Mobile Edge Computing (MEC) networks have attracted significant attention for enabling users to offload computation-intensive tasks to edge servers. Task offloading scheduling is a critical challenge, especially when complete information about tasks and edge servers is only partially accessible in practice. In general, each edge server can only obtain its own information but has no access to the complete real-time information of other edge servers, resulting in information incompleteness. To address this issue, this paper investigates the problem of Energy Minimization through Offloading with Incomplete Edge Information (EMO-IEI). Specifically, to address the uncertainty of real-time computing resources caused by incomplete edge information, we adopt the Exact Convex Regularization (ECR) method to estimate resource availability based on known expectations and variances. Utilizing these estimations, we reformulate the problem as a collapsing multi-knapsack problem and propose the GAP-ESM algorithm for efficient solution. Theoretical analysis validate that the GAP-ESM algorithm achieves an approximation ratio of (1 + κ/κ−ρκ−ρ ), where κ is system parameter associated with the energy requirements of computing tasks, and ρ is a tunable design parameter balancing approximation quality and computational complexity. Extensive simulations demonstrate that the proposed GAP-ESM algorithm outperforms baseline schemes in terms of overall energy consumption and task completion rate.

Collaborative edge computing enables multiple edge servers to cooperate by sharing information and computing resources, achieving resource complementarity and task collaborative processing. In areas with dense mobile users, edge servers are often deployed and managed by different business entities. Therefore, in a heterogeneous edge environment with multiple business entities, the collaborative relationship between edge servers is often constrained by their respective management strategies and profit considerations. Resource sharing is no longer unconditional, but instead involves a certain degree of competition and self-interested behavior. Existing works often assume unconditional collaboration, overlooking the rational and self-interested behavior of edge servers belonging to different business entities in practice. This paper proposes a coalition-based collaborative scheduling framework tailored for Directed Acyclic Graph (DAG) tasks with inter-task dependencies. It designs a coalition formation algorithm for edge servers and a multi-DAG task scheduling algorithm within the coalition to achieve optimal resource collaboration and edge server utility. Extensive experiments are conducted to evaluate the performance of the proposed algorithms. Experimental results show that the proposed algorithms consistently outperform all baselines across different network scales, achieving at least a 46.34% improvement in cooperative utility.

Camouflaged object detection (COD) relies on multi-granularity structural information and fine-grained details to distinguish objects from highly similar backgrounds. Whereas low-resolution data lacks high-frequency cues such as textures and sharp edges, retaining only coarse structures. These not only weaken discriminative features but also introduce resolution-induced camouflage beyond natural blending. Existing COD methods assume high-resolution data and fail to address this dual-source ambiguity, resulting in significant performance degradation and underscoring the need for approaches that explicitly explore essential spatial priors under low-resolution constraints. Therefore, we propose KRNet, the first framework explicitly designed for COD in low-resolution settings. KRNet presents a Leader-Follower framework where the Leader extracts dual gold-standard distributions: conditional and hybrid, from supporting data to drive the Follower in rectifying knowledge learned from low-resolution data. The framework further benefits from a cross-consistency strategy, and a stronger time-prompt conditional encoder that improve the rectification of these distributions. Extensive experiments on benchmark datasets demonstrate that KRNet outperforms state-of-the-art COD methods and SR-assisted COD approaches, highlighting its effectiveness in tackling the challenges of low-resolution data in COD. Code: https://github.com/whyandbecause/KRNet/tree/main.

Collaborative Edge Computing (CEC) enables effective load balancing by decomposing tasks across edge servers. However, due to limited computing and storage resources in CEC networks, eliminating computational redundancies becomes particularly important for improving overall efficiency and conserving resources. To address this, we propose a novel compute-storage cooperation framework that jointly optimizes task offloading and computation result caching to minimize system-wide delay and caching cost. The optimization problem is decomposed into two subproblems: reusable task scheduling and reusable data caching. Accordingly, the CoRe-S algorithm and the VaRe-C algorithm along with a proactive pre-caching mechanism are proposed to solve these subproblems, respectively. By leveraging temporal and spatial correlations among computational tasks, the proposed framework directly caches computation results to reduce redundant processing. In addition, the age of data is incorporated into the evaluation metric to better assess the value of cached results, thereby enhancing reuse efficiency. Theoretical analysis and extensive simulations are conducted to validate the effectiveness and superiority of the proposed algorithms. Compared with state-of-the-art baselines, our method reduces the total cost by up to 41.89% and achieves a cache hit rate of 53.1%.

Relation Extraction (RE) is a key task in table understanding, aiming to extract semantic relations between columns.However, complex tables with hierarchical headers are hard to obtain high-quality textual formats (e.g., Markdown) for input under practical scenarios like webpage screenshots and scanned documents, while table images are more accessible and intuitive.Besides, existing works overlook the need of mining relations among multiple columns rather than just the semantic relation between two specific columns in real-world practice.In this work, we explore utilizing Multimodal Large Language Models (MLLMs) to address RE in tables with complex structures.We creatively extend the concept of RE to include calculational relations, enabling multi-task learning of both semantic and calculational RE for mutual reinforcement.Specifically, we reconstruct table images into graph structure based on neighboring nodes to extract graph-level visual features.Such feature enhancement alleviates the insensitivity of MLLMs to the positional information within table images.We then propose a Chain-of-Thought distillation framework with self-correction mechanism to enhance MLLMs' reasoning capabilities without increasing parameter scale.Our method significantly outperforms most baselines on wide datasets.Additionally, we release a benchmark dataset for calculational RE in complex tables.

Successful video analysis relies on accurate recognition of pixels across frames, and frame reconstruction methods based on video correspondence learning are popular due to their efficiency. Existing frame reconstruction methods, while efficient, neglect the value of direct involvement of multiple reference frames for reconstruction and decision-making aspects, especially in complex situations such as occlusion or fast movement. In this paper, we introduce a Dynamic Memory Prediction (DMP) framework that innovatively utilizes multiple reference frames to concisely and directly enhance frame reconstruction. Its core component is a Reference Frame Memory Engine that dynamically selects frames based on object pixel features to improve tracking accuracy. In addition, a Bidirectional Target Prediction Network is built to utilize multiple reference frames to improve the robustness of the model. Through experiments, our algorithm outperforms the state-of-the-art self-supervised techniques on two fine-grained video object tracking tasks: object segmentation and keypoint tracking.

The Industrial Edge Computing (IEC) network has recently received considerable attention, where industrial devices offload their computation-intensive and delay-sensitive tasks to servers located at the network edge. Task offloading scheduling is a fundamental problem in IEC networks to achieve satisfactory quality of service. Many prior efforts have been devoted to scheduling task offloading for networks with complete information, while the complete information is hard or even infeasible to acquire by the scheduler. Therefore, their performance degrades in IEC networks with incomplete information. Scheduling task offloading for IEC networks with incomplete information is urgent and presents great technical challenges. This paper proposes a group-centric task offloading framework tailored for IEC networks with incomplete information, and models the minimum delay scheduling problem as a Partially Observable Markov Decision Process. Then, the SGOS algorithm integrating the Long Short-Term Memory with Soft Actor-Critic networks in reinforcement learning is proposed to devise online task offloading schedules for IEC networks with incomplete information. Extensive experimental results verify that the SGOS algorithm can achieve the best performance compared with base-line schemes in terms of major metrics, including convergence, delay, and workload balance.

Vision-Language Pre-training (VLP) models have achieved remarkable success in practice, while easily being misled by adversarial attack. Though harmful, adversarial attacks are valuable in revealing the blind-spots of VLP models and promoting their robustness. However, existing adversarial attacking studies pay insufficient attention to the key roles of different modality-correlated features, leading to unsatisfactory transferable attacking performance. To tackle this issue, we propose the Transferable MultiModal (TMM) attack framework, which tailors both the modality consistency and modality discrepancy features. To promote transferability, we propose the attention-directed feature perturbation to disturb the modality-consistency features in critical attention regions. In light of the commonly employed cross-attention can represent the consistent features among diverse models, it is more possible to mislead the similar model perception for activating stronger transferability. For improving attacking ability, we proposed the orthogonal-guided feature heterogenization to guide the adversarial perturbation to contain more modality-discrepancy features in the encoded embeddings. Since VLP models rely more on aligned features among different modalities during decision-making, increasing the modality-discrepant could confuse the learned representation for better attacking ability. Extensive experiments under diverse settings demonstrate that the proposed TMM outperforms the comparisons by large margins, i.e., 20.47% improvements in transferable attacking ability on average. Moreover, we highlight that our TMM also shows outstanding attacking performance on large models, such as MiniGPT-4, Otter, etc.

Camouflaged objects often blend in with their surroundings, making the perception of a camouflaged object a more complex procedure. However, most neural-network-based methods that simulate the visual information processing pathway of creatures only roughly define the general process, which deficiently reproduces the process of identifying camouflaged objects. How to make modeled neural networks perceive camouflaged objects as effectively as creatures is a significant topic that deserves further consideration. After meticulous analysis of biological visual information processing, we propose an end-to-end prudent and comprehensive neural network, termed IdeNet, to model the critical information processing. Specifically, IdeNet divides the entire perception process into five stages: information collection, information augmentation, information filtering, information localization, and information correction and object identification. In addition, we design tailored visual information processing mechanisms for each stage, including the information augmentation module (IAM), the information filtering module (IFM), the information localization module (ILM), and the information correction module (ICM), to model the critical visual information processing and establish the inextricable association of biological behavior and visual information processing. The extensive experiments show that IdeNet outperforms state-of-the-art methods in all benchmarks, demonstrating the effectiveness of the five-stage partitioning of visual information processing pathway and the tailored visual information processing mechanisms for camouflaged object detection. Our code is publicly available at: https://github.com/whyandbecause/IdeNet.

Matching tabular data to a knowledge graph (KG) is critical for understanding the semantic column types, column relationships, and entities of a table. Existing matching approaches rely heavily on core columns that represent primary subject entities on which other columns in the table depend. However, discovering these core columns before understanding the table’s semantics is challenging. Most prior works use heuristic rules, such as the leftmost column, to discover a single core column, while an insightful discovery of the core column set that accurately captures the dependencies between columns is often overlooked. To address these challenges, we introduce Dependency-aware Core Column Set Discovery ( DaCo ), an iterative method that uses a novel rough matching strategy to identify both inter-column dependencies and the core column set. Additionally, DaCo can be seamlessly integrated with pre-trained language models, as proposed in the optimization module. Unlike other methods, DaCo does not require labeled data or contextual information, making it suitable for real-world scenarios. In addition, it can identify multiple core columns within a table, which is common in real-world tables. We conduct experiments on six datasets, including five datasets with single core columns and one dataset with multiple core columns. Our experimental results show that DaCo outperforms existing core column set detection methods, further improving the effectiveness of table understanding tasks.

Traditional Chinese Medicine (TCM) is the crystallization of Chinese medical heritage for thousands of years and plays a huge role in human health. How to integrate TCM products with modern information intelligent production means to efficiently obtain standardized Chinese medicine services is a very worthy research direction. In this paper, the decocting service of TCM is taken as the research object, and combined with artificial intelligence, blockchain and other information technologies, an intelligent service system for trusted collaboration in the whole process of decocting TCM is innovatively proposed and designed. Under the intelligent collaboration of a variety of decocting devices, key links, such as automated prescription business management, intelligent dispensing and decocting, and blockchain-based quality monitoring, were studied, and collaborative decision-making optimization and information trusted traceability of the whole process of decocting Chinese medicine were realized. The service system was applied in a Chinese medicine piece processing company, realizing the ability to handle 880,000 prescriptions annually, 144 decocting stations with intelligent collaboration, and 35 indicators with full process traceability. The practice shows that the system can realize the trusted intelligent cooperation of the whole process of Chinese herbal decoction.

During the COVID-19 period, Lecture Recording System is of great significance to the remote and flexible education, and the key issue of the lecture recording automation is the detection of teacher and students. To achieve fast and accurate detection, We applied YOLO algorithm to the detection and adopted Eagleeye pruning method to reduce the amount of parameters and calculation of the YOLO model. We also designed and developed the Intelligent Lecture Recording System based on hisi Hi3531DV200. Experiments on the data set collected from Internet shows that the pruned model can achieve 94.8% precision and 31.25FPS speed on Hi3531DV200, which make the Intelligent Lecture Recording System fast, robust and accurate.

Mobile edge computing enables users to run resource-intensive applications at the network edge equipped with small server clusters. The mobile services are heterogeneous and edge servers are generally resource-limited. Only a subset of services can be processed by an edge server in a time slot. In this paper, we study the joint service placement and computation scheduling (JSPCS) problem to optimize both service quality and operation cost. We formulate the optimization problem to maximize the worst utility among all the services, under the constraints of multiple types of resources, and the JSPCS problem is proved to be NP-hard. By linear relaxation, we provide a dual decomposition approach to decouple this hard problem into a sequence of tractable sub-problems. We propose the Lagrange duality based joint optimal service placement and computation scheduling (LD-JSPCS) algorithm to derive the optimal solution to JSPCS problem with linear relaxation. By iteratively solving a series of feasibility problems, we prove the proposed algorithm guarantees the convergence to the optimum. Theoretical analysis and extensive simulations are performed, validating the efficiency of LD-JSPCS in service provision with limited resources.

Industrial Internet of Things (IIoT) has attracted increasing attention for improving the efficiency of manufacturing. Plenty of computation-intensive and latency-sensitive applications are required by IIoT networks, which pose significant challenges for the computation capacities of IIoT networks. To address these challenges, Edge-enabled Industrial Internet of Things (E-IIoT) emerges. Edge devices located at the edge of IIoT networks enlarge computation capacities of IIoT networks and improve their efficiency accordingly. How to schedule computation resources wisely is a major problem in E-IIoT networks. Since IIoT devices in an E-IIoT network monitor the industrial site collaboratively, tasks for processing sensory data collected by them are correlated accordingly. That means, scheduling highly correlated tasks to be processed at the same device can improve computation efficiency. Inspired by this fact, we propose a correlation aware scheduling (CAS) algorithm for E-IIoT networks in this article. In specific, computation model decision and processing order decision are made by considering computation resources of devices and correlations among tasks in the algorithm to minimize latency of E-IIoT networks. The NP-hardness of correlation aware latency minimization scheduling problem in E-IIoT networks is first proved. Theoretical analysis on approximation ratio of the CAS algorithm is provided, and simulation results demonstrate the effectiveness of the proposed algorithm in reducing latency.

Despite the salient feature of cloud computing, the cloud provider still suffers from electricity bill, which in part comes from 1) the power consumption of running physical machines (PMs) to guarantee the resource/time requirements of virtual machines (VMs), and 2) the dynamically varying electricity price offered by smart grids. In the literature, there exist viable solutions adaptive to electricity price variation to reduce the electricity bill. However, they are not applicable to serving time-sensitive VM requests. In serving time-sensitive VM requests, it is potential for the cloud provider to apply proper consolidation strategies to further reduce the electricity bill. Few prior works have provided theoretical solutions of VM consolidation strategies that are adaptive to electricity price variations in serving time-sensitive VM requests. In this work, to address this challenge, we develop electricity-price-aware consolidation algorithms for both the offline and online scenarios. For the offline scenario, we first develop a consolidation algorithm with constant approximation, which always approaches the optimal solution within a constant factor of 5. For the online scenario, we propose an$O(\log (\frac{L_{max}}{L_{min}}))$-competitive algorithm that is able to approach the optimal offline solution within a logarithmic factor, where$\frac{L_{max}}{L_{min}}$is the ratio of the longest length of the processing time requirement of VMs to the shortest one. Our trace-driven simulation results further demonstrate that the average performance of the proposed algorithms produce near-optimal electricity bill.

Transmission is crucial for big data analysis and learning in industrial Internet of Things (IoT) systems. To transmit data with limited energy is a challenge. This paper studies the problem of data transmission in energy harvesting systems with capacitor to supply energy where the energy receiving rate varies over time. The energy receiving rate is slower when the capacitor receives more energy. Based on this characteristic, we study the problem of how to transmit more data when the energy receiving time is not continuous. Given many packets that arrive at different time instances, there is a tradeoff between transmitting the packet right now or saving the energy to transmit the future arriving packets. We formalize two types of problems. The first one is how to minimize the total completion time when there is enough energy to transmit all the packets. The second one is how to transmit as many packets as possible when the energy is not enough to transmit all the packets. For the first problem, we give a 1 + α approximation off line algorithm when all the information of the packets and the energy receiving periods is known in advance, and a max{2, β} competitive ratio online algorithm where the information is not known in advance. For the second problem, we study three cases and give a 6 + [h/(b/R)] approximation off line algorithm for the general situation. We also prove that there does not exit a constant competitive ratio online algorithm.

In crowdsourcing markets, a requester announces a task and calls for contribution from potential participants. With strategic participants, the requester needs to reward the participants to introduce the incentives of participation. However, it is natural to ask whether it is worth introducing incentives if the total payment for eliciting incentives is too high. This paper addresses such a fundamental concern by designing a frugal mechanism with minimum payment used to procure the total amount of service contributions demanded. We design two mechanisms to provide the incentives of participation while minimizing the payment used by the requester. We first propose a frugal auction-based mechanism, which stimulates participants to truthfully report their information. We theoretically prove that the payment used is not more than the optimal cost (with no incentive considered) plus a bounded additive. We then design a Stackelberg-game-based mechanism, in which the requester fixes a certain total payment at the very beginning so as to encourage the participants to compete for it and participate in the task. We verify the existence of a unique Nash equilibrium (NE) and develop a novel algorithm to find the NE, as well as the optimal payment to extract the NE. Our simulation results show that the payment used in these mechanisms is close to the optimal solution with no incentive considered, while the extra payment caused by introducing truthfulness in auction-based mechanism is about twice that of the NE in Stakelberg-game-based mechanism.

Multicast is an important way to diffuse data in duty-cycled wireless sensor networks (WSNs), where nodes can receive data only in active state. The communication delay can be extremely large if inappropriate schedules are adopted. Unfortunately, most previous methods do not consider controlling multicast delay energy-efficiently. This paper studies the minimum active time slot augmentation for delay-bounded multicast (MAADM) problem in duty-cycled WSNs. The MAADM problem is proved to be NP-hard even under the node-exclusive interference model. An optimal algorithm is proposed for the MAADM problem when K = 2 and a heuristic latency bounding algorithm is proposed for source-to-all communications, where K denotes the number of the destination nodes. When K > 2, two (K-1)-approximation algorithms are designed for the MAADM problem. In addition, a low computation-complexity distributed algorithm is proposed. To the best of our knowledge, this is the first work that develops a series of efficient centralized and distributed algorithms for the MAADM problem in dutycycled WSNs. The theoretical analysis and experimental results verify that all the proposed algorithms have high performance in terms of delivery delay and energy consumption.

Myopia becomes a more and more serious worldwide problem as the number of myopic people (especially young people) grows rapidly. Efficient methods are required to monitoring the deterioration of nearsightness so as to take further treatment. This demo realizes a noval nearsightness monitoring system, called iSee, which utilizes the widely used smartphones to detect the deterioration of nearsightness by monitoring and analysing the the distance between the eyes and the smartphone screen. A prototype of iSee has been developed to evaluated the effectiveness under different environmental conditions.

Multi-radio WLAN mesh networks (WMNs) are intended to provide a wireless access infrastructure with high throughput and reliable transmission. With the increasing demand for ubiquitous Internet access, access networks tend to be large-scale, complex and dense, and energy consumption has become a critical concern. Since the networks are designed to support peak traffic demand but does not serve peak traffic demand all the time, which leads to significant energy wastage in off-peak conditions. This means that energy consumption can be effectively reduced through scheduling nodes on/off according to the current traffic demand. In multi-radio WMNs, each node is equipped with multiple radios, and radios can also be switched on/off for further energy saving. Therefore, we investigate the problem of joint node scheduling and radio switching for energy efficiency in this paper, which is proven to be NP-hard. We first formulate the problem as an integer linear programming model, which aims to minimize the power consumption of the network while satisfying node-radio-link, traffic demand and routing constraints. Then, we propose an efficient heuristic algorithm, which iteratively finds routing paths for all mesh access points (MAPs) while satisfying their traffic demand. In each iteration, exact one MAP is selected, and the nodes on its routing path are scheduled on and the corresponding radios are switched on. Finally, extensive simulations are conducted to evaluate the performance of the proposed algorithm in terms of the number of active nodes and radios. The results not only show that the algorithm can achieve energy saving efficiently and effectively, but also demonstrate that the joint optimization of node scheduling and radio switching can obtain better performance on energy efficiency.

Data sharing for data collection among multiple applications is an efficient way to reduce communication cost for Wireless Sensor Networks (WSNs). This paper is the first work to introduce the interval data sharing problem which is to investigate how to transmit as less data as possible over the network, and meanwhile the transmitted data satisfies the requirements of all the applications. Different from current studies where each application requires a single data sampling during each task, we study the problem where each application requires a continuous interval of data sampling in each task. The proposed problem is a nonlinear nonconvex optimization problem. In order to lower the high complexity for solving a nonlinear nonconvex optimization problem in resource restricted WSNs, a 2-factor approximation algorithm whose time complexity is$O(n^{2})$and memory complexity is$O(n)$is provided. A special instance of this problem is also analyzed. This special instance can be solved with a dynamic programming algorithm in polynomial time, which gives an optimal result in$O(n^{2})$time complexity and$O(n)$memory complexity. Three online algorithms are provided to process the continually coming tasks. Both the theoretical analysis and simulation results demonstrate the effectiveness of the proposed algorithms.

COUNT is a typical aggregation operation in Wireless Sensor Networks (WSNs). In such an operation, the total number of the items which are of the same kind is obtained and only one numerical value is returned as the result. This paper identifies the multiple count problem which counts items belonging to multiple categories. For each category, the total number of the items belonging to this category is calculated. Therefore, the returned result is a set of values instead of a single value. The multiple count problem is more challenging than the traditional count problem as the former incurs more communication overhead. This paper proposes a distributed approximate multiple count algorithm which can derive an error bounded result under a specified communication cost constraint for each node. The error of the derived result is hN/L, where h is the depth of the routing tree, N is the total number of all the items belonging to all the categories, and L is a representation of the communication cost constraint for each node. Furthermore, the weighted multiple count problem is investigated where different kinds of items to be counted have different weights. The proposed algorithms are evaluated through TOSSIM, a widely used simulation tool for WSNs. The theoretical analysis and simulation results both demonstrate the correctness and effectiveness of the proposed algorithms.

Data sharing for data collection among multiple applications is an efficient way to reduce the communication cost of Wireless Sensor Networks (WSNs). This paper is the first work to introduce the interval data sharing problem which is to investigate how to transmit as less data as possible over the network, and meanwhile the transmitted data satisfies the requirements of all the applications. Different from current studies where each application requires a single data sampling during each task, we study the problem where each application requires a continuous interval of data sampling in each task instead. The proposed problem is a nonlinear nonconvex optimization problem. In order to lower the high complexity for solving a nonlinear nonconvex optimization problem in resource restricted sensor nodes, a 2-factor approximation algorithm whose time complexity is O(n2) and memory complexity is O(n) is provided. A special instance of this problem is also analyzed. This special instance can be solved with a dynamic programming algorithm in polynomial time, which gives an optimal result in O(n2) time complexity and O(n) memory complexity. We evaluate the proposed algorithms with TOSSIM, a widely used simulation tool in WSNs. Theoretical analysis and simulation results both demonstrate the effectiveness of the proposed algorithms.

Data approximation is a popular means to support energy-efficient query processing in sensor networks. Conventional data approximation methods require users to specify fixed error bounds a prior to address the trade-off between result accuracy and energy efficiency of queries. We argue that this can be infeasible and inefficient when, as in many real-world scenarios, users are unable to determine in advance what error bounds can lead to affordable cost in query processing. We envision ε- approximate querying (EAQ) to bridge the gap. EAQ is a uniform data access scheme underlying various queries in sensor networks. It allows users or query executors to incrementally 'refine' previously obtained approximate data to reach arbitrary accuracy. EAQ not only grants more flexibility to in-network query processing, but also minimizes energy consumption through communicating data upto a just-sufficient level. To enable the EAQ scheme, we propose a novel data shuffling algorithm. The algorithm converts sensed datasets into special representations called multi-version array (MVA) . From prefixes of MVA , we can recover approximate versions of the entire dataset, where all individual data items have guaranteed error bounds. The EAQ scheme supports efficient and flexible processing of various queries including spatial window query, value range query, and queries with QoS constraints. The effectiveness and efficiency of the EAQ scheme are evaluated in a real sensor network testbed.