Jose Fernan Martinez

Structured data embodies vast reservoirs of actionable knowledge, yet non-technical users often lack the tools to effectively access or leverage these resources. This paper presents a novel middleware system architecture that harnesses Large Language Models (LLMs) to translate natural language queries (NLQs) into executable SQL statements, enabling seamless interaction with relational databases without requiring technical expertise. To bridge a critical gap in existing research, the proposed system integrates a robust security framework featuring multi-factor authentication and fine-grained, role-based access control. A comprehensive evaluation was conducted using over 813,000 real spatio-temporal IoT samples from the H2020-AFarCloud international project, comparing both general-purpose LLMs and models fine-tuned for NLQ-to-SQL translation. Two complementary evaluation metrics—Execution Accuracy (EX) and Valid Efficiency Score (VES)—were employed to assess syntactic correctness, semantic fidelity, and runtime efficiency. Experimental results show that GPT-o3 mini-high achieves the highest overall performance in terms of average VES, while DeepSeek R1 exhibits the most consistent behavior across heterogeneous query scenarios, as reflected by lower performance variability. These findings underscore the interplay between model capability and computational efficiency, demonstrating the potential of LLM-driven middleware to democratize structured data access and enhance data-driven decision-making.

The Internet of Things (IoT) landscape is hindered by a critical challenge: the lack of semantic interoperability among diverse data models. Existing IoT solutions often function as isolated data silos, impeding the seamless integration of heterogeneous data sources crucial for informed decision-making and streamlined processes. This research addresses this issue by introducing a pioneering solution: the Semantic Interoperability Support System (SISS). SISS is an innovative tool designed to bridge the semantic divide between disparate data models within a common application domain. To address the lack of interoperability between current IoT platforms, devices, and solutions that use native data models, SISS facilitates integration by enabling the generation of gateways or translator components. These components establish mappings between the semantic properties of source and target data models, leveraging advanced semantic analysis and inference techniques. The core principle underpinning SISS is its ability to discern and map the semantic content of data models. Through a meticulous analysis of the temporal and spatial dimensions inherent in the data, SISS establishes meaningful connections. This innovative approach fosters interoperability and enables a deeper understanding of the underlying information, enhancing the potential for data-driven insights. This paper delves into the pervasive issue of semantic interoperability in the current IoT paradigm and presents SISS as a transformative solution. By emphasizing its ability to transcend the limitations of existing solutions and its methodology to generate mappings between disparate data models, this research contributes to the achievement of global semantic interoperability in IoT.

Even the artificial intelligence (AI) has been widely used and significantly changed our life, deploying the large AI models on resource limited edge devices directly is not appropriate. Thus, the model split inference is proposed to improve the performance of edge intelligence (EI), in which the AI model is divided into different sub-models and the resource-intensive sub-model is offloaded to edge server wirelessly for reducing resource requirements and inference latency. Unfortunately, with the sharp increasing of edge devices, the shortage of spectrum resource in edge network becomes seriously in recent years, which limits the performance improvement of EI. Refer to the NOMA-based edge computing (EC), integrating non-orthogonal multiple access (NOMA) technology with split inference in EI is attractive. However, the NOMA-based communication aspect and the influence of intermediate data transmission fail to be considered properly in model split inference of EI in previous works, and the sophistication in resource allocation caused by NOMA scheme makes it further complicated. Thus, the Effective Communication and Computing resource allocation algorithm is proposed in this paper for accelerating the split inference in NOMA-based EI, shorted as ECC. Specifically, the ECC takes the energy consumption and the inference latency into account to find the optimal model split strategy and resource allocation strategy (subchannel, transmission power, computing resource). Since the minimum inference delay and energy consumption cannot be satisfied simultaneously, the gradient descent (GD) based algorithm is adopted to find the optimal tradeoff between them. Moreover, the loop iteration GD approach (Li-GD) is developed to reduce the complexity of the GD algorithm caused by parameter discretization. The key idea of Li-GD is that: the initial value of the i th layer’s GD procedure is selected from the optimal results of the former ( i - 1) layers’ GD procedure whose intermediate data size is the closest to i th layer. Additionally, the properties of the proposed algorithms are investigated, including convergence, complexity, and approximation error. The experimental results demonstrate that the performance of ECC is much better than that of the previous studies.

Sea cucumber detection represents an important step in underwater environmental perception, which is an indispensable part of the intelligent subsea fishing system. However, water turbidity decreases the clarity of underwater images, presenting a challenge to vision-based underwater target detection. Therefore, accurate, real-time, and lightweight detection models are required. First of all, the development of subsea target detection is summarized in this present work. Object detection methods based on deep learning including YOLOv5 and DETR, which are, respectively, examples of one-stage and anchor-free object detection approaches, have been increasingly applied in underwater detection scenarios. Based on the state-of-the-art underwater sea cucumber detection methods and aiming to provide a reference for practical subsea identification, adjacent and overlapping sea cucumber detection based on YOLOv5 and DETR are investigated and compared in detail. For each approach, the detection experiment is carried out on the derived dataset, which consists of a wide variety of sea cucumber sample images. Experiments demonstrate that YOLOv5 surpasses DETR in low computing consumption and high precision, particularly in the detection of small and dense features. Nevertheless, DETR exhibits rapid development and holds promising prospects in underwater object detection applications, owing to its relatively simple architecture and ingenious attention mechanism.

Robust and fast image recognition and matching is an important task in the underwater domain. The primary focus of this work is on extracting subsea features with sonar sensor for further Autonomous Underwater Vehicle navigation, such as the robotic localization and landmark mapping applications. With the assistance of high-resolution underwater features in the Side Scan Sonar (SSS) images, an efficient feature detector and descriptor, Speeded Up Robust Feature, is employed to seabed sonar image fusion task. In order to solve the nonlinear intensity difference problem in SSS images, the main novelty of this work is the proposed Underwater Wireless Sensor Network-based Delaunay Triangulation (UWSN-DT) algorithm for improving the performances of sonar map fusion accuracy with low computational complexity, in which the wireless nodes are considered as underwater feature points, since nodes could provide sufficiently useful information for the underwater map fusion, such as the location. In the simulated experiments, it shows that the presented UWSN-DT approach works efficiently and robustly, especially for the subsea environments where there are few distinguishable feature points.

Sea cucumber detection represents a significate step in underwater environmental perception, which is an indispensable part of the intelligent subsea fishing system. However, various complex factors such as water turbidity declines the clarity of underwater images, presenting a challenge to vision-based underwater target detection. Therefore, accurate, real-time and lightweight detection models are required. First of all, the development of subsea target detection is summarized in this presented work. Besides, since target detection methods based on deep learning such as YOLOv5 and DETR, which are respectively examples of one-stage and anchor-free deep leaning methods, have been increasingly applied in underwater detection scenarios. Based on the analysis of state-of-the-art underwater sea cucumber detection approaches and aiming to provide a reference for practical subsea identification, the sea cucumber detection based on the YOLOv5 and DETR are investigated and compared in detail. For each approach, the detection experiment is carried out on the derived dataset which contains a wide variety of sea cucumber sample images. The compared experiments demonstrate that the overall outperformance of YOLOv5 in terms of low computing consumption and high precision, particularly in detection of small and dense features. Nevertheless, the DETR exhibits rapid development and holds promising prospects in underwater object detection applications, owing to its relatively simple architecture and ingenious attention mechanism.