Network Technologies (United States)

ABSTRACT Through the lens of Nancy Fraser’s theory of social justice, this research critically examines whether and to what extent localisation, as practised, has realised its inherent social justice objectives. Using Start Network as a case study, it interrogates how localisation redistributes power, enhances accountability to at-risk and affected populations, and recognises the value of local knowledge, ultimately contributing to a more equitable and socially just humanitarian system. Drawing on a review of the literature, the paper demonstrates that for localisation to fulfil its social justice aims: (1) greater respect for and utilisation of indigenous and local knowledge is needed to challenge the dominance of Western epistemologies in humanitarian practice; (2) redistribution of power must be furthered by actively addressing structural inequalities in decision-making and resource allocation, supporting the decolonisation of aid; and (3) accountability mechanisms should be strengthened through meaningful and sustained community-level engagement.

The deployment of autonomous delivery vehicles (ADVs) is undergoing unprecedented expansion in China, fueled by escalating investments from online retail and logistics enterprises. Despite the multifacted benefits of ADVs, i.e. enhanced delivery efficiency, cost savings, and environmental sustainability, academic inquiry into consumes' intention to utilize such innovative technologies remains relatively scarce. To address this research gap, the study enhances and extends the Normative Activation Model (NAM) by integrating two critical constructs: perceived behavioral control (PBC) and perceived risk (PR). Using a structural equation model (SEM), we empirically analyze survey data collected from 354 online consumers via Wenjuanxing, a reputable online sampling platform. The results reveal that personal norms (PN) and PBC significantly affect consumers' intentions to adopt ADVs, while the impact of PR is statistically non-significant. Furthermore, ascription of responsibility (AR) partially mediates in the relationships between awareness of consequences (AC) and PN, and PN serves as a partial mediator in the relationships linking AC, AR, PBC and behavioral intention. These findings provide valuable theoretical and practical insights for logistics service providers - guiding the optimization of ADVs service design and risk mitigation strategies - and relevant government departments - informing the formulation of supportive policies for ADVs deployment and consumer protection regulations.

We experimentally demonstrate a GMI-estimated 254.7-Tb/s S+C+L band transmission applying DWDM over 200-km SSMF. NN-based equalizers applying transfer learning (TL) are deployed parallelly and the RX-DSP of each WDM channel is optimized. Covering 19.8-THz spectrum, the system SE reaches 12.86-bit/s/Hz. ©2025 The Author(s)

Large language models (LLMs) have demonstrated powerful capability to solve practical problems through complex step-by-step reasoning. Specifically designed LLMs have begun to be integrated into terrestrial communication networks. However, relevant research in the field of satellite communications remains exceedingly rare. To address this gap, we introduce SCNOC-Agentic, a novel architecture especially designed to integrate the management and control of satellite communication systems in LLMs. SCNOC-Agentic incorporates four components tailored to the characteristics of satellite communications: intent refinement, multi-agent workflow, personalized long-term memory, and graph-based retrieval. Furthermore, we define four typical real-world scenarios that can be effectively addressed by integrating with LLMs: network task planning, carrier and cell optimization, fault analysis of satellites, and satellite management and control. Utilizing the SCNOC-Agentic framework, a series of open-source LLMs have achieved outstanding performance on the four tasks under various baselines, including zero-shot CoT, CoT-5, and self-consistency. For example, qwen2.5-70B with SCNOC-Agentic significantly improves the parameter generation accuracy in the network task planning task from 15.6% to 32.2%, while llama-3.3-70B increases from 16.2% to 29.0%. In addition, ablation studies were conducted to validate the importance of each proposed component within the SCNOC-Agentic framework.

Recently, the attention of a research community has been attracted by the task of network traffic prediction, which reveals a potential to improve network control. However, the problem is often addressed using advanced (mainly machine learning) algorithms, which are computationally expensive and contribute significantly the greenhouse gas emissions. Since the reduction of emissions is a crucial issue according to the European Commission’s strategy, in this paper, we revisit the problem of network traffic prediction by considering not only the forecasting accuracy and processing time but also the following carbon footprint. We show that the selection of a prediction method can successfully compromise all these criteria.

Multiple-Input Multiple-Output (MIMO) technology has become a critical component in Beyond 5G (B5G) wireless communication systems, enhancing spectral efficiency and system capacity. However, accurate channel estimation remains a significant challenge, especially in high-mobility environments. In this study, conventional and deep learning-based channel estimation approaches were analyzed for 2×2, 4×4, and 8×8 Multi User MIMO, (MU-MIMO) systems. Least Squares (LS) method is compared with the proposed deep learning (DL)-based model, and the analyses are conducted under both perfect channel state information (P-CSI) and imperfect CSI (I-CSI) scenarios. The results demonstrate that the proposed DL-based approach achieves performance close to the P-CSI case and significantly improves estimation accuracy compared to traditional methods.

This study presents a transformer-based power prediction model for Raman-amplified wavelength division multiplexing (WDM) systems. Leveraging the transformer's self-attention mechanism, the model effectively captures global channel interactions, achieving a maximum prediction error of only 0.122 dB in a 200 km bidirectional Raman system with 47 channels and 12 pumps. The differential evolution (DE) technique was employed to optimize the hyperparameters of the models, further enhancing their performance. The comparative analysis was conducted to evaluate the performance of the two models under their respective optimal configurations. The model demonstrates robust generalization under varying pump powers and surpasses convolutional neural network (CNN)-based approaches in accuracy and interpretability.

We report the first real-time co-frequency co-time full-duplex optical switching network node using 400G and 800G mixed commercial coherent optical modules with ultra-high net capacity of 2.276Pbit/s for intra data center switching network application.