\u003c/sup\\u003e¹, a detection limit of 0.10 Pa, and a broad sensing range (0.10 Pa–1.40 MPa) with exceptional durability (\\u0026gt;12,000 cycles), outperforming state-of-the-art wearable sensors. A custom-designed flexible circuit wirelessly streams 16-channel pressure data to a companion APP, providing real-time visualization of dynamic force fields through chromatic mapping. The system’s energy autonomy is ensured by a hybrid perovskite solar cell/lithium-sulfur battery, enabling continuous operation across diverse environments. An embedded AI framework combines a random forest classifier (96% accuracy in foot arch abnormality detection) with a convolutional neural network (97.6% accuracy in classifying 12 pathological gait patterns), translating raw sensor data into clinical insights. This platform bridges the gap between wearable sensing and precision diagnostics, offering transformative potential for early disease detection, personalized rehabilitation, and telemedicine, and thus establishing a paradigm for next-generation intelligent wearables in global healthcare.\\u003c/p\\u003e\",\"manuscriptTitle\":\"Biomimetic Self-Powered Smart Insole with AI-Enhanced Mechanodiagnosis for Continuous Gait Monitoring\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2025-04-01 09:12:21\",\"doi\":\"10.21203/rs.3.rs-6314292/v1\",\"editorialEvents\":[{\"type\":\"communityComments\",\"content\":0}],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"researchsquare\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":true,\"externalIdentity\":\"\",\"sideBox\":\"\",\"snPcode\":\"\",\"submissionUrl\":\"/submission\",\"title\":\"Research Square\",\"twitterHandle\":\"researchsquare\",\"acdcEnabled\":true,\"dfaEnabled\":false,\"editorialSystem\":\"\",\"reportingPortfolio\":\"\",\"inReviewEnabled\":false,\"inReviewRevisionsEnabled\":true}}],\"origin\":\"\",\"ownerIdentity\":\"c49c1bcd-8f49-4b94-bfa6-fe74db51cee1\",\"owner\":[],\"postedDate\":\"April 1st, 2025\",\"published\":true,\"recentEditorialEvents\":[],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"published-in-journal\",\"subjectAreas\":[{\"id\":46349567,\"name\":\"Physical sciences/Engineering/Electrical and electronic engineering\"},{\"id\":46349568,\"name\":\"Physical sciences/Materials science\"},{\"id\":46349569,\"name\":\"Physical sciences/Energy science and technology/Renewable energy/Solar energy\"},{\"id\":46349570,\"name\":\"Health sciences/Medical research\"},{\"id\":46349571,\"name\":\"Physical sciences/Engineering/Biomedical engineering\"}],\"tags\":[],\"updatedAt\":\"2026-04-27T13:50:10+00:00\",\"versionOfRecord\":{\"articleIdentity\":\"rs-6314292\",\"link\":\"https://doi.org/10.34133/research.1063\",\"journal\":{\"identity\":\"research\",\"isVorOnly\":true,\"title\":\"Research\"},\"publishedOn\":\"2026-01-08 00:00:00\",\"publishedOnDateReadable\":\"January 8th, 2026\"},\"versionCreatedAt\":\"2025-04-01 09:12:21\",\"video\":\"\",\"vorDoi\":\"10.34133/research.1063\",\"vorDoiUrl\":\"https://doi.org/10.34133/research.1063\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-6314292\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-6314292\",\"identity\":\"rs-6314292\",\"version\":[\"v1\"]},\"buildId\":\"8U1c8b4HqxoKbykW_rLl7\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC-BY-4.0","license_restricted":false}