Communications engineering最新文献

Neuronal networks have driven advances in artificial intelligence, while molecular networks can provide powerful frameworks for energy-efficient information processing. Inspired by biological principles, we present a computational framework for mapping synthetic gene circuits into bio-inspired electronic architectures. In particular, we developed logarithmic Analog-to-Digital Converter (ADC), operating in current mode with a logarithmic encoding scheme, compresses an 80 dB dynamic range into three bits while consuming less than 1 µW, occupying only 0.02 mm², and operating at 4 kHz. Our bio-inspired approach achieves linear scaling of power, unlike conventional linear ADCs where power consumption increases exponentially with bit resolution, significantly improving efficiency in resource-constrained settings. Through a computational trade-off analysis, we demonstrate that logarithmic encoding maximizes spatial resource efficiency among power consumption and computational accuracy. By leveraging synthetic gene circuits as a model for efficient computation, this study provides a platform for the convergence of synthetic biology and bio-inspired electronic design.

The proliferation of distributed energy resources introduces multi-source uncertainties, including implicit uncertainties arising from third-party operators' partial observability of security constraints, challenging traditional distribution network planning methods dependent on model simplification and predefined scenarios. We address this gap via an adaptive hierarchical learning architecture that co-optimizes distributed energy resources location, capacity, and operational strategies data-drivenly, enabling autonomous learning of implicit constraints without full model knowledge. Our framework embeds a bi-level Stackelberg structure where Monte Carlo Tree Search autonomously generates planning schemes at the upper level, while multi-agent reinforcement learning directly learns operational policies from real-time data at the lower level under partial observability. Validation on both benchmark and large-scale practical distribution systems shows lower investment costs and faster solutions while maintaining voltage stability, demonstrating superior scalability and adaptiveness to implicit uncertainties versus scenario-based methods.

Intelligent metasurfaces, capable of shaping the electromagnetic field, have been extensively investigated in diverse scenarios, including beamforming, wireless communication, and electromagnetic imaging. Adaptable metasurface control is essential for their applications in practical communications engineering. Here we present a cyber-managed metasurface system to enhance the convenience of metasurface sub-array management, which integrates radio frequency energy harvesting with star-topology hybrid networks. By employing digitized phase-shifted transmission lines as tunable elements, the system not only enables electromagnetic manipulation and sensing capabilities but also achieves ultra-low power consumption. Each metasurface sub-array consists of 2 × 2 units, serving as a network node for data transmission and the sharing of harvested energy. Additionally, these metasurface sub-arrays, designed to resemble LEGO blocks, can be combined into various configurations, enabling flexible electromagnetic manipulation. The cyber-managed metasurface can be seamlessly integrated into wireless communication systems and passive wireless sensing networks, thereby providing versatility across diverse applications.

Neutron Resonance Transmission Analysis (NRTA) is a highly sensitive, non-destructive technique for nuclear material characterisation, but its application has been limited by its reliance on large, fixed, and costly installations. Here, we present a compact mobile NRTA system utilising a small ²⁵²Cf spontaneous neutron source, designated as a prototype "table-top NRTA system", to analyse nuclear materials, offering a mobile and cost-effective alternative to accelerators or deuterium-tritium generators. The pilot system, measuring 130 cm × 50 cm × 50 cm with a 42 cm flight path, enables time-of-flight measurements on nuclear material samples. The system's performance was demonstrated through NRTA measurements of simulated samples, including indium, hafnium, and cadmium metal plates. The experimental transmission spectra were compared with theoretical predictions using the PHITS Monte Carlo simulation and the JENDL-5 nuclear data library, enabling isotope identification below 5 eV. The obtained results underscore the system's potential as a complementary tool for nuclear security and safeguards verification, particularly in scenarios where access to large accelerator or reactor facilities is impractical, and where mobility, compactness, and cost-effectiveness are prioritised over throughput.

One of the main concerns affecting the uptake of battery packs is safety, particularly with respect to fires caused by cell faults. Mitigating possible risks from faults requires advances in battery management systems and an understanding of the dynamics of large packs. To address this, a machine learning classifier based upon a support vector machine was developed that detects cell faults within large packs using a limited number of current sensors. To train the classifier, a modelling framework for parallel-connected packs is introduced and shown to generalise to Doyle-Fuller-Newman electrochemical models. The fault classification performance was found to be satisfactory, with an accuracy of 83% using current information from only 27% of the cells. Validation on experimental pack data is also shown. These results highlight the potential to combine mathematical modelling and machine learning to improve battery management systems and deal with the complexities of large packs.

Fuelbreak placement is an important consideration in fire management. Historically, strategies for placing fuelbreaks have fallen on the experience of fire managers such as by following ridgelines, and recent searches for a formal placement strategy have struggled to scale to large areas. Here we present a basic strategy utilizing equal graph partitioning and quantum computing to efficiently determine placements. By posing partitioning as a quadratic constrained binary optimization problem, D-Wave's hybrid quantum optimization tool could complete the task in seconds. Results for the examined area show two alternatives to the ridgeline method in a so-called worst-case fire scenario: one with 2.9% improvement in land separation equality while clearing 76 less acres, and another with a 12.4% improvement by clearing 19 more acres. In a selected subsection, D-Wave's hybrid solver performed faster than the SCIP solver but slower than the CPLEX solver, with the prospect for increased speed-up on larger problems. These findings demonstrate the effectiveness of equal graph partitioning for fuelbreak placement and the potential of D-Wave's hybrid solvers.

Accurately predicting long-term degradation in chaotic systems remains a fundamental challenge due to their sensitive dependence on initial conditions and non-periodic dynamics. Conventional numerical models, which rely on fine time-step integration, are computationally demanding and prone to cumulative errors. Here we present a phase-space random walk framework for degradation modeling in chaotic systems. The approach characterizes local degradation velocity distributions through short-time averaging and reconstructs the long-term evolution as stochastic transitions across phase-space regions. Validation on chaotic electronic and mechanical systems demonstrates that the method improves computational efficiency by over two orders of magnitude while maintaining prediction errors below five percent. The analysis further reveals that chaotic systems experience transitions among dynamic regimes with varying degrees of chaos during degradation. This framework provides an efficient and generalizable way to modeling complex degradation processes, offering a other insights into the reliability design of electronic, mechanical, and mechatronic systems.

Magnetic particle imaging (MPI) is a tomographic imaging technique which determines the spatial distribution of magnetic nanoparticles (MNPs). Multi-contrast MPI provides the ability to detect environmental conditions of MNPs, such as temperature or viscosity. One parameter that has not been investigated but shows high potential for medical diagnosis is the pH value, as it is an indicator of inflamed or tumorous tissue. In this work, we present an approach to resolve the pH value using multi-contrast MPI. Our proof-of-concept is based on a stimuli-responsive, magnetic hydrogel that exhibits reversible swelling in response to a pH change. The pH contrast is generated indirectly via the pH-responsive hydrogel swelling modulating the signal of embedded MNPs. Magnetic particle spectrometry measurements show that the hydrogels' magnetic response correlates with the pH value, which could provide a new way of contactless pH monitoring. Finally, the feasibility of resolving different pH values in a multi-contrast MPI image is demonstrated.

Computational image enhancement for microscopy facilitates cutting-edge biological discovery. While promising, the commonly used deep learning methods are computationally expensive owing to the use of general-purpose architectures, which are inefficient for microscopy data. Here, we propose a sparsity-efficient neural network for image enhancement as a deep representation learning solution to inverse problems in imaging. To maximize accessibility, we developed a framework named DeBCR, consisting of a modular Python library and a user-friendly point-and-click DeBCR plugin for Napari, a popular bioimage analysis tool. We provide a detailed protocol for using the DeBCR as a library and a plugin, including data preparation, training, and inference. We compare the image restoration performance of DeBCR to ten current state-of-the-art models over four publicly available datasets spanning crucial modalities in advanced light microscopy. DeBCR demonstrates more robust performance in denoising and deconvolution tasks across all assessed microscopy modalities while requiring notably fewer parameters than existing models.