Long-distance inspection of water conveyance tunnels remains difficult because each measurement must retain an accurate position estimate over extended missions, thereby coupling sensing, navigation and motion control in ways that conventional modular inspection architectures do not readily resolve. Here we present a circumferential multibeam sonar (CMBS)-centric unmanned underwater vehicle system in which CMBS supports both tunnel profiling and marker-based absolute position correction. Marker-triggered Kalman updates reduce INS/DVL drift and improve positional accuracy under the designed marker-spacing conditions. Autonomous heading control maintains near-centreline motion for consistent profile acquisition, enabling successive cross-sectional profiles to be assembled into spatially co-registered tunnel geometric models. In controlled experiments, marker-based correction reduced positional error from ± 26.59 m in unaided dead reckoning to ± 0.22 m. Field deployments across three hydraulic infrastructure sites, including the South-to-North Water Diversion Project, demonstrate geometric measurement accuracy, stable operation and spatially consistent geometric modelling under in-service conditions. An unmanned underwater vehicle integrates circumferential multibeam sonar with marker-based navigation correction and autonomous heading control for in-service tunnel inspection. Marker-triggered calibration reduces long-distance positional drift and enables spatially consistent tunnel geometric models in controlled and field deployments.
Passive heat transfer enhancement utilizing local interfacial bumps is important for thermal management, yet conventional forward geometry optimization relies on costly trial-and-error parameter scanning, and empirical models often fail to capture nonlinear flow responses, temperature-nonuniformity control, and overall performance. Here, a reverse-design framework for fluid-solid interfacial bump morphologies is developed by coupling the Interface-Coupled Dynamic Mesh (ICDM) method, Large Eddy Simulation (LES), Gaussian Process Regression (GPR), and gradient-based optimization. The ICDM method enables continuous interfacial evolution from a single base mesh and achieves about 33-fold acceleration. Utilizing 400 single-bump cases and 720 homogeneous three-bump structures, GPR maps the discrete geometry-physics database onto a continuous parameter space; the predicted optima are independently verified through numerical simulations. The framework identifies optimal morphologies under unconstrained and constrained conditions, enabling efficient inverse design and large-scale optimization of passive heat transfer structures. Guangchao Yang and colleagues develop an inverse-design framework combining dynamic-mesh simulations, Gaussian-process surrogates, and constrained optimization for fluid-solid interfacial bumps. The framework accelerates database construction and identifies morphologies that balance heat transfer, temperature uniformity, and pressure loss.
The signal anomalies of global navigation satellite system (GNSS) are the key factors for the performance of shipborne strapdown gravimetry. In order to obtain reliable gravity information in real-time and at low operating cost in the conditions of signal outages or interference, we show a high-precision strapdown gravimetry method adaptively robust to GNSS signal anomalies. During signal outages, the method can reliably output gravity information by the strapdown inertial navigation system/pseudo log (SINS/PL) integration based on covariance matching adaptive Kalman filter. During interference or normal signal, the method uses the SINS/GNSS integration based on linear Huber-M estimation robust Kalman filter to perform gravimetry. The marine shipborne experiments show that, under the conditions of GNSS signal outages and interference scenarios, the overall accuracy of the method reaches 0.771mGal. By adaptively adjusting the measurement modes and advanced filters, the method can effectively handle GNSS signal anomalies. Compared to existing methods, it has superior comprehensive performance, such as higher computational efficiency and robustness. Fangjun Qin, Leiyuan Qian and colleagues report a high-precision strapdown gravimetry method adaptively robust to GNSS anomaly. This method overcomes the influence of GNSS signal interference and outages without increasing the cost of operation or equipment.
Effortless object finding by humans, even in cluttered or unseen environments, relies on the seamless integration of perception, memory, and contextual inference. In contrast, embodied robots operating under egocentric perception and partial observability frequently struggle with dynamic spatial relations and long-term consistency, leading to inefficient, repetitive search behaviors. Here we present Human-like Memory Navigation (HM-Nav), a brain-inspired architecture that bridges this cognitive gap by integrating transient sensory inputs with an evolving internal model to enable long-term object navigation. HM-Nav employs three synergistic pillars: (i) Perception: a multi-view fusion module that reconciles viewpoint inconsistencies into unified representations; (ii) Memory: an adaptive dynamic knowledge graph that accumulates semantic-spatial associations over extended timescales; and (iii) Inference: an experience-driven trajectory optimization mechanism that learns from past failures to suppress cyclic and suboptimal search patterns. Validated in simulations and real-world trials, HM-Nav demonstrates superior navigation performance and robust sim-to-real transfer, significantly outperforming existing benchmarks. Our findings suggest that emulating human-like memory structures is essential for achieving resilient, long-term autonomy in complex, open-ended environments. Ying Zhang and colleagues develop a human-like memory mechanism that enables embodied robots to achieve long-term object navigation in dynamic and unfamiliar environments. The results show that memory-guided navigation substantially improves robot autonomy, efficiency, and robustness compared with existing approaches.
To address a core bottleneck in the development of Urban Air Mobility (UAM), namely the physical simulation of complex, dynamic low-altitude wind fields, this study successfully developed and validated the large-scale integrated simulation facility named “Wind-Matrix.” Through three integrated designs oriented toward low-altitude aircraft testing, the facility extends the application scope of existing multi-fan wind tunnels and extreme-wind-field simulators: (1) It uses a horizontal wind wall composed of 81 independently programmable high-speed axial-flow permanent magnet synchronous fans, enabling precise and rapid (sub-second) control of wind speed, wind shear, and gusts. (2) It integrates a movable vertical jet, a bottom swirl flow system, and the horizontal wind wall into a synchronized facility, allowing the stable reproduction of the complete evolution of extreme wind fields—such as tornadoes and downbursts—within a single experiment. (3) It enables precise generation of complex three-dimensional non-uniform flow fields, including sinusoidal distributions, through independent coordinated control of all 81 fans, with dynamic wind-field adjustment based on real-time feedback. Experimental results demonstrate that the platform can accurately generate maximum wind speeds of 55 m/s, customizable turbulence spectra, wind shear gradients, transient gusts, tornadoes and downbursts, successfully simulating multidimensional urban wind environments ranging from steady flows to unsteady composite flows. This work helps narrow the critical “data gap” between numerical simulation and real flight, providing an essential data foundation for aerodynamic testing, control system verification, and the establishment of safety standards for aircraft. It is expected to significantly advance the safety and engineering maturity of UAM technologies. Chen Zhao and colleagues developed and validated Wind-Matrix, a large-scale low-altitude Urban Air Mobility wind simulator with 81 programmable fans, vertical jet, and swirl system. It generates 55 m/s max winds, turbulence spectra, shear, gusts, tornadoes, downbursts, and 3D non-uniform flows, bridging the critical sim-to-flight data gap and supporting aerodynamic testing, control system verification, and safety standards.
Intracortical brain-computer interfaces (iBCIs) hold promise for restoring motor, sensory, and cognitive functions, including applications in paralysis treatment and speech decoding. High-density microelectrode arrays (MEAs) provide fine spatial and temporal discrimination of neural activity, but the considerable upsurge in data generation poses challenges for wireless transmission from miniaturized implants due to constraints on power, bandwidth, heat dissipation, and device size. To address these constraints, we propose a two-stage wireless iBCI architecture comprising a transdural galvanic-coupled body channel communication (BCC) link from a free-floating MEA to an intracranial unit, followed by a transcutaneous link to an external unit. This study focuses on the transdural BCC telemetry system, which provides compact, wideband, and energy-efficient data transmission. Phantom tests, and ex vivo experiments using a human cadaveric head specimen validate the system, demonstrating wireless transmission up to 500 Mbps with 20% duty cycling and bit error rates below 10⁻⁵. The system incorporates the send-on-delta encoder (SODA), achieving up to 11.4× data compression and reducing thermal load for meeting the safety guidelines. Safety is further examined using brain-on-a-chip models, which demonstrate that the system does not evoke unintended neural activity, supporting the platform's long-term viability for high-resolution iBCIs.
Hypersonic flight and atmospheric re-entry simulations conducted in high-enthalpy ground test facilities aim to recreate the natural conditions of high altitude and thermally demanding flights. The current instrumentation used for measuring flow dynamics parameters and validating computational models lacks the capability for direct, nonintrusive measurements, largely due to its limited applicability in such hostile environments. Here we demonstrate direct and nonintrusive velocimetry measurements in the undisturbed hypersonic air flow and in the air flow near an object’s surface within an arc-heated test facility using Femtosecond Laser Electronic Excitation Tagging (FLEET). Tagging the intrinsic nitrogen molecules in air flow using a femtosecond laser and tracking them over time allowed for direct flow velocity measurements, offering high resolution while preserving the integrity of the flow. Our results pave the way towards the next generation of measurement and testing instrumentation for hypersonic aerothermal test facilities. While the characterization of the flow fields of arc-heated test facilities is complicated by the hostile testing environment, this work demonstrates that reliable flow characterization can be achieved using Femtosecond Laser Electronic Excitation Tagging (FLEET) optical velocimetry diagnostics. Aaron Gieder and colleagues report nonintrusive velocimetry measurements in both undisturbed hypersonic air flow and near an object’s surface within an arc-heated test facility.
Accurately predicting the acoustic properties of noise-reduction materials in real-world, flow-exposed environments remains a significant challenge in engineering. Classical models of microslit panels are unsuited for grazing flow conditions due to complex, nonlinear flow-acoustic interactions. To overcome this challenge, this paper proposes a machine-learning-based surface impedance prediction model for microslit panels. A unit cell comprising a microslit panel with a backing cavity is selected as the research subject. Parameters including geometric features, flow and acoustic conditions are used to create a dataset. The correction term, which accounts for the flow effect on the surface impedance, can be derived through flow-acoustic simulations using COMSOL Multiphysics. The resulting dataset is then fed into a backpropagation neural network model. The combined machine learning model, constructed by the trained backpropagation neural network model and classical microslit panel formulas, is experimentally validated and demonstrates rapid surface impedance prediction ability. Using this model, this study further conducts a statistical analysis of the coupling effects of various features on surface impedance. Our framework offers a powerful tool for the rational design and analysis of high-performance acoustic liners, with substantial application potential for aerospace and environmental noise control. Sidong Zhang and colleagues develop a Machine-Learning-based Surface Impedance Prediction model that enables fast acoustic characterization prediction of microslit panels under grazing flow. This work demonstrates that the effect of flow on acoustic resistance is predominantly positive, while its effect on acoustic reactance is negative, with a few cases where the reversal of the influence of structural parameters occurs and exhibiting complex flow-acoustic coupling phenomena.
Achieving intuitive neural control of supernumerary robotic limbs remains a major challenge in human movement augmentation. Although brain-machine interfaces (BMIs) can translate motor imagery (MI) into robotic commands, MI of non-biological effectors often produces weak and unreliable neural signals. We used electroencephalography (EEG) to examine neural signatures of MI for a supernumerary robotic thumb performed alone and concurrently with the natural thumb. Thirty-three healthy participants completed an XR-based protocol involving motor observation, motor execution, and repeated kinesthetic MI under three conditions: natural thumb, supernumerary thumb, and both concurrently. Concurrent MI, primed by motor observation and execution, elicited significantly stronger left (contralateral) alpha-beta event-related desynchronization in motor-parietal regions compared to either condition alone. Functional connectivity analysis showed enhanced motor-parietal coupling during concurrent MI, whereas supernumerary-only MI exhibited stronger occipital-to-parietal connectivity. These findings provide a neurophysiological characterization of how supernumerary and natural effectors are represented during MI and offer a basis for future studies examining their implications for BMI applications. Achieving intuitive neural control of extra robotic limbs is a challenge in human movement augmentation. Haneen Alsuradi and colleagues use EEG to show that imagining a supernumerary robotic thumb and the natural thumb together elicits stronger motor-related brain activity than imagining either alone.
In this article, I explore how urban design perpetuates gender inequity through neglecting the specific ways women experience the built environment. The women-specific dimensions are women’s disproportionate role as caregivers, their heightened perceptions of safety and exposure to harassment, and the persistence of a default-male approach in planning and design. These factors restrict women’s mobility, access to education and employment, and sense of belonging in public spaces. Drawing on interdisciplinary literature, case studies, and existing gender mainstreaming frameworks, I translate theory into practical actions for built environment professionals which covers participatory design processes, gender-sensitive planning interventions across transport and public space and increasing workforce diversity within the sector. Ultimately, I argue that gender equity as a systemic design responsibility and that professionals must actively apply existing tools and knowledge to create inclusive, safe, and accessible urban environments for all. Natasha Watson, an engineering practitioner, synthesises how caregiving, safety, and male-centred design shape the navigation of and access to our urban environments for women and girls. Drawing on interdisciplinary literature, case studies, and existing gender mainstreaming frameworks, she explored what can be done to make cities more equitable for all.
Reliable prediction of structural vulnerability under complex multiaxial loading remains challenging due to nonlinear load couplings, imbalanced failure-critical samples, and limited model interpretability. Here, we present a mechanics-informed, risk-aware learning framework integrating polynomial-harmonic feature augmentation, Weibull-based risk reweighting, and a unified Degradation Risk Score that combines stress margins and bolt pretension loss. Demonstrated on a bolted steering-knuckle assembly with finite-element-derived multiaxial load-response data, the framework improves multiple linear regression from R² = 0.37 to 0.96 with a 67% reduction in RMSE, while enhancing robustness and sensitivity in high-risk regimes across ensemble and neural network models. SHAP analysis confirms that physically meaningful multiaxial interaction features dominate predictions, revealing critical load paths associated with structural vulnerability. Finally, Bayesian logistic and Weibull calibration provide a route to link Degradation Risk Score to component-level failure probabilities, enabling probabilistic risk assessment and reliability-centred decision-making in practical engineering systems.
Seismic ambient noise enables non-invasive monitoring of high-rise buildings, but existing indicators trade floor-level localization for operational simplicity. Resonance-frequency changes track whole-building stiffness but poorly localize structural variations, whereas inter-floor interferometry localizes them through computationally intensive comparisons with reference records. Here we use interpretable machine learning to retain floor-level localization without repeated interferometric analysis. From 22 days of recordings on every floor of a 21-storey building on thick sediments in Shanghai, interferometry identifies wind-associated velocity reductions of about 2% at F6–F7 and F9–F10. An XGBoost model using only basement and roof resonance and polarization parameters predicts intermediate-floor dynamics, and its attribution profile shows transitions at the same levels. This correspondence indicates that the model retains floor-level spatial sensitivity. The candidate zones may reflect modular stiffness and load-transfer contrasts. Once trained, the model predicts the resonance frequency, vibration azimuth, and polarization dip of intermediate floors using only basement and roof recordings, without repeated inter-floor interferometry or pre-event reference records. Linpeng Qin and colleagues combine seismic ambient noise analysis with interpretable machine learning to monitor floor-level dynamics in a high-rise building on thick sediments. The model identifies transitions near floors F6–F7 and F9–F10 using only basement and roof recordings, consistent with interferometric velocity reductions.
Achieving an ultra-high dynamic extinction ratio (ER) in optical modulators is crucial for enhancing signal fidelity, dynamic range, and energy efficiency in optical communication systems. However, conventional Mach-Zehnder modulator (MZM) suffers from limited ER due to fabrication-induced imbalances, thermal drift, and imperfect Y-branch splitters, making it difficult to maintain high contrast without complex calibration or temperature control. Here, we propose and experimentally demonstrate a dual-branch ER-Boost MZM that achieves a dynamic ER over 38 dB without requiring perfect 50:50 splitters or bandwidth compromise. The architecture introduces an auxiliary extinction-enhancement branch that actively suppresses residual optical leakage through phase-controlled destructive interference. An adaptive bias controller provides sub-millisecond feedback to stabilize the ER against environmental fluctuations. Experimental results confirm long-term ER stability within a 3 dB drift, while the ER-Boost concept itself can be readily extended to silicon photonic platforms. Ziyuan Shi and colleagues develop an interference-engineered Mach-Zehnder modulator that actively suppresses residual optical leakage using an auxiliary extinction-enhancement branch and adaptive bias control. The device achieves a dynamic extinction ratio above 38 dB and significantly improves the performance of photon-counting optical communication systems.
Efficient phase modulation at mm-wave and THz frequencies is an ongoing challenge. Here, we introduce a method for high-resolution dynamic spatial phase modulation of mm-waves, via selective photo-activation of a silicon-based resonant metasurface. In contrast to conventional methods of regulating the phase of reflected waves by shifting the resonance frequency of a modulator, we control the resonant coupling strength to realise low-loss binary phase modulation. We demonstrate a system operating at a wavelength of 6 mm, realising dynamically encoded holograms enabling focussing, beam steering, multi-casting and orbital angular momentum mode generation. Moreover, we show that optical losses scale favourably with frequency, making this approach highly promising for higher frequency applications. G. Neville White and colleagues demonstrate high-resolution, dynamic spatial phase modulation of mm-waves via selective photo-activation of a silicon-based resonant metasurface. Regulating the phase by controlling the resonant coupling strength, they realise holograms enabling focusing, beam steering, multi-casting and orbital angular momentum mode generation.
The aerodynamic performance of industrial flow components is crucial in fluid transport systems in buildings, such as Heating, Ventilation and Air Conditioning (HVAC) systems, as excessive flow resistance can lead to increased pressure loss. Most of the existing low-resistance design topology optimization (TO) methods for building fluid distribution systems are based on two-dimensional models, ignoring the resistance in the thickness direction, which limits the optimization accuracy. We propose a pseudo-3D TO method. The nonlinear turbulent friction source term is incorporated into the Navier–Stokes equations to account for friction from the top and bottom walls of the corresponding three-dimensional flow channel, thereby capturing out-of-plane wall-friction effects in the two-dimensional model. The pressure loss coefficient and energy dissipation are set as multi-objectives, comprehensively characterizing resistance through boundary pressure drop and internal viscous dissipation. Z-bend verification shows that the pseudo-3D objective value is 5% lower than the traditional model value and deviates by only 4.5% from the full 3D result. The optimal Z-bend maintained a resistance reduction of 78-81% at different Reynolds numbers and aspect ratios. This pseudo 3D TO method overcomes the limitations of 2D models and provides an approach for low resistance design of industrial flow components. Yan Tian and colleagues propose a pseudo-three-dimensional topology optimization framework for resistance reduction in Z-bend channels. The results show that considering the assumed channel thickness improves the transferability of two-dimensional optimized designs to three-dimensional flow structures.
Eyeglasses are widely used to correct refractive errors, but optical fit evaluation usually emphasizes lens design alone and mainly assumes uniform use of the lens aperture. This can increase design complexity and cost while overlooking how the eye interacts with the lens during everyday wear. Here we show that favorable eyeglasses optical fit can be achieved through head-anthropometry-based parametric frame customization within the context of natural eye and lens interaction. We combined parametric frame design with mathematical lens modeling and evaluated both myopia and presbyopia groups. The results show that appropriate frame customization can maintain rational control of oblique aberrations for spherical and basic aspheric lenses without relying on high-order lens terms. Natural eye and lens interaction also produced a non-uniform pattern of lens use. These findings support a practical, frame-focused route to improving eyeglasses optical fit. Eyeglass performance depends not only on lens design, but also on how the eye uses the lens in daily viewing. Luwei Chen and colleagues propose a parametric frame customization approach to improve optical fit using head shape and eye-behavior data.
Here we present five short case studies in which key authors of papers originally published in Communications Engineering in 2022 describe the impact of their work four years on. These reflections provide a glimpse of what the papers meant for the key researchers involved, their research groups, goals, and broader communities. The pieces also give an insight into what the actual pathway to real world implementation of applied science and engineering research looks like.
Polarization-resolved near-infrared imaging adds a useful optical contrast mechanism to eye tracking by measuring the polarization state of light reflected by ocular tissues in addition to its intensity. In this paper we demonstrate how this contrast can be used to enable eye tracking. Specifically, we demonstrate that a polarization-enabled eye tracking (PET) system composed of a polarization–filter–array camera paired with a linearly polarized near-infrared illuminator can reveal trackable features across the sclera and gaze-informative patterns on the cornea, largely absent in intensity-only images. Across a cohort of 346 participants, convolutional neural network based machine learning models trained on data from PET reduced the median 95th-percentile absolute gaze error by 10–16% relative to capacity-matched intensity baselines under nominal conditions and in the presence of eyelid occlusions, eye-relief changes, and pupil-size variation. These results link light–tissue polarization effects to practical gains in human–computer interaction and position PET as a simple, robust sensing modality for future wearable devices. Mantas Zurauskas and colleagues present a single-shot polarization-enabled eye-tracking system for more robust gaze estimation in wearable devices. Across 346 participants, polarization-resolved imaging improves 95th-percentile gaze accuracy by 10–16% versus matched intensity-only imaging.
Formwork for concrete construction remains one of the most resource-intensive, polluting, and constraining components of the construction sector, relying on prefabricated molds that demand extensive material, labor, and on-site preparation. We introduce a self-deploying formwork system based on multistable tubes that transform from a compact configuration to a full-scale structural mold with integrated steel reinforcement. We demonstrate this concept by deploying a steel-reinforced 2.36 m high structure within 14 seconds. We further observed that the polymer shell of the multistable tubes increases the structural strength by a factor of three, significantly enhancing their load-bearing capacity. The demonstrated concept paves the way for the creation of concrete castings with minimal on-site intervention, enabling a scalable pathway toward rapid, low-carbon construction methods.
Nanofiber yarns offer unique opportunities for developing next-generation flexible materials that combine nanoscale functionality with macroscopic structural integrity. Their high surface-to-volume ratio, fine open porosity, and fibrous morphology enable applications in tissue engineering, drug delivery, filtration, sensors, energy storage, and functional textiles. However, the transition of nanofiber yarn technology from laboratory to industrial scale has been limited by low productivity and poor control over morphology, linear density, twist, and tensile strength. Here, we report a simple and robust method with strong scale-up potential for the continuous production of 100% nanofiber yarns using high-throughput needleless and collectorless alternating-current (AC) electrospinning. A plume of nanofibers generated from a rotating disc spinning-electrode is continuously deposited on a rotating drum and withdrawn through a spinning triangle, where a twirling device imparts controlled twist. Using this setup, nanofiber yarns composed of poly(vinyl butyral) (PVB), poly(ε-caprolactone) (PCL), polyamide 46 (PA46), and poly(vinyl alcohol) (PVA) were produced at speeds up to 55 m min⁻¹ with a 200 mm disc spinning-electrode. Scanning electron microscopy revealed densely packed nanofibrils within the yarn cross-sections and a dominant pore size around 1 µm. The yarns were braided and woven into scaffolds supporting cell adhesion and proliferation. This high-performance alternating-current electrospinning approach effectively bridges the gap between electrospun fibers—whose two of three characteristic dimensions reside in the nanoscale—and macroscopic textile manufacturing, thereby enabling a practical and scalable pathway for the industrial production of nanofiber yarns for biomedical and advanced material applications. Jaroslav Mikule and colleagues developed a continuous AC-electrospinning method for producing core-free, 100% nanofiber yarns. The method forms yarns from several polymers at up to 55 m min⁻¹ and enables their processing into braided and woven scaffolds that support fibroblast proliferation.
