Autonomous underwater vehicles (AUVs) have become a promising solution for data collection in underwater acoustic sensor networks (UASNs), and deep reinforcement learning (DRL) has been widely applied to enhance collection performance. However, our preliminary experiments indicate that existing DRL-based studies still face two critical challenges: 1) Collection blind spots. The sparse collection rewards and the requirement for energy-efficient trajectory planning during data collection jointly restrict AUVs' ability to explore and collect data from all sensor nodes (SNs), ultimately resulting in some SNs remaining uncollected. 2) Collection collisions. Simultaneous data collection by multiple AUVs can lead to packet collisions and collection failures, further decreasing the collection rate. To address these challenges, we propose ahierarchical DRL-basedcollision-freedatacollection scheme (HCDC). Specifically, we leverage a hierarchical DRL framework to decompose the multi-AUV-assisted data collection (MADC) problem into a high-level global target selection (GTS) and a low-level local trajectory planning (LTP) subproblems. For GTS, we design a multi-agent GTS (MA-GTS) algorithm to assign the next target SN for collection to each AUV. The MA-GTS incorporates both global and local rewards to collaboratively optimize the overall energy consumption while avoiding individual penalties. Based on the assigned target SN, a deep deterministic policy gradient-based LTP (DDPG-LTP) algorithm is proposed to conduct AUV trajectory planning, utilizing intrinsic rewards to enhance learning efficiency and eliminate collection blind spots. Furthermore, to avoid packet collisions, we analyze the conditions for collision-free data collection and propose an adaptive back-off slot (ABS) algorithm to schedule AUVs' collection slots. With the collision-free slots, DDPG-LTP dynamically adjusts AUVs' velocities to ensure collision-free collection while reducing energy consumption. Extensive simulation results demonstrate that HCDC can achieve better collection rate and energy efficiency than state-of-the-art schemes.

Autonomous Underwater Vehicles (AUVs) provide a new prospect for data collection in underwater wireless sensor networks (UWSNs). For dynamic underwater environments, researchers typically apply deep reinforcement learning (DRL) to design multi-AUV collection schemes for UWSNs. However, these methods suffer from the following issues. 1)Overloaded observations. The importance of various observations for an AUV varies over time. Considering all observations equally complicates decision-making for subsequent actions. 2)Dynamic scale of AUVs and sensors. Once the number of AUVs or sensors is changed, traditional static neural networks require retraining, lacking scalability across diverse scenarios. To solve the above issues, we propose a Generalizable Attention-based Data collection scheme (GAMD) for Multi-AUV UWSNs, while enhancing AUVs’ collection efficiency. GAMD incorporates the attention mechanism with multi-agent DRL framework, which enables AUVs to prioritize observations more critical for action decisions. Moreover, we propose an adaptive information processing approach, enabling the AUV policy model to seamlessly adapt to various scenarios without retraining. Additionally, we develop a training paradigm with incremental complexity across different scale scenarios to simplify training process and accelerate convergence. Simulation results demonstrate that GAMD alleviates the training cost compared to the state-of-the-art methods, and simultaneously optimizes collection energy efficiency, collection time, and trajectory distance.

Large Language Models (LLMs), constrained by limited context windows, often face significant performance degradation when reasoning over long contexts.To address this, Retrieval-Augmented Generation (RAG) retrieves and reasons over chunks but frequently sacrifices logical coherence due to its reliance on similarity-based rankings.Similarly, divideand-conquer frameworks (DCF) split documents into small chunks for independent reasoning and aggregation.While effective for local reasoning, DCF struggles to capture longrange dependencies and risks inducing conflicts by processing chunks in isolation.To overcome these limitations, we propose ToM, a novel Tree-oriented MapReduce framework for long-context reasoning.ToM leverages the inherent hierarchical structure of long documents (e.g., main headings and subheadings) by constructing a DocTree through hierarchical semantic parsing and performing bottom-up aggregation.Using a Tree MapReduce approach, ToM enables recursive reasoning: in the Map step, rationales are generated at child nodes; in the Reduce step, these rationales are aggregated across sibling nodes to resolve conflicts or reach consensus at parent nodes.Experimental results on 70B+ LLMs show that ToM significantly outperforms existing divide-andconquer frameworks and retrieval-augmented generation methods, achieving better logical coherence and long-context reasoning.Our code is available at https://github.com/gjn12- 31/ToM.

Simulators are essential to troubleshoot and optimize Underwater Acoustic Network (UAN) schemes (network protocols and communication technologies) before real field experiments. However, due to programming differences between the above two contents, most existing simulators concentrate on one while weakening the other, leading to non-generic simulations and biased performance results. Moreover, novel UAN schemes increasingly integrate Artificial Intelligence (AI) techniques, yet existing simulators lack support for necessary AI frameworks, failing to train and evaluate these intelligent methods. On the other hand, these novel schemes consider more UAN characteristics involving more complex parameter configurations, which also challenge simulators in flexibility and fineness. To keep abreast of advances in UANs, we propose the Fourth Generation (FG) network simulator-3 (ns-3)-based simulator Aqua-Sim FG, enhancing the general and intelligent simulation ability. On the basis of retaining previous generations’ functions, we design a new general architecture, which is compatible with various programming languages, including MATLAB, C++, and Python. In this way, Aqua-Sim FG provides a general environment to simulate communication technologies, network protocols, and AI models simultaneously. In addition, we expand six new features from node and communication levels by considering the latest UAN methods’ requirements, which enhances the simulation flexibility and fineness of Aqua-Sim FG. Experimental results show that Aqua-Sim FG can simulate UANs’ performance realistically, reflect intelligent methods’ problems in real-ocean scenarios, and provide more effective troubleshooting and optimization for actual UANs. The basic simulator is available at https://github.com/JLU-smartocean/aqua-sim-fg.

Quantitatively reconstructing the 3D structure of individual leaves within tree canopies is critical for understanding forest function and environmental responses to climate change. While quantitative structure models (QSMs) using terrestrial laser scanning (TLS) effectively capture woody structures, they lack the capability to accurately reconstruct non-woody leaf components. This study proposes AdLeaf (Accurate and Detailed Leaf), a novel approach for fine-scale reconstruction of individual leaves using TLS point clouds. AdLeaf combines wood-leaf separation, individual leaf segmentation, detection and repair of incomplete leaves, explicit reconstruction, and parameter extraction. It automates semantic segmentation at the tree scale to separate woody and leafy components. Instance segmentation is refined through similarity graphs. Incomplete leaves are detected and repaired using shape concavity analysis and symmetry-based mirroring. AdLeaf enables direct measurement of leaf attributes, including count, area, inclination, volume, and azimuth. Validation using field scans, synthetic data, and both in-situ and destructive measurements shows high accuracy: leaf counting errors ranged from 0.58% to 8.23% for trees with 201-4,000 leaves. Reconstructed leaf geometries had mean and standard deviations below 0.83 cm and 0.70 cm, respectively. Leaf area measurements (10–180 cm2) achieved a coefficient of determination (R²) of 0.95, bias of -0.20 cm², and root mean square error of 5.63 cm2. Incomplete leaf detection errors were below 28%, with the repaired area relative RMSE reduced by 9.4%. By addressing QSM limitations, AdLeaf enables explicit 3D leaf reconstructions that support detailed analysis of canopy light interception, spatial heterogeneity, and photosynthesis. It provides a robust framework for linking leaf structure to function at the tree level, advancing forest structure and radiative transfer research.

Autonomous Underwater Vehicle (AUV)-assisted Underwater Acoustic Networks (UANs) are promising for complex ocean applications. In essence, an AUV-assisted UAN is still dominated by fixed nodes, and Time Division Multiple Access (TDMA)-based Medium Access Control (MAC) protocols have undisputed practicability in such fixed nodes-dominated UANs since they are simple and easy to deploy. However, AUV-assisted UANs may exist dynamic bidirectional data streams, while most existing protocols assume UANs have a unidirectional data stream, and their fixed scheduling sequence results in the long end-to-end delay in AUV-assisted UANs. In this paper, we first reveal a phenomenon between the data stream and the scheduling sequence, derived from real-world experiments: their consistent direction decreases the packet waiting delay but increases the slot length, and vice versa. To optimize the end-to-end delay, UANs with dynamic bidirectional data streams expect the MAC protocol to provide a flexible scheduling sequence. To this end, we propose a low-delay Adaptive Scheduling MAC protocol (AS-MAC) based on TDMA for AUV-assisted UANs. In AS-MAC, we analyze the relationship between scheduling sequence and data stream, extracting two significant factors: slot length and packet delay. Afterwards, we design Slot Length Model (SLM) and Packet Delay Model (PDM) to analyze the end-to-end delay of different data streams. Based on these two models, we present a Scheduling Sequence and Slot Length allocation Algorithm (SSSLA) to adaptively provide the minimum end-to-end delay for current bidirectional data streams. Extensive simulation results show that AS-MAC efficiently addresses severe queue congestion of the state-of-the-art protocols and reduces the end-to-end delay of different dynamic streams in various scenarios.

Underwater acoustic sensor networks (UASNs) drive toward strong environmental adaptability, intelligence, and multifunctionality. However, due to unique UASN characteristics, such as long propagation delay, dynamic channel quality, and high attenuation, existing studies present untimeliness, inefficiency, and inflexibility in real practice. Digital twin (DT) technology is promising for UASNs to break the above bottlenecks by providing high-fidelity status prediction and exploring optimal schemes. In this article, we propose a Digital Twin-based Network Architecture (DTNA), enhancing UASNs’ environmental adaptability, intelligence, and multifunctionality. By extracting real UASN information from local (node) and global (network) levels, we first design a layered architecture to improve the DT replica fidelity and UASN control flexibility. In local DT, we develop a resource allocation paradigm (RAPD), which rapidly perceives performance variations and iteratively optimizes allocation schemes to improve real-time environmental adaptability of resource allocation algorithms. In global DT, we aggregate decentralized local DT data and propose a collaborative Multi-agent reinforcement learning framework (CMFD) and a task-oriented network slicing (TNSD). CMFD patches scarce real data and provides extensive DT data to accelerate AI model training. TNSD unifies heterogeneous tasks’ demand extraction and efficiently provides comprehensive network status, improving the flexibility of multi-task scheduling algorithms. Finally, practical and simulation experiments verify the high fidelity of DT. Compared with the original UASN architecture, experiment results demonstrate that DTNA can: (i) improve the timeliness and robustness of resource allocation; (ii) greatly reduce the training time of AI algorithms; (iii) more rapidly obtain network status for multi-task scheduling at a low cost.

Medium Access Control (MAC) protocols are critical for scheduling resources to access multiple users without collisions in Underwater Acoustic Networks (UANs). Due to harsh marine environments and limited communication resources, UANs lack a standard MAC protocol to adapt to various scenarios. The specific UAN scenario suffers from how to analyze multiple basic MAC protocols’ performance boundaries and modify the most potential one. A practical solution is to evaluate MAC protocols’ performance by modeling data loss (collisions and packet errors) and service time. However, existing models provide inaccurate performance results, since they ignore the effects of unique UANs’ characteristics and MAC protocol diversity on data loss and service time. In this paper, we propose a MAC Performance Analysis Framework (MPAF) for UANs to consider both unique UANs’ characteristics and MAC protocols’ diversity. We design Successful Transmission Probability (STP) model and Packet Service Time (PST) model in MPAF to estimate nodal throughput, delay, and energy consumption. STP model analyzes data loss types of different MAC protocols by considering long propagation delay, half-duplex communication, and random backoff to achieve a superior STP result from a view of real underwater communication conditions. Based on STP model, we employ Markov chain to deduce the retransmission number in PST model. In this way, MPAF ensures effectiveness and applicability in real-ocean environments. Extensive simulation results show that MPAF can accurately evaluate different MAC protocols’ performance boundaries, select the most appropriate basic protocol, and provide modified suggestions for a specific UAN scenario.

Mobile Underwater Acoustic Networks (UANs) leverage Autonomous Underwater Vehicles (AUVs) to enhance flexibility and mobility, playing an essential role in ocean research. Similar to static UANs, the Medium Access Control (MAC) protocol is still critical for mobile UANs to achieve efficient communication. However, mobile UANs are delaysensitive and suffer from low Signal-to-Noise Ratio (SNR), which presents significant challenges for the design of MAC protocols. As a hybrid technology combining spread spectrum and multicarrier modulation, Multi-Carrier Code Division Multiple Access (MC-CDMA) offers simple multi-path channel equalization and flexible multi-user access, aiding the MAC protocol in achieving robust communication in mobile UANs. Along this line, we propose an MC-CDMA-based MAC (MC-MAC) protocol, which considers both characteristics of mobile UANs and MC-CDMA to achieve efficient concurrent communication. Specifically, to adequately utilize the limited underwater communication resources, we design an adaptive node clustering algorithm, classifying nodes based on propagation distance, relative mobile velocity, data size, and data grade. Meanwhile, the algorithm determines non-random initial center nodes and adaptively decides the optimal number of clusters to decrease the computational complexity. Based on the clustering results, we present a manyobjective optimization algorithm, which jointly allocates specific spreading code length, spreading code number, subcarrier range, and transmission power to optimize throughput, delay, and energy consumption in mobile UANs. Extensive simulation results demonstrate that MC-MAC fully leverages the advantages of MC-CDMA, providing efficient concurrent communication with lower energy consumption for mobile UANs compared to stateof-the-art protocols.

Performing a high-capacity Medium Access Control (MAC) protocol suffers from low bandwidth and long propagation delay in Underwater Acoustic Networks (UANs). Non-Orthogonal Multiple Access (NOMA) is a promising technology to assist MAC protocols in overcoming the above restrictions and improving UANs’ capacity. It enables multiple users to access the same frequency-time resource based on power or code differences of user classes. However, UANs lack MAC research employing NOMA’s physical advantages. Most existing NOMA-based MAC protocols are designed for terrestrial networks, which are inapplicable to UANs. In classifying users, they ignore the effects of harsh marine environments on acoustic channels and fail to decrease channels’ interference, resulting in conflicting communications. Moreover, such unreasonable classification results further affect resource allocation, leading to low transmission rate and high energy consumption in UANs. In this paper, we propose a Hybrid NOMA-based MAC protocol (HN-MAC) to achieve efficient concurrent communication for UANs. Specifically, HN-MAC combines power-domain and code-domain NOMA to classify users and allocate communication resources. For the user classification, we propose an Adaptive Clustering Algorithm (ACA), which dynamically determines the clusters’ number and classifies users based on channel gain and channel correlation under multipath conditions. In this way, HN-MAC decreases interference among multiple users in various ocean scenarios. During the resource allocation, we formulate a joint allocation problem of transmission power and codebook based on the clustering result to optimize transmission rate and energy consumption. Further, a genetic algorithm is proposed to solve the allocation problem by considering resource constraints. Simulation results show that HN-MAC provides more stable concurrent communications with less resource consumption than the state-of-the-art protocols in various UANs.

Performing an effective media access control (MAC) protocol suffers from strong dependencies between Underwater Acoustic Networks’ upper and lower layers: 1) the network layer frequently uses geo-routing protocols, which do not provide the specific next-hop for MAC protocols, resulting in serious data collisions and 2) in such scenarios with the unknown next-hop, fixed orthogonal frequency-division multiplexing (OFDM) resource does not adapt to the changing environment, and degrades network performance (OFDM is a mature modulation technology in the physical layer). However, there is scant research on MAC protocols considering the network layer and the physical layer simultaneously, to solve data collisions and resource allocation. To this end, we present a cross-layer MAC protocol to integrate Geo-routing protocols and OFDM technology (GO-MAC) at the same time. GO-MAC employs a handshake scheme to allocate optimal communication resources and select the next-hop concurrently. First, we formulate the OFDM resource allocation as a joint optimization problem based on transmission mode, subcarrier spacing, guard interval, and transmission power, to decrease transmission delay and energy consumption. Then, a Karush–Kuhn–Tucker conditions-based Heuristic algorithm (KKT-H) is proposed to solve this problem. Finally, we consider node congestion and channel quality to assist geo-routing protocols with the next-hop selection, and decrease packet collisions. Simulation results show that our protocol matches geo-routing protocols and OFDM technology better than the state-of-the-art protocols, providing higher end-to-end reliability with lower costs.

YouTube videos are one of the most effective platforms for disseminating creative material and ideas, and they appeal to a diverse audience. Along with adults and older children, young children are avid consumers of YouTube materials. Children often lack means to evaluate if a given content is appropriate for their age, and parents have very limited options to enforce content restrictions on YouTube. Young children can thus become exposed to inappropriate content, such as violent, scary or disturbing videos on YouTube. Previous studies demonstrated that YouTube videos can be classified into appropriate or inappropriate for young viewers using video metadata, such as video thumbnails, title, comments, etc. Metadata-based approaches achieve high accuracy, but still have significant misclassifications, due to the reliability of input features. In this paper, we propose a fusion model, called Samba, which uses both metadata and video subtitles for content classification. Using subtitles in the model helps better infer the true nature of a video improving classification accuracy. On a large-scale, comprehensive dataset of 70K videos, we show that Samba achieves 95% accuracy, outperforming other state-of-the-art classifiers by at least 7%. We also publicly release our dataset.

Underwater acoustic sensor networks (UASNs) have been widely applied in marine monitoring, military reconnaissance, hydrology surveys, etc. Their location information is an important apriori knowledge when they are carried out underwater. However, the complex underwater environments impose great challenges on location acquisition, especially for autonomous underwater vehicles (AUVs), because of their mobility and finite power. In UASNs, existing location and navigation methods can offer AUVs position information, but they may either need a doppler velocity log (DVL), which is inefficient due to the complex underwater environments, or they may require additional localization infrastructure to deploy underwater, which suffers from large communication latency among AUVs, and costs enormous power. In this article, an efficient velocity estimation and location prediction method (VELP) in UASNs is proposed to avoid the above restrictions. It only utilizes collaborations based on communication among AUVs to achieve higher precision location with lower cost. Specifically, we apply an AUV-assisted velocity estimation algorithm with Doppler shift estimation in the physical layer of UASNs to improve the velocity estimation accuracy instead of the DVL. Meanwhile, we build a belief propagation-neural network-based location prediction model, which decreases the communication requirements and obviates introducing modeling errors. Extensive experimental results show VELP achieves superior performance on both accuracy and efficiency, demonstrating its great advantage in offering AUVs’ location information.

For an autonomous underwater vehicle (AUV), navigation is a key functionality. Dead-reckoning (DR) navigation is an important class among all the AUV navigation methods. In DR, the measurement errors of inertial sensors (such as gyroscopes and accelerometers) lead to accumulated errors with time, which affect navigation accuracy significantly. Especially, accumulated errors in fast-changing environments, such as waves near or on the surface, are tough to handle. In this article, we propose a neural-network-based AUV navigation method for fast-changing environments, called NN-DR. NN-DR employs the neural network to predict pitch angles accurately, which is our core contribution. In NN-DR, we smoothly integrate the Kalman filter, neural network, and velocity compensation to reduce accumulated errors. Extensive simulation experiments are conducted to test the correctness and stability of NN-DR, and the results show that NN-DR is very effective in lowering accumulated errors. For instance, at time 300 s, NN-DR achieves superior performance on accuracy for navigation, about 160 times than the state-of-the-art DR methods, demonstrating great advantage on AUV navigation for fast-changing environments.

With the advent of powerful network processors (NPs) in the market, many computation-intensive tasks such as routing table look-up, classification, IPSec, and multimedia transcoding can now be accomplished more easily in a router. An NP consists of a number of on-chip processors to carry out packet level parallel processing operations. Ensuring good load balancing among the processors increases throughput. However, such multiprocessing also gives rise to increased out-of-order departure of processed packets. In this paper, we first propose an Ordered Round Robin (ORR) scheme to schedule packets in a heterogeneous network processor assuming that the workload is perfectly divisible. The processed loads from the processors are ordered perfectly. We analyze the throughput and derive expressions for the batch size, scheduling time and maximum number of schedulable processors. To effectively schedule variable length packets in an NP, we propose a Packetized Ordered Round Robin (P-ORR) scheme by applying a combination of deficit round robin (DRR) and surplus round robin (SRR) schemes. We extend the algorithm to handle multiple flows based on a fair scheduling of flows depending on their reservations. Extensive sensitivity results are provided through analysis and simulation to show that the proposed algorithms satisfy both the load balancing and in-order requirements for parallel packet processing.

We consider a cluster-based multimedia Web server that dynamically generates video units to satisfy the bit rate and bandwidth requirements of a variety of clients. The media server partitions the job into several tasks and schedules them on the backend computing nodes for processing. For stream-based applications, the main design criteria of the scheduling are to minimize the total processing time and maintain the order of media units for each outgoing stream. In this paper, we first design, implement, and evaluate three scheduling algorithms, First Fit (FF), Stream-based Mapping (SM), and Adaptive Load Sharing (ALS), for multimedia transcoding in a cluster environment. We determined that it is necessary to predict the CPU load for each multimedia task and schedule them accordingly due to the variability of the individual jobs/tasks. We, therefore, propose an online prediction algorithm that can dynamically predict the processing time per individual task (media unit). We then propose two new load scheduling algorithms, namely, Prediction-based Least Load First (P-LLF) and Prediction-based Adaptive Partitioning (P-AP), which can use prediction to improve the performance. The performance of the system is evaluated in terms of system throughput, out-of-order rate of outgoing media streams, and load balancing overhead through real measurements using a cluster of computers. The performance of the new load balancing algorithms is compared with all other load balancing schemes to show that P-AP greatly reduces the delay jitter and achieves high throughput for a variety of workloads in a heterogeneous cluster. It strikes a good balance between the throughput and output order of the processed media units.

Several companies have introduced powerful network processors (NPs) that can be placed in routers to execute various tasks in the network. These tasks can range from IP level table lookup algorithm to application level multimedia transcoding applications. An NP consists of a number of on-chip processors to carry out packet level parallel processing operations. Ensuring good load balancing among the processors increases throughput. However, such multiprocessing also gives rise to increased out-of-order departure of processed packets. In this paper, we first propose a dynamic batch co-scheduling (DBCS) scheme to schedule packets in a heterogeneous network processor assuming that the workload is perfectly divisible. The processed loads from the processors are ordered perfectly. We analyze the throughput and derive expressions for the batch size, scheduling time and maximum number of schedulable processors. To effectively schedule variable length packets in an NP, we propose a packetized dynamic batch-coscheduling (P-DBCS) scheme by applying a combination of deficit round robin (DRR) and surplus round robin (SRR) schemes. We extend the algorithm to handle multiple flows based on a fair scheduling of flows depending on their reservations. Extensive sensitivity results are provided through analysis and simulation to show that the proposed algorithms satisfy both the load balancing and in-order requirements in packet processing.

Several companies have introduced powerful network processors (NP) that can be placed in active routers to execute application level tasks in the network. An NP consists of a number of on-chip processors to carry out packet level parallel processing operations. We propose to employ them for multimedia streaming (transcoding) to convert the incoming video streams to low bit-rate media units as per the requirements of the clients. To effectively schedule the parallel transcoding operations in an active router, we propose a static sequentialized batch-coscheduling (SSBC) scheme to meet both load balancing and real-time requirements for media streaming, based on divisible load theory (DLT). We first analyze the feasibility and optimality of the load distribution schemes from the theoretical perspectives, and then present separate solutions for non-delay-sensitive streams and delay-sensitive streams. Rigorous simulations and experiments have been carried out to evaluate the performance.