Recent advances in network technologies and Multi Agent Reinforcement Learning (MARL) have accelerated the development of Autonomous Underwater Vehicle (AUV) cluster networks, enabling intelligent applications such as target tracking and cooperative target encirclement. However, existing MARL models are typically designed for single-task scenarios, limiting their scalability in real-world multi-task environments. To ad dress this, we propose a lightweight MARL framework capable of handling multiple AUV tasks with reduced reliance on underwater sampling. Specifically, a unified state space representation is constructed to support task generalization, while a hybrid online offline MARL training paradigm is introduced by leveraging the logical reasoning and sample generation capabilities of Large Language Models (LLMs). This reduces the demand for real-time data collection. Furthermore, a supervised pretraining strategy is incorporated to improve convergence and learning stability. Based on these components, we develop the Large Language Model-driven Hybrid online-offline MARL algorithm towards Multi-Task scenarios (LLM-HMT), which supports intelligent deployment of multi-task AUV cluster systems with minimal state representation, reduced sample requirements, and limited training iterations. Extensive experiments demonstrate that LLM HMT outperforms mainstream MARL baselines in convergence speed, task success rate, and resource efficiency, highlighting its potential for practical underwater applications.

Currently, the Internet of Underwater Things (IoUT) plays an important role in ocean exploration, monitoring, and protection. However, it faces many security threats due to resource constraints, such as denial-of-service attacks. To overcome these challenges, a novel underwater network architecture that incorporates cloud computing and edge computing technologies, namely, the ocean-oriented cloud edge collaboration networks (O-OCECNs), is designed. O-OCECNs, while enhancing computational capabilities, still faces network threats, such as energy depletion attacks, data pollution attacks, etc. On this basis, a trust management mechanism called ETrust is proposed in this article, which uses ensemble learning to ensure network security. In the ETrust mechanism, nodes collect trust evidences by monitoring the behavior and communication results of other nodes and then deliver the evidences to the cluster head node. The cluster head node uploads the collected evidences to the edge server in its region to complete the trust value computation. Then, the edge server uploads the trust values to the cloud server, which utilizes the built-in learning to complete the trust evaluation. Finally, the cloud server outputs the trust evaluation value to the cluster head node and it adjusts the network topology. The experimental result demonstrates that the proposed scheme can detect malicious nodes with a trust evaluation accuracy of up to 98.55%. It also performs well in capturing selective forwarding attacks occurring in the network. Furthermore, the mechanism enhances network lifetime more effectively compared to existing schemes. Experimental result shows that our scheme maintains the average residual energy of devices at approximately 63%, compared to 50% with other methods.

Coverage is a fundamental issue in the Solar Insecticidal Lamps IoTs (SIL-IoTs). Compared to complete coverage, partial coverage emerges as the preferred strategy for deploying SILs within a limited budget, as this deployment solution offers the highest cost-effectiveness. In this paper, we concentrate on studying the constrained SILs deployment problem, taking into account partial coverage and energy harvesting requirements, which we refer to as the cSILDP-PCEH problem. In this context, the positions for deploying SILs are restricted to a weighted set of candidate locations on the ridges. The weight assigned to each candidate location reflects the energy harvesting potential of the SIL deployed at that position. Our objective is to deploy a group of SILs in a subset of these candidate locations, ensuring a high overall energy harvesting potential, network connectivity, and achieving partial coverage. Due to the NP-hard nature of the problem, we introduce an approximation algorithm with a provable performance ratio tailored to our problem. Finally, we conduct a theoretical analysis of our proposed algorithm and perform extensive simulations. The simulation results demonstrate that the proposed algorithm achieves a minimum improvement of 16.45% in energy harvesting potential while preserving network connectivity and maintaining a comparable coverage level.

Federated Learning is a promising technique that facilitates cloud–edge–terminal collaboration in Artificial Intelligence of Things (AIoT). It will enable model training without centralizing data, addressing privacy, and security concerns. However, when applied to AIoT, this technique faces several challenges, such as low communication efficiency among terminal devices, edges, and cloud platforms. In this article, we propose a novel approach called asynchronous federated learning with quantization (QuAsyncFL), which combines asynchronous federated learning with an unbiased nonuniform quantizer to address the issue of low communication efficiency. Moreover, we provide a detailed theoretical analysis of convergence with quantized gradients proving that the model could converge to a certain bound. Our experiments demonstrate that QuAsyncFL outperforms the original approach, achieving significant improvements in terms of communication efficiency. The research results represent a further step toward developing cloud–edge–terminal collaboration enabled AIoT.

This paper presents a low overhead multimedia data compression method for Solar Insecticidal Lamps Internet of Things (SIL-IoTs), achieving efficient audio data transmission. First, the audio and video data generated by working solar insecticidal lamps are collected to form an original dataset in the solar insecticidal lamps applications. Then, we propose SILGAN, a generative adversarial network to compress multimedia audio data in the SIL-IoTs. Specifically, a depth-wise separable convolution is adopted to reduce the computational resources for training networks and running programs. Moreover, the network parameters are optimized so that obtaining a better neural network model with stable operation on the nodes of the SIL-IoTs. Finally, experimental evaluation is conducted, showing the effectiveness of the proposed SILGAN approach.

Solar insecticidal lamps (SILs) Internet of Things is an emerging and environmentally friendly technology for preventing and controlling agricultural pests. As the disk coverage model only provides a coarse approximation of the sensing area in reality, the probabilistic coverage model (PCM) is appropriate for the deployment of SILs. However, most of the current studies on coverage problem under PCM have focused on point$\epsilon $-probability coverage, whereas a few referred to the area coverage problem since it is extremely difficult to verify the coverage of a complete continuous area under PCM, especially for irregular shaped area. In this article, we study how to deploy the minimum number of SILs with PCM to provide complete area$\epsilon $-probability coverage for actual farmland with irregular shape, where the locations used to deploy SILs are a limited set of candidates located on field ridges. We first formulate the complete area$\epsilon $-probability coverage problem into the minimum point$\epsilon $-probability coverage problem and prove that it is NP-complete. After that, we present an approximation algorithm with provable approximation rations to our problem. Finally, we analyze the performance of the proposed algorithm theoretically and perform extensive simulations to demonstrate its effectiveness.

Solar insecticidal lamps Internet of things (SIL-IoTs) has a long-term application trend, because it makes agricultural pest control more environmentally friendly and intelligent. However, the increase of lamps deployed has resulted in the challenge about maintenance burden. In this paper, we therefore provide a constrained SIL Deployment Problem under Coverage and Maintenance Cost considerations, referred to as cSILDP-CMC, where the positions used to deploy SIL nodes are a limited set of weighted Candidate Locations (CL) located on the ridges. A novel method is proposed to quantify the maintenance cost of each CL based on their comprehensive weight of coverage and maintenance cost considerations. Then we formulate the cSILDP-CMC and propose an Iterative Deployment Method (IDM) to solve the defined optimization problem. Finally, the experimental results show that our proposal equips better performance in terms of deployment cost, total comprehensive weight and coverage uniformity compared with the other four peer algorithms.

Wireless Sensor Networks (WSNs) have been widely used in agricultural productions. As a critical problems in WSNs, node deployment has a direct impact on the effectiveness of routing and data fusion operations as well as on the accuracy of anticipated coverage in several agricultural scenarios, e.g., mono-crop and mixed-crop farmlands. Network simulations are necessary for testing the effectiveness of deployment algorithms, but some of them lack the accuracy of real-world deployments. This is due to the fact that analogue maps in these network simulations do not take into account the nature of the terrain, for example obstacles such as buildings and trees in the line of vision for sensors, uneven surfaces and elevations for hilly terrains, the node locations obtained by the deployment algorithms in these analogue maps cannot be mapped to the actual farmland. In this paper, we present a methodology of constructing analogue map called AnaMap for providing both simulation and visualization of the actual farmland with the characteristic of partition structure and irregular boundary to assist the investigation of node deployment algorithms in agricultural environment. One case study is described to prove the usability of AnaMap.

Solar insecticidal lamps (SILs) play a vital role in green prevention and control of pests. By embedding SILs in wireless sensor networks (WSNs), we establish a novel agricultural Internet of Things (IoT), referred to as the SIL-IoTs. In practice, the deployment of SIL nodes is determined by the geographical characteristics of an actual farmland, the constraints on the locations of SIL nodes, and the radio-wave propagation in a complex agricultural environment. In this article, we mainly focus on the constrained SIL deployment problem (cSILDP) in a mixed-crop farmland, where the locations used to deploy SIL nodes are a limited set of candidates located on the ridges. We formulate the cSILDP in this Scenario as a connected set cover (CSC) problem and propose a hole-aware node deployment method (HANDM) based on the greedy algorithm to solve the constrained optimization problem. The HANDM is a two-phase method. In the first phase, a novel deployment strategy is utilized to guarantee only a single coverage hole in each iteration, based on which a set of suboptimal locations is found for the deployment of SIL nodes. In the second phase, according to the operations of deletion and fusion, the optimal locations are obtained to meet the requirements on complete coverage and connectivity. Experimental results show that our proposed method achieves better performance than the peer algorithms, specifically in terms of deployment cost.

As an important physical control technology, solar insecticidal lamp (SIL) can effectively prevent and control the occurrence of pests. The combination of SILs and wireless sensor networks (WSNs) initiates a novel agricultural Internet of Things (IoT), i.e., SIL-IoTs, to simultaneously kill pests and transmit pest information. In this article, we study the weighted SIL deployment problem (wSILDP) in SIL-IoTs, where weighted locations on ridges are prespecified and some of them are selected to deploy SILs. Different from the existing studies whose optimization objective is to minimize the deployment cost, we consider the deployment cost and the total weight of selected locations jointly. We formulate the wSILDP as the weighted set cover (WSC) problem and propose a layered deployment method based on greedy algorithm (LDMGA) to solve the defined optimization problem. The LDMGA is composed of two phases. First, SILs are deployed layer by layer from the boundary to the center until the entire farmland is completely covered. Second, on the basis of three design operations, i.e., substitution, deletion and fusion, the suboptimal locations obtained in the first phase are fine-tuned to achieve the minimum deployment cost together with the maximum total weight for meeting the coverage and connectivity requirements. Simulation results clearly demonstrate that the proposed method outperforms three peer algorithms in terms of deployment cost and total weight.

Solar insecticidal lamp (SIL) is a green prevention and control technology for pests. With the development of wireless sensor networks (WSNs), the combination of SILs and WSNs forms a novel agricultural Internet of Things-SIL Internet of Things (SIL-IoTs). However, the complex geographical characteristic of actual farmland has a great impact on SIL deployment. In this article, we study the SIL deployment problem (SILDP) with characteristics of full coverage, penetrable obstacles, irregular boundary, and partition structure. According to the partition structure caused by natural physiognomy feature, the actual farmland is divided into many subareas by ridges, and each subarea can be considered as a separate partition. Then, we formulate the SILDP in the scenario with the partition structure as the quadratic assignment problem. After that, we propose two deployment methods based on the genetic algorithm to address the SILDP. These two methods are the same in optimization objectives, but different in deployment sequence. The experimental results show that the proposed deployment methods equips better performance in terms of deployment cost compared with the other six peer algorithms.