Abstract：To meet the demands of emerging services such as global seamless coverage, emergency communications, IoT, and aviation and maritime applications, 3GPP Release 19 has systematically enhanced NTN. Compared to the transparent forwarding-dominated architecture in Release 18, 3GPP R19 NR NTN introduced a novel onboard regeneration processing architecture for the first time, enabling flexible deployment of RAN and core network functions on satellites. This significantly reduces end-to-end latency and enhances system capacity. The 3GPP R19 NR NTN access network standard has been enhanced and expanded in multiple aspects compared to Release 18, making the 6G technical solution more comprehensive and further strengthening its capabilities, thereby providing a solid technical foundation for the commercialization of 5G NR NTN. This paper takes the challenges faced in the current development of 6G satellite Internet as the starting point and focuses on analyzing aspects such as the core architecture of R19 NR NTN, air interface and performance optimization, mobility and service continuity, terminal and scenario expansion, etc. Finally, the technological development of R20 was prospected.

Abstract：For the research on the expansion of marine area communication coverage, deploying small base stations(SBSs) inside ships reduces the penetration loss for cabin. However, the marine limits the wired backhaul for SBSs via internet protocol radio access network(IPRAN) or STN to transmit data back to the core network on land. Moreover, ordinary customer premise equipment(CPE) for backhaul of SBSs has low power and limited uplink transmission. Therefore, to solve marine communication problem, a scheme for expanding 5G coverage is proposed via introducing high-power CPE and high-gain servo antenna systems for heterogeneous network based on the macro BS(MBS) near coast and SBSs in marine areas. The high-power CPE provides backhaul for SBSs in cabin with wired connection. The servo antenna system is used to enhance uplink(UL) for the CPE to connect to the MBS. In particular, the list information about MBS is used for the wireless network searching process of multiple user equipments(UEs), and optimizes the target shore-based cell access algorithm for the servo antenna through collaboration of multiple UEs. The current network results show good performance for the scheme proposed.

Abstract：To address the limitations of existing amplitude phase shift keying (APSK) modulation schemes in satellite communications, specifically their reliance on commercial chips or FPGA implementations and performance evaluations confined to ideal Gaussian channels, this study proposes a dynamically reconfigurable verification framework based on an open-source software defined radio (SDR) platform. Aligned with the DVB-S2 standard, a joint simulation system for high-order modulation and satellite channels is constructed on the GNU Radio platform. This system supports flexible configuration of 16APSK and 32APSK modulation-demodulation links and deeply integrates Loo and Corazza models to accurately emulate multipath delay, shadow fading, and Doppler shift characteristics. End-to-end hardware-in-the-loop validation from baseband generation to RF transceiver operations and receiver demodulation is implemented via the HackRF SDR hardware platform. Adopting a modular design, the system enables dynamic switching of modulation orders and channel parameters, overcoming reconfiguration constraints inherent in closed-source systems. This framework provides a scalable testbed for evaluating spectral efficiency, fading resilience, and power amplifier nonlinearity adaptability of high-order modulation techniques in 6G integrated space-ground networks.

Abstract：In broadband Multiple Input Multiple Output (MIMO) communication systems, the use of a uniform circular array (UCA) at the base station introduces a significant near-field beam-splitting phenomenon due to the array's wide bandwidth and large-scale antenna characteristics. This renders traditional fast beam training methods based on phase shifter (PS) beamforming ineffective in broadband environments. This paper first demonstrates that the beam-splitting effect in the UCA architecture is controllable. Specifically, by introducing time delay (TD) beamforming, the extent of near-field beam splitting can be effectively adjusted. Building on this, a novel fast broadband beam training strategy is designed. The scheme employs TD beamforming to simultaneously generate beams at multiple frequencies, each focused on distinct locations in three-dimensional space. The training is conducted in two stages: the first stage searches for the user's azimuth information, and the second stage searches for the user's distance and elevation information. Simulation results show that the proposed scheme has a low training overhead of only 1 051, while its average achievable rate reaches 96.28% of the perfect channel state information benchmark.

Abstract：Aiming at the severe threat of unmanned aerial vehicle (UAV) “black flight” to the safe operation of substations, as well as the problems of low positioning accuracy, weak anti-interference ability and poor scene adaptability of traditional low-altitude protection technologies, this paper proposes a precise parameter estimation algorithm for low-altitude UAVs in power grids based on 5th Generation-Advanced (5G-A) integrated sensing and communication（ISAC）. First, a perception parameter system covering UAV motion characteristics and electromagnetic scattering characteristics is constructed, and a positioning environment reconstruction parameter system including interference signal power spectral density, multipath attenuation coefficient and signal occlusion rate is designed. Environmental adaptation is realized through the process of "parameter collection - model correction - error compensation". Finally, a two-stage continuous time-delay-Doppler channel parameter estimation algorithm of "2D orthogonal matching pursuit (2D-OMP) sparse coarse estimation + continuous domain golden section fine estimation" is proposed to solve the problems of grid mismatch and high complexity of traditional algorithms. The results show that the root mean square error (RMSE) of velocity estimation of the proposed algorithm remains optimal. Compared with the traditional single pilot estimation and one-step search algorithms, the accuracy is improved by about 55% and 33% respectively, which can achieve the target perception of decimeter-level positioning and low false alarm rate. The research results provide a high-precision and high-reliability technical scheme for low-altitude safety protection of substations, and also lay a foundation for the large-scale application of 5G-A technology in t

Abstract：In industrial automated production, metal surface defect detection faces challenges such as small target missed detection and insufficient positioning accuracy. To address these challenges, this paper proposes three optimizations for the YOLOv11 model. Firstly, a C3k2_MLCA module is constructed, which integrates a multi-scale local contextual attention mechanism to enhance the network's ability to capture subtle features. Secondly, the loss function is improved from CIoU to EIoU, which explicitly constrains the width-height differences of bounding boxes to improve regression accuracy. Finally, a Detect_LADH detection head is designed, which optimizes multi-scale defect recognition through a lightweight adaptive feature aggregation structure. Experiments on the self-constructed dataset show that the mAP@0.5 of the improved model is 8.9% higher than that of the original model, with the inference time remaining at 1.4 ms per picture. It provides a high-precision and high-robustness technical solution for industrial defect detection.

Abstract：To solve the issues of poor recognition performance of artificial intelligence models due to the scarcity of transformer fault samples and the high missed detection rate of small target faults in existing infrared diagnostic techniques, this paper presents a parallel operation-based infrared image anomaly recognition method for transformers. Firstly, the consistency analysis is conducted on transformers in parallel operation, and the "comparison-recognition" abnormal infrared images diagnostic strategy is developed. Next, the YOLO-based object detection model is built to segment five key transformer components. Finally, the region growing algorithm is employed to locate abnormal areas in infrared images. Experimental results show that the proposed method can detect online thermal faults in transformers under zero fault sample conditions, and effectively addresses the difficulty of detecting small target faults, providing a novel and feasible scheme for transformer thermal fault detection.

Abstract：For multi-chiplet heterogeneous complex structures, traditional functional testing methods are unable to accurately locate internal interconnection faults. A boundary-scan-based interconnect testing method for multi-chiplet is proposed, which enables the detection of interconnect faults between chiplets. Based on the IEEE 1149.1 boundary-scan protocol and an optimized interconnect testing algorithm, this method uses an Automatic Test Equipment (ATE) testing system to identify faulty internal interconnection lines, thereby accurately detecting chip faults and defects. Compared with traditional testing methods, the boundary-scan-based interconnect testing method for multi-chiplet is stable and reliable. It can accurately pinpoint internal interconnection faults within the chip, significantly improves testing efficiency, and ensures the reliable application of chiplet systems.

Abstract：This paper presents a command-driven 16-bit successive approximation register analog-to-digital converter (SAR ADC) designed to enable precise real-time monitoring of target devices. The ADC performs conversion only upon receiving commands from the host, allowing accurate data acquisition and fault detection. The proposed ADC consists of three main components: the first-stage ADC, the second-stage ADC, and the residue amplifier. A one-time on-chip calibration technique is employed to mitigate the capacitor mismatch. An incremental and binary hybrid capacitor array is implemented in the first-stage capacitive digital-to-analog converter (CDAC) design. After calibration, measurement results show that the ADC achieves an SNDR of 82 dB and an SNR of 84 dB. The measured differential nonlinearity (DNL) and integral nonlinearity (INL) are -0.9/+1.5 LSB and -5.5/+4 LSB, respectively. The circuit is fabricated in a 180 nm CMOS process, occupies an active area of 2.9 mm², operates at a supply voltage of 5 V, and consumes 6 mW of power.

Abstract：In order to meet the complex requirements of electronic equipment such as miniaturization, high performance, rapid development and information processing capabilities, a big data cleaning integrated information processing SiP circuit based on advanced System-in-Package (SiP) technology was designed. The circuit adopts a domestically produced bare core, realizes the system integration of multiple micro components through 3D Through-Silicon Via (3D TSV) and modular multiplexing technology, supports the adaptive interference suppression function of multiple airspace directions at the same time, supports the interference suppression ability of large-scale unmanned swarm, and meets the real-time computing capability requirements of big data cleaning. The circuit is 90% smaller and 85% lighter than the board. Through multi-physics simulation and practical testing, the circuit meets the development requirements and has application prospects in the field of UAV swarms.

Abstract：At present, most of the analog circuits of integrated circuits are verified in the form of behavior models during SoC (System on Chip) verification. However, as the complexity and scale of circuits in integrated circuits in the world become higher and larger, there are more and more analog modules integrated by SoCs or MCUs, which are becoming more and more complex, and the complexity of digital-analog verification has increased dramatically. Based on the above problems, we are committed to developing a platform implementation method for Digital Mixed-Signal (DMS) verification based on SoC, so as to solve the correctness of the functions, timing, low-power functions, and low-power indicators of chip-level digital-analog interaction that can only be verified by waiting for the return test, and improve the completeness of verification.

Abstract：In response to the technical limitation of traditional in-ear active noise cancellation headphones, which only rely on the reference microphone to collect the main noise source for noise reduction while ignoring the impact of interfering noise on the error microphone, resulting in a significant reduction in system noise reduction performance when dealing with complex working conditions of mixed high and low-frequency noise, this paper proposes an optimized solution based on the dual-channel FxLMS algorithm. By constructing an error separation subsystem at the error microphone and combining it with the feedback FxLMS algorithm, the interference noise is suppressed. Theoretical analysis and experiments show that in the 250~500 Hz mixed-frequency interference noise scenario, the system's active noise reduction performance reaches 38 dB. When the noise spectrum expands to 165~600 Hz, the noise reduction is 19 dB. For high-frequency narrowband interference noise, the noise reduction is 30 dB. The improved algorithm significantly enhances the robustness of active noise control in complex noise environments through a multi-dimensional noise separation and compensation mechanism.

Abstract：The phenomenon of ice coating on transmission lines is a significant issue in power system operation, where accurate estimation of ice weight holds critical importance for ensuring the safety and stability of transmission lines. This paper proposes an image stitching-based mass estimation method for transmission lines. Addressing the insufficient edge preservation capability of existing deblurring algorithms under complex motion blur scenarios, this paper employs improved total variation regularization to remove blur from poorly captured images, thereby enhancing image details and sharpness. Subsequently, the SIFT algorithm is utilized to extract and match features from the enhanced images, combined with a weighted fusion algorithm to achieve image stitching. Finally, the maximum sag of conductors is calculated based on the proportion of conductors in the stitched image. Using the parabolic model of transmission lines and relevant conductor parameters, this paper analyzes the correspondence relationship between ice weight and sag to derive ice weight quantification. The combined strategy of total variation regularization and SIFT dynamic matching proposed in this paper maintains 95.8% matching accuracy even under blurred and low-light conditions, outperforming mainstream algorithms. This approach provides power system engineers with reliable ice weight estimation references, effectively enhancing the anti-icing and disaster mitigation capabilities of transmission lines.

Abstract：As one of the key technologies in remote sensing-based 3D reconstruction, point cloud and image fusion faces challenges in traditional approaches, such as high computational cost and difficulty in balancing efficiency and data precision. This paper proposes an efficient point cloud color fusion strategy that integrates image pyramids with bilinear interpolation, while introducing OpenMP for parallel optimization. The method leverages the pyramid structure to enable rapid multi-resolution localization and adaptive regional interpolation, effectively improving fusion efficiency while preserving fine details. Experiments conducted on two datasets demonstrate that the proposed approach maintains high fusion accuracy while achieving a 3.6~3.8 fold improvement in processing speed. The performance gain is especially notable for high-density point cloud data. This method not only provides high-precision data support for 3D modeling of power transmission corridors but also offers important technical support for automated inspection.

Abstract：The initial productivity of oil reservoir is affected by many factors such as geology, engineering and development, which is a complex nonlinear time series. The traditional prediction method using a single model has low prediction accuracy and weak data adaptability. A prediction model of initial reservoir productivity based on sparse auto encoder (SAE) and ensemble learning is proposed. Firstly, SAE is used to analyze various factors affecting the initial production, and the five-dimensional characteristics of oil saturation, effective thickness of perforated interval, fracturing sand addition, sand addition intensity and energy retention state are automatically extracted as the main control factors. Then, the prediction models are established by using linear regression (LR), support vector regression (SVR) and long short-term memory neural network (LSTM) to predict the initial production. Finally, Bagging ensemble learning is used to comprehensively integrate the prediction results of the three methods, so as to obtain the final prediction results of the initial production of the reservoir. Using the actual data of an ultra-low permeability oilfield to carry out the verification test, the results show that compared with the single LR, SVR and LSTM methods, the proposed method has higher prediction accuracy, stronger data adaptability and higher application prospect.

Abstract：This paper introduces a high-speed data acquisition and playback system for UAVs. Built on the Xilinx Zynq-7035 platform with PS and PL co-design, the system enables real-time acquisition, storage, and playback of radio signals across 70 MHz to 6 GHz. It offers precise geolocation to synchronize sampling coordinates for interference localization and synthetic aperture applications. High-speed UDP transmission over Gigabit Ethernet reaches 900 Mbps. The lightweight design suits UAV deployment for airborne spectrum monitoring and jammer localization. The Qt-based host software provides UDP reception, real-time spectrum visualization, parameter configuration, and controllable playback, delivering an integrated "acquisition–processing–storage–playback" solution.

Abstract：In response to the challenges of exponential growth in software complexity and frequent vulnerability in vehicle electronic control units (ECUs), coupled with the inefficiencies and cost challenges associated with traditional physical disassembly-dependent flashing methods utilizing proprietary protocols, this study proposes a standardized CAN bus based Bootloader solution which is implemented on the NXP S32K144 microcontroller with adherence to UDS protocol specifications (ISO 14229-1 and ISO 15765-2). The developed framework incorporates critical UDS application layer services and transport layer packet segmentation mechanisms to establish a comprehensive three-phase firmware update process encompassing pre-programming, programming, and post-programming phases. Through rigorous Flash partitioning architecture, hardware-based protection mechanisms, and cyclic redundancy check (CRC) verification protocols, the system ensures robust boot reliability and secure update operations. Based on this, a self-developed Delphi-based client tool was designed and extensive Bootloader programming testing was conducted. Experimental results demonstrate the solution's capacity for efficient and reliable ECU software updates, exhibiting exceptional robustness and reliable recovery capabilities even under adverse conditions including power interruptions and CAN communication failures. This implementation establishes a standardized and highly reliable software update and upgrade framework for automotive critical ECUs, providing fundamental technical bases for future ECU software development, optimization, calibration, and upgrade. The proposed methodology also offers a comprehensive engineering reference for software upgrade development of similar ECUs.

Abstract：In view of the high precision requirements for energy storage battery cell voltage sampling and battery cluster voltage, current, and insulation detection, this paper presented a liquid cooled battery management system based on ADBMS6830, which uses ADI's advanced AFE ADBMS6830 and TI's precision ADC ADS1120 to achieve accurate measurement of battery voltage information, cluster voltage, cluster current, and system insulation impedance detection. The test results show that the battery voltage acquisition can achieve measurement errors within±2 mV within the full temperature range of standard voltage 3.3 V, cluster voltage acquisition can achieve testing accuracy of±0.06%, cluster current acquisition can achieve testing accuracy of ±0.06%, and insulation detection can achieve testing accuracy of ±5%. The proposed liquid cooled battery management system has been applied in a series 5 MWh liquid cooled energy storage system, verifying the feasibility of the system.

Abstract：Targeting the demanding requirements of aerospace and industrial control fields, including high computing real-time performance, massive data processing, high-speed communication, and complex task execution, this paper implements cross-platform Symmetric Multi-Processing (SMP) multicore porting of SpaceOS, a real-time operating system for spacecraft computers, using programmable fusion chips. First, the FMQL45 platform is selected as the hardware/software foundation for porting SpaceOS, enabling core functionalities such as the Board Support Package (BSP) and task scheduling. Subsequently, an SMP multicore solution is designed for the FMQL45 platform, incorporating multicore memory partitioning, stack configuration, interrupt handling, and task management. Cross-core synchronization and communication are achieved through inter-core interrupts (IPI) and shared memory. Finally, a representative SMP multicore engineering project is developed, establishing a debugging framework with the IAR Embedded Workbench IDE and I-JET debugger. Functional and performance tests validate the system’s multicore processing capabilities.

Abstract：In infrared and visible image fusion, conventional convolutional networks typically rely on fixed convolution kernels, which restrict their ability to adaptively extract complementary cross-modal features and simultaneously model local textures and global semantic relationships. To address these limitations, we propose a fusion framework that integrates adaptive convolution with a multi-scale dynamic Transformer. Adaptive convolution enhances spatial alignment and interaction between heterogeneous modal features, while the dynamic Transformer further strengthens long-range dependency modeling and mitigates local detail degradation. In addition, the multi-scale architecture enables comprehensive extraction of critical information across hierarchical feature spaces. The decoder reconstructs the fused image in an end-to-end manner, optimized by a triplet loss comprising pixel consistency, gradient preservation, and structural maintenance. Extensive experiments conducted on the TNO and RoadScene datasets demonstrate that the proposed method achieves superior performance in visual quality, information retention, and detail enhancement compared with six representative approaches. Moreover, it provides a good balance between inference efficiency and performance stability, effectively preserving salient targets in infrared imagery while maintaining fine textures from visible images.

Abstract：With the development of intelligent transportation, the convenience and safety of rail driving operation are getting lots of attention. In traditional rail driving, drivers need to frequently manually operate various physical controls to regulate on-board equipment, which is not only inefficient but also distracting and increases safety risks. To improve this situation, this paper proposes a on-board auxiliary control system based on speech recognition technology. This system allows drivers to conveniently control on-board equipment via voice commands, significantly improving human-machine interaction efficiency. The experimental results show that this system achieves high recognition accuracy and real-time response capabilities in high-noise environments, effectively reducing the operational burden of drivers. This research provides a new technical approach for improving intelligence level and operational safety of rail driving.

Abstract：Discrete manufacturing enterprises face multiple challenges in advancing digital transformation, including prominent data silos, inadequate data governance systems, inconsistencies between system data and operational scheduling mechanisms, insufficient application collaboration, and unclear implementation pathways. This paper first systematically analyzes the advantages and limitations of existing mainstream technical solutions such as data middle platform, industrial internet platform, microservice architecture, and enterprise service bus, pointing out their common problems in addressing the needs of small and medium-sized discrete manufacturing enterprises, including data-business separation, insufficient data reliability assurance, cost-benefit contradictions, and lack of unified scheduling mechanisms. To address these issues, this paper proposes a new paradigm for digital transformation, which mainly includes the following innovative measures: Using data standards as the breakthrough point, establish a comprehensive data governance system that covers all aspects, encompasses all data, and operates in real time across manufacturing enterprises; Centered around the corporate brain, build a radial digital architecture supported by large models; Using the data foundation as a platform, establish a unified scheduling mechanism for multi-source heterogeneous system data and business operations; Adopting the principles of small, fast, lightweight, and precise, to build a micro-unit collaborative application cluster model based on low-code development; Using transparent factories as the implementation vehicle, to explore new pathways for enterprise digital transformation and future factory construction. The new paradigm effectively addresses the limitations of existing solu