Biomimetic Underwater Vehicles (BUVs), as an important research direction in the field of intelligent equipment, have made significant progress in propulsion system optimization, environmental perception and autonomous control in recent years. The widespread application of Artificial Intelligence (AI) technology has greatly promoted the intelligent and multifunctional development of BUVs in complex mission scenarios. This paper focuses on the key applications of AI technology in BUVs, including intelligent control based on reinforcement learning, intelligent navigation with fusion of multiple sensors, data-driven structural optimization design, and intelligent diagnosis and adaptive maintenance for reliability. At the same time, the core challenges currently faced by BUVs in terms of algorithm robustness, extreme environment adaptability and coordinated execution of complex tasks are summarized as well. For future development, potential research directions such as intelligent material integration, multi-environment collaborative operation, digital twin technology and efficient energy management are proposed to provide technical support for the continuous innovation and practical application of BUVs.

The growing demand for sustainable energy conversion, storage, and environmental remediation solutions has driven researchers toward materials that simultaneously deliver higher performance, greater durability, and lower environmental footprints. Conventional synthetic approaches, while widely used, often suffer from sluggish transport kinetics, mechanical fragility, and high embodied energy. Here, inspired by biological systems refined through 3.8 billion years of evolution, bioinspired materials are unlocking unprecedented capabilities across multiple domains. Benefiting from the uniqueness of these nature-derived architectures ranging from virus-templated nanostructures and coral-like hierarchies to woodaligned channels, nacre-mimetic layers, and aquaporin-inspired membranes, virus-templated solar cells now reach 10.6% efficiency, thylakoid-mimetic systems achieve 3.1% with 580 fs charge separation, bacterial nanowires deliver 307 μW cm-2 power density, coral-inspired supercapacitors attain 1661 F g-1 with 97% rate retention, nacre-mimetic electrolytes increase fracture toughness 3.1-fold while suppressing dendrites, pollen-inspired hollow structures retain 78–82% capacity after 500 cycles, hydrogenase-mimetic catalysts exhibit 2750 h-1 turnover frequency, and aquaporin-inspired designs enable 100 h continuous CO2 reduction at pH 1.0 with 60% energy efficiency; meanwhile, mucus-mimetic filters provide 25-fold adhesion enhancement, corn-based biodegradable media realize 99.9994% PM0.3 removal at 45 Pa, and MOF-COF hybrids achieve 99.8% tetracycline degradation alongside 1447 mg g-1 adsorption capacity. Life-cycle assessments further confirm their edge, with agricultural waste-derived bioinspired materials requiring only 5–10 MJ kg- 1 of embodied energy, versus 50–80 MJ kg-1 for synthetic frameworks, and demonstrating 14-day biodegradability. This review presents a design philosophy for bioinspired materials that are advancing from simple structural mimicry to full functional mastery, from laboratory demonstrations to emerging commercial relevance, and from performance optimization to complete lifecycle accountability.

To achieve human-like autonomy and adaptability in complex unstructured environments, robots must undergo a paradigm shift in multimodal perception systems by drawing inspiration from neuroscience. However, existing studies often remain at superficial descriptions of biological mechanisms, failing to deeply demonstrate how these principles can systematically guide innovation in robotic perception theory and technology. This review aims to bridge this gap by centering on insights from neuroscience to systematically construct a technological blueprint ranging from bio-inspired sensors to brain-like fusion algorithms. First, this review provides an in-depth analysis of the neural circuitry underlying multisensory integration in the brain, extracting engineering-ready computational principles. It then systematically examines cuttingedge sensor technologies such as neuromorphic vision and flexible electronic skin, emphasizing their applications in tasks like real-time state estimation. At the algorithmic level, the review focuses on how deep learning techniques, including variational autoencoders, cross-modal attention mechanisms, and spiking neural networks, can be employed to implement predictive coding and active perception in robotic systems. Finally, the article critically discusses core challenges in translating neuroscientific inspiration into engineering practice and outlines future directions such as neuromorphic computing, standardized embodied datasets, and machine self-body awareness. This work provides a clear development pathway for building next-generation embodied intelligence systems capable of genuine environmental perception and interaction.

As a key structure between Cells and Extracellular Matrix (ECM), Basement Membranes(BM)play an irreplaceable role in maintaining tissue morphology and physiological functions, and its damage is closely related to the pathological process of many organ diseases. In recent years, bionic Artificial Basement Membranes (ABM), as innovative biomaterials, have received extensive attention in tissue engineering and regenerative medicine. In the present review, we systematically sorted out the distribution characteristics of BM in various tissues and organs of the human body, and then deeply analyzed the intrinsic connection between their fine structure and physiological functions. The pathological mechanisms of BM damage and related diseases (e.g., kidney disease, pulmonary fibrosis, etc.) have been described on this basis. Then, from the perspective of material science design, the latest research progress of ABM in the direction of structural biomimetic construction, functional component optimization, and bioactive modification has been comprehensively reviewed. Finally, the potential development of biomimetic ABM in clinical translational applications has been discussed concerning the technical bottlenecks in the current research. This review aims to promote the applications of ABM in biomedical fields and provide new solutions for tissue repair and regeneration.

The spinal joint enhances the terrain adaptability of quadruped robots, but it also significantly increases control complexity and intensify energy consumption, thereby limiting their operational endurance. To address these challenges, this paper proposes an energy consumption optimization method based on hybrid force-position control and energy characteristic parameters, aiming to reduce the energy consumption of the spinal joint quadruped robot in trot gait. Firstly, based on the characteristics of the spinal joint, the kinematic model of the robot is established, and the Jacobian matrix is derived. A hybrid force-position control framework for spinal joint quadruped robot is designed. A Virtual Model Controller (VMC) for trot gait is developed to coordinate the motion of the stance and swing legs, with flexible control integrated into the spinal joint. Subsequently, the actuator energy consumption model is established, and the energy characteristics of the robot with different gait and spinal joint parameters are tested. A fitness function combining energy consumption metrics, stability indicators and heuristic reward terms is constructed, and the energy-saving strategy of the robot is designed using the improved Particle Swarm Optimization (PSO) algorithm. Finally, a prototype of the spinal joint quadruped robot is developed and tested across diverse terrains. Both the simulation and experimental results demonstrate that the method proposed in this paper effectively reduces the locomotion energy consumption of the spinal joint quadruped robot in trot gait.

This study focuses on the design, development, and testing of a soft robot that emulates the contraction pattern of the Craspedacusta sowerbyi jellyfish, which has been analyzed using image processing techniques. This prototype is mainly composed of a 20.4 cm diameter silicone body that possesses a set of backbone structures supporting a pair of braided Shape Memory Alloy wires that, once activated, produce a change in the shape of the body. Each pair of braided wires provides agonist–antagonist motion; the first has a 1.4 mm diameter and produces a bending of 2.3 cm in the backbone structure that generates the contraction of the jellyfish bell; when the second wires, which possess a 1 mm diameter, are activated, the bell of the jellyfish expands by changing the backbone structure to a straight shape. The implementation of the contraction pattern is achieved through the sequential activation of the braided wires using electric relays; this pattern takes 10 s to execute and provides a motion speed of 15.6 mm/s. The proposed architecture for the jellyfish robot allows for a smooth shape change that generates underwater motion without creating vibrations that affect the life forms in the environment and includes a camera to perform exploration tasks.

Limbless creatures such as snakes and worms can effectively move in various environments such as narrow spaces, ruins, rugged terrain, soft or muddy ground, and so on, becoming the imitation targets for biomimetic robots. However, existing worm-like robots mainly move in pipes or on a hard plane, and their adaptability to environments is still limited. This paper presents a new design and prototype of a worm-like robot that can crawl in various environments. The robot consists of a soft skin and a metameric skeleton. The skin has biomimetic setae that assist in movement, and integrates ripple and folding structures to achieve tight wrapping of the skeleton and low resistance adaptation to the large deformation of robot peristalsis. The skeleton is constructed based on a novel telescopic module that mimics the deformations of the earthworm segments through a space symmetric parallel coupled gear linkage mechanism. A kinematic analysis of the telescopic module provides the quantitative relationship between the input torque and output forces of the telescopic module to ensure necessary output forces for the robot crawling. Three gait modes of a prototype robot with four telescopic modules are introduced based on the concept of anchoring segments. A kinematic analysis method for the worm robot under different gait modes is provided. The slip rate to evaluate the efficiency of robot crawling has been defined. Experimental results indicate that the prototype robot can crawl more efficiently with more anchoring segments in various environments such as pipes, sand, water, and so on.

Recent trends in robots mechanical design include the use of compliant mechanical parts for different purposes. Here, we consider the trunk of a quadruped. Many of such mechanical parts can be modelled as flexible beams. However, classic cantilever dynamics analysis considers small bending angles of the flexible parts. When the bending is large, the resonant frequency of the beam changes as a result of mechanical deformation. Therefore, the classical cantilever beam formulation cannot be used for dynamic modelling, which is key for the development of suitable control strategies. In this work, a dynamic that takes into account the changes resulting from the large deflections of the beam is formulated. In addition, the effects on the dynamics are studied when an extra mass is connected to a rigid bar at the free end of the beam. As a result, our mathematical formulation is applicable to quadruped robotic structures composed of both rigid and flexible components, allowing the development of a control system capable of achieving energetically efficient galloping by tuning to the quasi-resonant frequency of the system. Experimental results carried out using a physical prototype representing the trunk-hind legs of a quadrupedal robot confirm the precision of the proposed model.

Development of dexterous robotic joints is essential for advancing manipulation capabilities in robotic systems. This paper presents a design and an implementation of a tendon-driven robotic wrist joint, together with an efficient Sliding Mode Controller (SMC) for precise motion control. The wrist mechanism is modelled using the Timoshenko-based approach to accurately capture its kinematic and dynamic properties, which serve as the foundation for tendon force calculations within the controller. The proposed SMC is designed to deliver fast dynamic response and computational efficiency, enabling accurate trajectory tracking under varying operating conditions. The effectiveness of the proposed controller is validated through comparative analyses with existing controllers for similar wrist mechanisms. The proposed SMC demonstrated superior performance, validated through both simulation and experimental studies. The Root Mean Square Error (RMSE) range in simulation is found to be approximately 1.67 × 10-2 radians, while experimental validation yielded an error of 0.2 radians. Additionally, the controller achieved a settling time of less than 3 seconds and a steady-state error below 10-1 radians, consistently observed across both simulation and experimental evaluations. Comparative analyses with other controllers confirmed that the developed SMC surpassed alternative control strategies in motion accuracy, rapid convergence, and steady-state precision, contributing to enhanced dexterity. This work establishes a foundation for future exploration of tendon-driven wrist mechanisms and control strategies in robotic applications.

We study the link between wing deformation and aerodynamic force production of an insect-inspired two-vein flexible flapping wing. The three-dimensional deformation of the wing is monitored during experiments, while the instantaneous aerodynamic forces are simultaneously recorded using a force balance. We demonstrate that the optimal wing stiffness distribution, controlled by the angle between the two veins, results from a subtle passive control of the phase lag between the leading and trailing edges. Specifically, we identify an optimal inter-vein angle β in the range of 15–20 degrees, which maximizes the thrust coefficient ?CT ? by enabling a phase lag of approximately π/2 between leading and trailing edge kinematics. This optimal deformation timing enhances wing curvature during peak flapping velocity, thereby improving aerodynamic efficiency. A complementary modal analysis provides a structural interpretation of the observed deformation dynamics and clarifies the role of stiffness anisotropy. Our findings indicate that insects may exploit passive structural properties to optimize flight performance without active control, highlighting the importance of venation patterns in governing aerodynamic effectiveness.

During avian flight, the wing flapping motion exhibits pronounced time-asymmetry. However, the time-asymmetric flapping data of avian species are extremely scarce, and there are little well-established theories and methodologies. Inspired by this, a study on time-asymmetric flapping strategies and control systems for morphing flight robot was conducted, with the aim of exploring the practical implementation of this mechanism in such vehicles and evaluating its effects on aerodynamic performance and energy efficiency. Through the analysis of flight videos of the Milvus migrans lineatus, time-asymmetry wing joint rotations were extracted, thereby the Tang model was proposed. Specifically, the Tang model consists of a three-rods model and avian flapping formula. In this model, the wing bones of the Milvus migrans lineatus are simplified as three hinged links. By combining a cosine-based flapping angle formula with modulation polynomials, the wing joints rotational motion of the Milvus migrans lineatus is reconstructed in uniform form. Meanwhile, the Tang model serves as a dedicated framework for wing–tail coordinated deformation and provides essential theoretical support. Simulation analyses were conducted to evaluate morphing flight robots’ performance across downstroke time ratios k and flapping frequencies to optimize the lift-to-drag ratio. Subsequently, a bionic flapping mechanism and a time-asymmetric flapping control system based on Field-Oriented Control (FOC) were designed, enabling precise adjustment of the flapping speed and the duration of the downstroke. Experimental results show that under different downstroke time ratios, the system exhibits rapid dynamic response and high steady-state accuracy. Under the condition of a downstroke time ratio k= 0.4, its energy efficiency improves by 4.8%, and the motion posture of the bionic wing closely resembles that of the Milvus migrans lineatus, demonstrating superior bionic performance. The proposed model and findings are expected to enhance aerodynamic efficiency and reduce energy consumption in morphing flight robots, providing practical guidance for future morphing flight robots’ design.

This study presents a force estimation system for the X and Y axes based on surface electromyography (sEMG) signals acquired from eight channels of the Myo armband. The ground truth forces were recorded using an ATI 6-axis force sensor in two degrees of freedom (2-DOF) setup. The proposed system aims to estimate both static directional forces (X+, X-, Y+, and Y-) and dynamic movements, providing a framework for muscle force analysis. A total of five regression models—Fine Tree (FT), Interaction Linear (IL), Medium Gaussian Support Vector Machine (MG-SVM), Exponential Gaussian Process Regression (Exponential GPR), and Wide Neural Network (WNN)—were evaluated across sixty-three case combinations derived from six features in time domain descriptors: Root Mean Square (RMS), Integrated EMG (IEMG), Waveform Length (WL), Zero Crossing (ZC), Variance (VAR), and Willison’s Amplitude (WAMP). This resulted in the assessment of 630 models for both the X and Y axes. The performance of each model was assessed using the Root Mean Square Error (RMSE). Experimental results demonstrated that the Exponential GPR model achieved the highest accuracy in both axes when utilizing the RMS and IEMG feature combination, yielding an average RMSE of 1.0550 N. The WNN followed as the second-best model, with an RMSE of 1.2631 N using IEMG and VAR in both axes. In addition to predictive accuracy, model selection also considered computational aspects such as training time, model complexity, and inference speed. Among all candidates, the WNN offered the most balanced trade-off between accuracy and efficiency, highlighting its potential for real-time force estimation in human-robot interaction applications.

Energy Storage and Return (ESR) prosthetic feet are commonly used by individuals with tibial amputations, but their fixed stiffness cannot accommodate varying walking tasks and user preferences. In contrast, semi-active prostheses can adjust stiffness through control mechanisms, though their design still leaves room for improvement in biomimetic characteristics, mass, and dynamic response. In this paper, we present a bionic, variable-stiffness, semi-active ankle–foot prosthesis. Stiffness is adjusted before each ground contact by controlling the motor to alter the length of the moment arm from the ankle joint center to the point of force application, with a range of 1.78–7.01 N·m/°. Additionally, the device features a biomimetic footplate design based on the transverse arch of the human foot, which reduces weight (total weight: 933 g) while guaranteeing support capacity. The prosthesis can precisely identify the phase for stiffness adjustment and achieve the widest range of adjustment during this phase. A unilateral tibial amputee participated in a preliminary clinical test involving variable-speed walking, stairs, and ramps. Preliminary single-subject data suggest that, compared to the subject's previous fixed-stiffness prosthesis, the newly designed prosthesis increases the range of motion, peak power, and energy storage during level walking, while reducing the adduction moment on the sound limb's knee joint. These findings offer initial support for the prosthesis' potential to improve the biomechanics of walking and other activities, providing insights for the development of biomimetic design.

Collective navigation is inspired by the long-distance migrations of biological swarms, which maintain stable collective behavior in the whole journey. Inspired by these biological navigation mechanisms, multi-agent systems can implement such strategies to eliminate their reliance on inter-agent communication. However, existing bio-inspired collective navigation methods often rely on noisy relative positioning measurements and lack sufficient fault tolerance for absolute positioning failures. To address these issues, we propose a robust bio-inspired collective navigation method for drone swarms based on information fusion and adaptive fault tolerance. At the perception level, an interactive multiple modelbased fusion framework integrates kinematic predictions with visual positioning measurements, thereby suppressing sensor noise and improving relative positioning accuracy. At the decision level, we develop an adaptive fault-tolerant bioinspired swarm decision-making algorithm, which consists of (1) a dual-gradient proximal term that enables rapid collision avoidance and swarm aggregation and (2) an adaptive migration term that allows agents to dynamically adjust decision weights based on their positioning confidence. During positioning failures, agents reduce reliance on the migration term and increase their dependence on neighbors. Numerical simulations and real-world flight experiments demonstrate that the proposed method significantly enhances visual relative positioning precision and maintains stable collective performance during positioning failures, thereby improving the collective navigation robustness.

Bilateral robotic training offers a promising solution for upper-limb motor recovery; however, conventional Mirror Therapy (MT) systems typically provide fixed assistance and rely solely on the motion of the unimpaired limb. This paper presents a bilateral intention-driven control strategy for an upper-limb rehabilitation exoskeleton that fuses motor intentions from both the healthy and affected arms. By integrating inertial signals, surface Electromyography (sEMG), and interaction torque, the system employs a fuzzy variable admittance controller combined with sliding mode control to generate adaptive assistive torque that reflects the user’s physical engagement. Experiments involving 4 healthy participants and 11 post-stroke patients were conducted under three control conditions: no interlimb coupling, fixed admittance, and variable admittance. Compared to fixed-parameter control, the proposed variable admittance strategy improved bilateral coordination by 18.6%, reduced interlimb motion lag from 1.19 s to 0.54 s, enhanced intention consistency by 18.33%, and increased average interaction torque by 23.13%. These results demonstrate that combining bilateral intention fusion with adaptive admittance control can significantly enhance movement synchronization and voluntary engagement, offering a promising control paradigm to enhance patient engagement and motion coordination for rehabilitation robotics.

This study introduces a novel, cost-effective, and user-friendly methodology for evaluating the performance of upper limb exoskeletons. The proposed experimental procedure is demonstrated through the assessment of a passive exoskeleton developed and assembled at the University of Calabria. The evaluation method focuses on analyzing the stability of key anatomical landmarks on the user, tracked using color-coded markers during the execution of a predefined task. The measurement system employs two low-cost, commercially available cameras, each capable of detecting three distinct markers attached to the user’s limbs. This setup enables accurate tracking of joint motion without the need for expensive motion capture systems. The principal innovation of this work lies in democratizing exoskeleton performance evaluation through a dual-camera, marker-based system that offers comparable assessment capabilities to high-end solutions at a fraction of the cost. Additionally, the modular design enables adaptation to different exoskeleton configurations and anatomical regions, supporting systematic evaluations across diverse applications while facilitating wider adoption of assistive technologies.

During stop-jumping landings, variations in toe box space (TBS) may influence lower limb movement strategies. This study aimed to investigate the effects of TBS on metatarsal (MT) stress distribution and lower limb biomechanics. Twentyeight basketball players participated, wearing tight-fitting (TF) and loose-fitting (LF) shoes. An integrated foot-ankleshoe-knee finite element model (FEM) was established to obtain MT1-MT5 stress. The peak metatarsophalangeal joint dorsiflexion angle (PMDA) was measured and its relationships with peak MT stress and 13 ankle and knee parameters were analyzed. TF shoes significantly restricted metatarsophalangeal joint (MPJ) dorsiflexion (P<0.001), and PMDA was correlated with all variables except peak ankle inversion angle (PAIA) (R2= 0.209). MT stress was more concentrated in the central region. These findings suggest that restricted TBS limits MPJ mobility, and lower PMDA is linked to central MT2 stress distribution and compensatory loading at proximal joints, potentially increasing injury risk. This study provides biomechanical insights for basketball shoe selection and injury prevention in high-impact sports.

The lower-limb prosthesis is used to assist patients with dysfunction of motor dysfunction or aging through Brain-Machine Interface (BMI) based on Electroencephalography (EEG) signals to control cognitive tasks. This paper presents a remarkable model to improve the estimation of the EEG signal and further help improve the control performance for the lowerlimb prosthesis, and then improve the rehabilitation. It is based on an optimized Multiclass Support Vector Machine (MSVM) using Snake Optimizer (SO) to get the best possible parameter tuning for classifying different cognitive tasks to control of lower-limb exoskeleton. A public EEG dataset for a lower-limb exoskeleton using Motor Imagery (MI) during the control of the prosthesis and attention to gait (Att) on two surfaces, including flat (Experience) and non-flat (Slopes), has been used as benchmark data sets for this work. The results of the proposed model revealed the superiority of this technique in accuracy, compared with two optimization methods, including Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). By comparing the outcomes of SO-MSVM with state of the arts, it achieved an accuracy of more than 85% for MI and Att metric, demonstrating intriguing results for solving the rehabilitation challenge. The devised technique could help people with neurological conditions who have trouble using manual controls.

Displacement sensors with large measurement ranges, high linearity and low hysteresis are crucial for wearable devices, underwater environments and precision numerical control machines. However, due to the limitations of design principles, traditional displacement sensors find it difficult to achieve these characteristics simultaneously. This work presents a capacitive displacement sensor with a variable-angle plate. Moreover, the sensor exploits edge capacitance to extend the linearity interval. The sensor provides a measurement range of 10 mm with a resolution of 30 μm in contact sensing, a linearity exceeding 0.9967, and a response time of less than 30 ms. Additionally, the sensor possesses non-contact sensing capabilities, enabling it to detect the approach and departure of objects such as hands. This dual-mode displacement sensing design strategy exhibits significant application potential in wearable devices, underwater activities, and the safe production of mechanical workshops.

Multidimensional flexible sensing units serve as key components for realizing tactile perception. However, existing technologies mainly focus on uniaxial normal pressure detection and have limitations in directional sensing of tangential forces. Inspired by petal structures, here we propose a multi-dimensional force flexible sensor with a multi-lobed symmetric electrode structure. Through Finite Element Method (FEM) analysis of the electrode structure, the design demonstrates improved sensitivity characteristics to both normal and tangential displacement. An integrated experimental-FEM model established quantitative relationships between applied forces and capacitive responses, achieving prediction of normal force magnitude, tangential force magnitude, and tangential direction. Validation via a dual-mode force-controlled testbed demonstrated multidimensional sensing capabilities with: 5.7% normal force error, 5.5° tangential directional error, and 0.14 N tangential magnitude error. A preload-dependent correction factor was further developed to compensate for detection deviations under varying compression states. This work provides technical support for multidimensional force tactile perception.

The airflow vector information is crucial for ensuring flight safety and flow field identification for the increasingly miniaturized drones. However, the measurement of the airflow velocity and direction on small-sized and compatible sensors is still challenging. Natural creatures, such as scorpions, use the trichobothria organ to resolve airflow vectors in a complicated airflow field. Inspired by this sensing capability, we investigate the elliptical-top and cylindrical-root structure of the trichobothria hair shaft. The elliptical top enlarges the deflection difference of the hair shaft between the airflow from sensitive and insensitive directions, which makes the hair shaft an airflow-vector sensor. To replicate this capability, a Bioinspired Elliptical-top and Cylindrical-root Hair sensor (BECH) is designed. The sensor exhibits sensitivities of 16.46 and 9 mV/m/s for the sensitive and insensitive directions, respectively, and detects the airflow velocity and direction with the resolution of 0.064 m/s and 5.08°. To accomplish the recognition of airflow vector, a BECH array was constructed. Based on a Back Propagation Neural Network (BP-NN), the airflow vector is accurately detected with Root Mean Squared Error (RMSE) of 0.00186 and 3.149 for the velocity and direction, respectively. The proposed BECH, inspired by scorpions, provides a new concept for the design of hair-like sensors aiming to decouple airflow vectors.

The connection between metal and polymer is crucial for lightweight manufacturing in the electronics, automotive, aerospace industries and so on. Inspired by biological curves, this study proposes a novel biomimetic interlocking structure, designed specifically to enhance the metal-polymer joint strength. Through this approach, superior connection strength is achieved compared to conventional structures. Three different interlocking structures—Dragonfly Head-Neck interlocking structure, Kelvin interlocking structure, and the propopsed Curved Body-Centered Cubic Lattice interlocking structure— were additively manufactured with stainless steel, followed by injection molding to form the metal-polymer connection structures. The bonding performance of these structures was evaluated through finite element analysis and experiment. The results indicate that the Curved Body-Centered Cubic lattice interlocking structure with a 10% fill rate exhibited the highest bonding strength, outperforming both the Kelvin interlocking structure and the Dragonfly Head-Neck interlocking structure. Reducing the stiffness of the metal subatrate near the metal-polymer connection rigon can establish efficient load transfer path, which leads to a uniform stress distribution within the polymer, and allows the polymer to better withstand tensile forces during loading, finally achieve the goal of enhance the bonding strength of of the metal-polymer jointing. This research offers an innovative approach to enhancing mechanical connection interface strength, with significant implications for improving the durability and performance of metal-polymer composites.

Owing to their excellent strength-to-weight ratio, honeycomb sandwich structures have long been favored in impact mitigation and energy absorption applications. Integrating wood-plastic composites into 3D-printed bioinspired sandwich designs offers a sustainable path toward enhanced mechanical performance. In this study, the out-of-plane compressive behavior of two 3D-printed bioinspired configurations, traditional Honeycomb Plate (HP) and End-trabecular Beetle Elytron Plate (EBEP), was systematically investigated using both experimental testing and finite element simulations. Fabrication of the sandwich plates involved two material systems: Polylactic Acid (PLA) and PLA reinforced with Wood Fibers (WF). In addition, the microstructures of the two material systems were examined, and a cost analysis was conducted. Results demonstrate that the incorporation of WF significantly improves both the specific load-bearing capacity and specific energy absorption of the structures, with EBEP showing greater performance gains than HP. Microstructural analysis reveals that the performance enhancement stems from multiple reinforcement mechanisms contributed by the WF. The close agreement between experimental and simulation results confirms the reliability of the proposed model and sheds light on the reinforcing role of WF. The EBEP structure made from PLA/WF demonstrates a lightweight, high-strength, and cost-effective solution. These findings offer theoretical support for the development of cost-effective and eco-friendly high-performance energy-absorbing structures.

This study presents a novel bionic thin-walled tube with a complex cross-section (BS), inspired by the protective thorns of the durian fruit, to enhance crashworthiness and impact protection. The spiky durian shell dissipates impact energy and shields non-impact regions, motivating the biomimetic design. Quasi-static compression tests demonstrate that BS5 achieves a 9% higher Specific Energy Absorption (SEA) than Sinusoidal corrugated tubes (SIN) and 22% higher than Double Corrugated Tapered tubes (DT). The three BS configurations also exhibit 31–60% higher Crushing Force Efficiency (CFE) while reducing Undulation of Load-carrying Capacity (ULC) by 26–74%, resulting in smoother force–displacement responses. Bulkheads provide no advantage in axial energy absorption; in fact, BS5 without bulkhead achieves a 33% higher SEA. Structures with polygonal mid-sections show lower imperfection sensitivity than square ones, and when the side length-to-thickness ratio is preserved, larger-scale structures retain stable absorption efficiency, indicating potential for protective applications such as shelters. By integrating two antiprism units, BS5 delivers 45% higher SEA, 71% higher CFE, and 35% lower ULC than a single antiprism tube. A calibrated linear elastic constitutive model accurately predicts crushing behavior under quasi-static loading. Unlike SIN and DT double-layer structures that collapse in thick-walled modes, deformation analysis reveals that double-layer BSs possess an optimal interlayer spacing to avoid such degeneration, exhibit a more complex plastic hinge evolution mechanism and higher energy absorption efficiency. Low-velocity drop-weight tests confirm superior impact resistance of BSs compared with SIN and DT, while high-velocity simulations reveal only a 21% increase in Initial Peak Crushing Force (IPCF) from 5 to 50 m/s, demonstrating robust dynamic performance. Overall, the durian-inspired BS tubes exhibit excellent crashworthiness and strong potential for advanced defense and engineering impact protection applications.

Modern ballistic protection equipment demands a critical balance between high impact resistance and lightweight, a challenge unresolved by conventional approaches. Biological systems achieve such synergy through evolutionarily optimized structures, offering promising biomimetic solutions, yet a limited understanding of impact-resistance mechanisms hinders their full potential. This study employed a multi-technique approach combining ultra-depth-of-field microscopy, SEM, EDS, FTIR, nanoindentation, and impact testing to investigate the convex hull of Lq (desert scorpion) tergum and the locally thickened regions at the hulls relative to the inter-hull gaps. Convex hulls deflect and disperse impact loads, while local thickening guides cracks and dissipates energy, balances impact resistance, and lightweighting. Hp tergum (rainforest scorpion) served as a control, confirming this convex hull–local thickening dual mechanism as a key adaptation of desert scorpions to particle impacts. Finite element models of the central (BM1) and lateral (BM2) convex hull arrangements were subjected to particle impacts. The lateral (BM2) configuration exhibited superior stress regulation and damage mitigation. Guided by this lateral prototype, bio-inspired ceramic-fiber protectors were designed and fabricated. Ballistic tests (7.62 mm armor-piercing incendiary projectiles, 800–815 m/s) showed a 24.9% reduction in BFS compared with conventional flat composites of equal areal density, lowering potential human impact. Its impact-resistance mechanism, which involves projectile deflection combined with stress homogenization and crack guidance, closely replicates the desert scorpion tergum’s biological mechanism. This work provides a mechanism-driven design paradigm for lightweight composite armor.

Voronoi structures are widely present in nature, and highly ordered Voronoi structures such as honeycomb structures have gained extensive recognition and in-depth research in the field of sound absorption structure design. However, Voronoi structures in biological tissues are not all highly ordered. Stochastic Voronoi structures are equally prevalent and exhibit excellent multifunctional properties. To further explore the acoustic value of stochastic Voronoi structures, this study proposes a Voronoi sound absorbing porous structure that features both structural stochasticity and performance robustness. First, a theoretical calculation model is established based on microperforated panel theory and Helmholtz resonance theory, enabling the rapid calculation of the structure’s sound absorption coefficient. Then, a systematic analysis is conducted on the effective conditions for absorption performance robustness from four dimensions: unit number, structural randomness, manufacturing errors, and boundary cutting. Results indicate that there exists a unit number threshold associated with absorption bandwidth in the Voronoi structure. When this threshold is exceeded, the structure can exhibit favorable sound absorption robustness against structural stochasticity, manufacturing errors, and boundary cutting. Experimental verification shows that under significant boundary changes, the structure still maintains an average sound absorption coefficient of approximately 0.8 within an absorption bandwidth of approximately 400 Hz. Its favorable low-frequency broadband sound absorption performance and robustness endow it with promising application prospects in engineering fields where cost control, environmental adaptability, and construction efficiency need to be balanced.

Poly-Ether-Ether-Ketone (PEEK) emerges as a reliable thermoplastic used in various orthopedic applications, notably for its mechanical properties and mitigating stress shielding properties. Its thermoplastic nature enables 3D printing, allowing the use of established techniques for fabricating Triply Periodic Minimal Surface (TPMS) architectures like Gyroid, Diamond, and Primitive. This study proposes four strategies to explore the Young’s modulus E and ultimate strength σ? design space: uniform latticing and three functional grading configurations based on relative density, unit cell size, or their hybrid combination. PEEK-based lattices are fabricated, and deviations between printed samples (as-built) and designed models are evaluated. The apparent mechanical properties: ultimate and yield strength (σ?, σY), toughness U, and Young’s modulus E, are estimated. The 3D-printed samples demonstrated high fabrication fidelity, with sheet thickness deviations of ~ 9.5% for uniform and ~7–17% for graded lattices. In situ tracking of spatial variations within unit cells reveals how grading influences Poisson’s ratio and auxetic behavior. Overall, PEEK scaffolds achieve E values within ~7% of trabecular bone and exceed its strength by over >11.8 times. To objectively assess property matching, we introduce a Bone Matching Index using radar plots to quantify deviations in σ?, σY, E, and U. Among TPMS architectures, Gyroid matches trabecular bone most closely in E, while the hybrid Primitive configuration shows the most balanced mechanical response and highest index.

In scaffolds designed for ideal bone healing, bone regeneration efficacy is a crucial consideration. Our recent findings indicate that the contact area between the defect and scaffold significantly influences regeneration efficacy, with experimental evidence showing that morphological factors at this interface can enhance bone regeneration. Building upon these insights, this study experimentally examines how the directionality of channels within the scaffold affects bone regeneration efficacy. We designed two scaffold types: a Non-Radial-Stacking Scaffold (NRS) and a Radial-Stacking Scaffold (RS). Both scaffolds were engineered to have identical porosity, pore size, and defect-scaffold contact areas, differing only in channel direction. Numerical analysis of oxygen diffusion demonstrated that the RS scaffold facilitated better diffusion compared to the NRS. The in-vivo results further indicated that the RS scaffold significantly enhanced bone regeneration relative to the NRS. These findings suggest that prioritizing channel direction in scaffold design is essential for optimizing regenerative outcomes.

LiDAR SLAM plays a vital role in the automation of engineering vehicles, yet most existing algorithms are designed for static environments and face challenges when applied to dynamic construction scenarios. To address this, this paper proposes a novel dynamic LiDAR SLAM algorithm, Dyna-LIO, inspired by amphibian vision principles. First, a bio-inspired mapping framework is established to systematically describe the theoretical connection between biological principles and SLAM modules. Then, linear interpolation between LiDAR odometry and IMU odometry is performed to initialize the pose of the current frame; ground points are extracted and the point cloud is transformed into depth images and bird’seye views for dynamic object detection and removal. Finally, geometric feature-based registration of static point clouds is carried out, followed by global optimization using a factor graph to obtain high-precision LiDAR odometry results. Validation on the KITTI dataset shows that the proposed method achieves significant improvements in localization accuracy compared with existing algorithms in dynamic environments. Simulation experiments in typical construction and mining scenarios also demonstrate a substantial reduction in localization errors, further verifying the applicability and robustness of Dyna-LIO in real operational environments.

This work proposes a system that uses wireless sensor nodes and surface electromyography (sEMG) data to monitor and recognize human activities in bed. The contributions of this work are that we implement 2.4 GHz IEEE 802.15.4 wireless sensor nodes combined with a human body-attached sEMG sensor, where sEMG data are wirelessly sent to a receiver connected to a computer that serves as the processing center. Human activities in bed, including rapid breathing, seizure sleeping, falling from the bed, and lying on the ground as the critical events, are considered. Activity recognition is carried out using a machine learning-based classification framework with 49 and 73 features and 31 classifiers, where the effects of the number of features and variety of datasets are also studied. Experimental results demonstrate that with 49 features and using datasets from all subjects, average classification accuracies of 79.7% (training) and 80.9% (testing) can be obtained from the ensemble bagged trees. Additionally, results indicate that using datasets from all participants for training and testing results in a poorer classification accuracy than using datasets from each subject separately. For 73 features, with the ensemble bagged trees, classification accuracies are 99.5% for both training and testing. Particularly, there are 99.5% of rapid breathing, 99.7% of seizures, 97.9% of falling from the bed, and 99.8% of lying on the ground. The Fine-KNN classifier also provides satisfactory accuracies of 92.1% and 93.0%, where the prediction speeds and the training times are better.

Accurate skin cancer diagnosis is vital for early treatment and improved patient outcomes. Deep learning models have shown promise in automating skin cancer classification, yet challenges remain due to data scarcity and limited uncertainty awareness. This study presents a comprehensive evaluation of deep learning-based skin lesion classification with transfer learning and UQ on the HAM10000 dataset. We benchmark several pre-trained feature extractors (including Contrastive Language-Image Pre-training (CLIP) variants, ResNet50, DenseNet121, VGG16, EfficientNet-V2-Large, and ConvNeXt Large) combined with traditional classifiers such as SVM, XGBoost, and logistic regression. Multiple PCA settings (64, 128, 256, 512) are explored, with LAION CLIP ViT-H/14 and ViT-L/14 at PCA-256 achieving the strongest baseline results. In the UQ phase, Monte Carlo Dropout (MCD), Ensemble, and Ensemble Monte Carlo Dropout (EMCD) are applied and evaluated using uncertainty-aware metrics (UAcc, USen, USpe, UPre). Ensemble methods with PCA-256 provide the best balance between accuracy and reliability. Further improvements are obtained through feature fusion of top-performing extractors at PCA-256. Finally, we propose a feature-fusion–based model trained with a Predictive Entropy (PE) loss function, which outperforms all prior configurations across both standard and uncertainty-aware evaluations, advancing trustworthy deep learning-based skin cancer diagnosis.

Imbalanced data classification is one of the most critical challenges in machine learning. Standard classifiers do not provide adequate accuracy for detecting minority samples because these classifiers are biased toward the majority samples. To overcome this drawback, many methods have been proposed. One prominent method is the Gravitational Fixed Radius Nearest Neighbor (GFRNN) algorithm, which applies Newton’s law of universal gravitation to determine the class of a test sample based on two parameters: mass and radius. Although GFRNN shows good performance on some imbalanced datasets, it faces several fundamental problems, including ignoring the data distribution and the improper calculation of radius and mass. In this study, a Gravitational Dynamic Radius Nearest Neighbor trained by a Fuzzy Enhanced Hiking Optimization Algorithm (FEHOA-GDRNN) is proposed to improve GFRNN performance. In FEHOA-GDRNN, Enhanced Hiking Optimization Algorithm (EHOA) applies a new spider web search to find better solution based on a Mamdani Fuzzy Inference System (FIS). FEHOA-GDRNN is evaluated on 40 imbalanced datasets and its results are compared with GDRNN trained by Fuzzy HOA (FHOA-GDRNN), GFRNN trained by HOA (HOA-GFRNN), GFRNN and its various versions (IGFRNN, I-GFRNN, and EGDRNN), Cost-Sensitive Support Vector Machine with an RBF kernel (CSSVM-RBF), Cost-Sensitive Support Vector Machine with a Linear kernel (CS-SVM-Linear), Cost-Sensitive Na?ve Bayes (CS-NB), Binary Decision tree (BDT), Random Forest (RF), Gaussian-Probabilistic Neural Network (GaussianPNN) and Skew-Probabilistic Neural Network (SkewPNN). Moreover, the results of the proposed classifier are compared with those of several Fuzzy K-Nearest Neighbor (FKNN). The results demonstrate that FEHOA-GDRNN outperforms other methods in key metrics, including Average Accuracy (AAcc) and Geometric Mean (GM).

Accurate identification of diverse skin conditions is critical for effective patient care, particularly as Monkeypox (Mpox) has emerged as a significant global concern with an increasing number of cases worldwide. Individuals affected by Mpox often exhibit distinctive skin symptoms, and the disease carries a risk of transmission through contact, highlighting the need for reliable diagnostic systems. Existing studies in skin disease detection predominantly rely on single convolutional neural networks or traditional ensemble methods, which may struggle to generalize across diverse conditions. To address these limitations, we propose a novel fuzzy rank-based ensemble framework for the detection of multiple skin diseases. Our approach employs transfer learning to enhance the feature extraction capabilities of multiple base learners, including DenseNet169, MobileNetV2, and ResNet101V2, augmented with additional fully connected layers. Fuzzy logic principles are then applied to aggregate decision scores from these models, using two non-linear functions to adaptively weigh the confidence of each base learner for every input sample. We conduct a comprehensive evaluation on three datasets, comprising two Mpox datasets and one multi-class skin disease dataset, comparing our method against baseline models and recent state-of-the-art approaches. Results demonstrate that the proposed ensemble improves classification reliability and robustness, while interpretability analyses using Grad-CAM and LIME offer insights into model decision-making. Despite its effectiveness, the approach depends on dataset quality and diversity, and the inherent complexity of deep learning models may limit full interpretability in clinical scenarios. Overall, the proposed framework provides a confidence-aware, adaptable solution for accurate and reliable skin disease diagnosis.

Accurate and quick diagnosis of COVID-19 remains a global challenge, as conventional RT-PCR testing suffers from high false-negative rates, delays, and limited availability. Chest X-ray imaging, while more accessible, requires expert interpretation and is prone to human error. Although several deep learning methods have been applied, many face issues of overfitting, insufficient feature selection, and reduced reliability. In order to overcome these constraints, we suggest a hybrid framework, COVID-19-DSCNN-XISSP, which integrates a Deep Spatial Convolutional Neural Network (DSCNN) for robust spatial excision of features with the Exponential Incremental Social Ski Driver with Sine Cosine Algorithm (XISSP) for optimal feature selection. This strategy enhances classification accuracy while reducing computational overhead. Extensive experiments on benchmark COVID-19 chest X-ray datasets demonstrate that the proposed model outperforms existing approaches in terms of accuracy, precision, and robustness. The findings highlight the potential of COVID-19-DSCNN-XISSP as an economical, automated, and expandable tool for assisting radiologists in early and reliable detection of COVID-19,

Object detection in remote sensing images presents numerous challenges, such as significant variations in object scale and pronounced differences in aspect ratios. Current mainstream approaches typically employ large convolutional kernels or multi-scale convolutional architectures to address these challenges. However, large convolutional kernels tend to introduce substantial background noise, while the use of kernels with varying scales may lead to feature information redundancy. To tackle these issues, we propose a robust and efficient rotation-aware object detection method–MKS-YOLO. MKS-YOLO explicitly decouples high- and low-informative features through an information-aware threshold gating mechanism, effectively suppressing background noise interference. Furthermore, to reduce noise and enhance the model’s ability to perceive objects at multiple scales, we design a Multi-Kernel Dynamic Selection Module (MKSM). This module integrates multiscale convolutional kernels and introduces a spatial attention mechanism to achieve dynamic weighted feature fusion. In addition, we incorporate a Contextual Large Kernel Attention (CLKA) mechanism to enhance feature representation capabilities for objects with high aspect ratios. Moreover, during the feature extraction stage, we introduce an Attention-based Internal Feature Interaction module (AIFI), which further strengthens the correlations among deep semantic features. To validate the effectiveness of the proposed method, we conduct extensive experiments on two widely used remote sensing image datasets: DOTA-v1.0 and DIOR-R. Experimental results demonstrate that MKS-YOLO achieves state-of-the-art detection accuracy, with Mean Average Precision (mAP) reaching 77.71% and 82.50%, respectively, fully validating the superior performance of the proposed method.

The Distributed Flexible Job Shop Scheduling Problem (DFJSP) is critical in modern manufacturing; however, existing research has not sufficiently addressed dynamic disturbances, particularly unexpected worker absences. This study extends the DFJSPW model to introduce an enhanced framework, DFJSPWA, which optimizes maximum makespan, worker workload, and total energy consumption by integrating worker load factors with random absenteeism. To solve this complex problem, we propose a Q-learning-based Hyper-heuristic Evolutionary Algorithm (QLHHEA). This algorithm features a segmented encoding scheme that implicitly captures absenteeism information, utilizing a decoding process tailored for both standard and rescheduling contexts. Additionally, we construct a pool of twelve efficient Low-Level Heuristics (LLHs) combined with Q-learning to enable the adaptive selection of operators. Furthermore, a Hybrid Rescheduling Method (HRM) is developed, employing three response strategies based on worker status and the urgency of the absenteeism. Comprehensive experiments on 58 benchmark instances demonstrate that QLHHEA significantly outperforms six established algorithms, including MOEA/D and NSGA-II. Statistical validation confirms the superiority of the proposed method. This research provides a robust theoretical and methodological framework for improving scheduling efficiency and resource utilization in distributed production systems facing worker absenteeism.

To tackle the issue where current techniques for combining infrared and visible imagery often fail to equilibrate thermal saliency and background textures in complicated heterogeneous scenarios (especially under few-shot scenarios), we present a subregional adaptive framework inspired by the rattlesnake visual mechanism. The method first employs an enhancement model inspired by the rattlesnake visual mechanism to improve the contrast and details of infrared images. Subsequently, a lightweight object detection network is integrated to precisely partition the image into target and non-target regions. Differentiated fusion strategies are designed for these regions: focusing on preserving infrared thermal radiation information in target regions, while maintaining visible texture structures in background regions, with a Gaussian blur mechanism ensuring smooth transitions. Finally, a colorization step is employed to enhance visual perception by restoring chromatic information to the fused image. Extensive evaluations on widely used benchmark datasets demonstrate that our framework consistently outperforms state-of-the-art bio-inspired and deep learning methods. Qualitatively, it significantly excels in enhancing the thermal contrast of targets and preserving high-fidelity background textures. Quantitatively, it delivers notable improvements in key metrics such as Edge Preservation (QAB/F ), Average Gradient, and Cross Entropy, reflecting superior edge information transfer and structural fidelity. Most importantly, the proposed framework achieves these superior results specifically under few-shot settings, demonstrating exceptional data efficiency and robust generalization.