Federated Learning (FL) offers a pioneering distributed learning paradigm that enables devices/clients to build a shared global model that can be obtained through frequent model transmissions between clients and a central server, causing high latency, energy consumption, and congestion over backhaul links. To overcome these drawbacks, Hierarchical Federated Learning (HFL) has emerged, which organizes clients into multiple clusters and utilizes edge nodes (e.g., edge servers) for intermediate model aggregations between clients and the central server. Current research on HFL mainly focus on enhancing model accuracy, latency, and energy consumption in scenarios with a stable/fixed set of clients. However, addressing the dynamic availability of clients – a critical aspect of real-world scenarios – remains underexplored. This study delves into optimizing client selection and client-to-edge associations in HFL under intermittent client participation so as to minimize overall system costs (i.e., delay and energy), while achieving fast model convergence. We unveil that achieving this goal involves solving a complex NP-hard problem. To tackle this, we propose a stagewise methodology that splits the solution into two stages, referred to as Plan A and Plan B. Plan A focuses on identifying long-term clients with high chance of participation in subsequent model training rounds. Plan B serves as a backup, selecting alternative clients when long-term clients are unavailable during model training rounds. This stagewise methodology offers a fresh perspective on client selection that can enhance both HFL and conventional FL via enabling low-overhead decision-making processes. Through evaluations on diverse datasets, we show that our methodology outperforms existing benchmarks on crucial factors such as model accuracy and system costs.

Surveillance videos play a crucial role in public security. However, current tasks related to surveillance videos primarily focus on classifying and localizing anomalous events. Despite achieving notable performance, existing methods are restricted to detecting and classifying predefined events and lack satisfactory semantic understanding. To tackle this challenge, we introduce a novel research avenue focused on Video-and-Language Understanding for surveillance (VALU), and construct the first multimodal surveillance video dataset. We manually annotate the real-world surveillance dataset UCF-Crime with fine-grained event content and timing. Our newly annotated dataset, UCA (UCF-Crime Annotation), contains 23,542 sentences, with an average length of 20 words, and its annotated videos are as long as 110.7 hours. Moreover, we evaluate SOTA models on five multimodal tasks using this newly created dataset, establishing new baselines for surveillance VALU, from small to large models. Our experiments reveal that mainstream models, which perform well on previously public datasets, exhibit poor performance on surveillance video, highlighting new challenges in surveillance VALU. In addition to conducting baseline experiments to compare the performance of existing models, we also propose novel methods for multimodal anomaly detection tasks and finetune multimodal large language model models using our dataset. All the experiments highlight the necessity of constructing this multimodal dataset to advance surveillance AI. Upon the experimental results mentioned above, we conduct further in-depth analysis and discussion. The dataset and codes are provided athttps://xuange923.github.io/Surveillance-Video-Understanding.

This paper explores an interesting worker recruitment challenge where the mobile crowd sensing and computing (MCSC) platform hires workers to complete tasks with varying quality requirements and budget limitations, amidst uncertainties in worker participation and local workloads. We propose an innovative hybrid worker recruitment framework that combines offline and online trading modes. The offline mode enables the platform to overbook long-term workers by pre-signing contracts, thereby managing dynamic service supply. This is modeled as a 0-1 integer linear programming (ILP) problem with probabilistic constraints on service quality and budget. To address the uncertainties that may prevent long-term workers from consistently meeting service quality standards, we also introduce an online temporary worker recruitment scheme as a contingency plan. This scheme ensures seamless service provisioning and is likewise formulated as a 0-1 ILP problem. To tackle these problems with NP-hardness, we develop three algorithms, namely,i)exhaustive searching,ii)unique index-based stochastic searching with risk-aware filter constraint,iii)geometric programming-based successive convex algorithm. These algorithms are implemented in a stagewise manner to achieve optimal or near-optimal solutions. Extensive experiments demonstrate our effectiveness in terms of service quality, time efficiency, etc.

Vehicular clouds (VCs) play a crucial role in the Internet-of-Vehicles (IoV) ecosystem by securing essential computing resources for a wide range of tasks. This paPertackles the intricacies of resource provisioning in dynamic VCs for computation-intensive tasks, represented by undirected graphs for parallel processing over multiple vehicles. We model the dynamics of VCs by considering multiple factors, including varying communication quality among vehicles, fluctuating computing capabilities of vehicles, uncertain contact duration among vehicles, and dynamic data exchange costs between vehicles. Our primary goal is to obtain feasible assignments between task components and nearby vehicles, calledtemplates, in a timely manner with minimized task completion time and data exchange overhead. To achieve this, wepropose ahybrid graphtaskscheduling (P-HTS) methodology that combines offline and online decision-making modes. For the offline mode, we introduce an approach called risk-aware pilot isomorphic subgraph searching (RA-PilotISS), which predicts feasible solutions for task scheduling in advance based on historical information. Then, for the online mode, we propose time-efficient instantaneous isomorphic subgraph searching (TE-InstaISS), serving as a backup approach for quickly identifying new optimal scheduling template when the one identified by RA-PilotISS becomes invalid due to changing conditions. Through comprehensive experiments, we demonstrate the superiority of our proposed hybrid mechanism compared to state-of-the-art methods in terms of various evaluative metrics, e.g., time efficiency such as the delay caused by seeking for possible templates and task completion time, as well as cost function, upon considering different VC scales and graph task topologies.

Surveillance videos are important for public security. However, current surveillance video tasks mainly focus on classifying and localizing anomalous events. Existing methods are limited to detecting and classifying the predefined events with unsatisfactory semantic understanding, although they have obtained considerable performance. To address this issue, we propose a new research direction of surveillance video-and-language understanding (VALU), and construct the first multimodal surveillance video dataset. We manually annotate the real-world surveillance dataset UCF-Crime with fine-grained event content and timing. Our newly annotated dataset, UCA (UCF-Crime Annotation)11The dataset is provided at https://xuange923.github.io/Surveillance-Video-Understanding., contains 23,542 sentences, with an average length of 20 words, and its annotated videos are as long as 110.7 hours. Furthermore, we benchmark SOTA models for four multimodal tasks on this newly created dataset, which serve as new baselines for surveillance VALU. Through experiments, we find that mainstream models used in previously public datasets perform poorly on surveillance video, demonstrating new challenges in surveillance VALU. We also conducted experiments on multimodal anomaly detection. These results demonstrate that our multimodal surveillance learning can improve the performance of anomaly detection. All the experiments highlight the necessity of constructing this dataset to advance surveillance AI.

We present FlexDID, a decentralized identity management system with flexible credential presentation and efficient revocation. FlexDID allows identity holders to perform both vertical disclosure and horizontal disclosure. Vertical disclosure allows an identity holder to derive a credential to prove only a subset of all attributes it holds. Meanwhile, horizontal disclosure allows multiple identity holders with the same attributes to aggregate their credentials for fast verification. We also introduce a three-layer architecture. Besides the roles of identity holders and issuers in conventional identity management systems, we introduce a layer of secondary brokers. The secondary brokers can be viewed as delegates for identity holders to perform flexible disclosure. Together with our system optimizations, such as batch verification, FlexDID is able to manage the credentials for a large number of identity holders, making it a perfect fit for applications, such as the Industrial Internet of Things. Our evaluation shows all the operations of FlexDID can complete in millisecond level and the most computationally extensive operation has a latency of up to 34% lower compared to existing identity management systems.

Cloud-aidedmobileedgenetworks (CAMENs) allow edge servers (ESs) to purchase resources from remote cloud servers (CSs), while overcoming resource shortage when handling computation-intensive tasks of mobile users (MUs). Conventional trading mechanisms (e.g., onsite trading) confront many challenges, including decision-making overhead (e.g., latency) and potential trading failures. This paper investigates a series of cross-layer matching mechanisms to achieve stable and cost-effective resource provisioning across different layers (i.e., MUs, ESs, CSs), seamlessly integrated into a novel hybrid paradigm that incorporates futures and spot trading. In futures trading, we explore anoverbooking-drivenaforehandcross-layermatching (OA-CLM) mechanism, facilitating two future contract types: contract between MUs and ESs, and contract between ESs and CSs, while assessing potential risks under historical statistical analysis. In spot trading, we design two backup plans respond to current network/market conditions: determination on contractual MUs that should switch to local processing from edge/cloud services; and anonsitecross-layermatching (OS-CLM) mechanism that engages participants in real-time practical transactions. We next show that our matching mechanisms theoretically satisfy stability, individual rationality, competitive equilibrium, and weak Pareto optimality. Comprehensive simulations in real-world and numerical network settings confirm the corresponding efficacy, while revealing remarkable improvements in time/energy efficiency and social welfare.

By opportunistically engaging mobile users (workers), mobile crowdsensing (MCS) networks have emerged as important approach to facilitate sharing of sensed/gathered data of heterogeneous mobile devices. To assign tasks among workers and ensure low overheads, we introduce a series of stable matching mechanisms, which are integrated into a novel hybrid service trading paradigm consisting offutures tradingandspot tradingmodes, to ensure seamless MCS service provisioning. In futures trading, we determine a set of long-term workers for each task through anoverbooking-enabledin-advancemany-to-manymatching (OIA3M) mechanism, while characterizing the associated risks under statistical analysis. In spot trading, we investigate the impact of fluctuations in long-term workers' resources on the violation of service quality requirements of tasks, and formalize a spot trading mode for tasks with violated service quality requirements under practical budget constraints, where the task-worker mapping is carried out viaonsitemany-to-manymatching (O3M) andonsitemany-to-onematching (OMOM). We theoretically show that our proposed matching mechanisms satisfy stability, individual rationality, fairness, and computational efficiency. Comprehensive evaluations confirm the satisfaction of these properties in practical network settings and demonstrate our commendable performance in terms of service quality, running time, and decision-making overheads, e.g., delay and energy consumption.

Mobile Crowdsensing (MCS) has become an emerging paradigm to solve complex urban sensing problems by utilizing the ubiquitous sensing capacities of the crowd. One critical issue in MCS is to efficiently allocate tasks to users. We focus on addressing the task allocation problem in a distributed manner, where each user individually and freely makes his decision to undertake tasks. Existing distributed schemes simply consider users behave completely selfishly, which leads to inefficient solutions and damages the overall benefit of all users. Different from existing schemes, in this paper we integrate the social relationship into users' decision making and build a socially aware utility model for each user, which consists of both user's own utility and the weighted sum of his social neighbors' utilities. Based on this, we formulate a novel Socially Aware Task Selection (SATS) game for users in MCS. We theoretically prove the existence of pure Nash equilibrium in the SATS game with the help of a potential game framework. We further propose a distributed user selection algorithm to actually achieve the pure Nash equilibrium. Extensive simulations based on both real and synthetic social relationship graph datasets demonstrate that our approach can achieve more efficient solutions which improve users' overall benefit compared with existing schemes.

Channel qualities between UAVs vary drastically due to the mobility of UAVs. Conventional channel coding relies on channel state information (CSI) estimation and active bit rate selection and thus cannot adapt well to such varying channel conditions. In contrast, rateless codes can achieve almost optimal bit rate under varying channel conditions without CSI estimation and explicit rate selection. In rateless codes, Spinal codes are one of the most prominent solutions and perform much better than other rateless codes over the mobile fading channel between UAVs. However, Spinal codes still face the challenge of error accumulation effect, which largely hurts the transmission efficiency. In this paper, we for the first time analyze the error accumulation effect and its impact on the performance of Spinal codes under mobile fading channel conditions between UAVs. Furthermore, we propose a model for helping the decoder estimate the quality of each received symbol. Based on such model, trust function based Spinal codes (TFSC) are then proposed. Its main idea is to treat received symbols differently according to their qualities so that those symbols with better qualities can contribute more to the decoding process. Simulation results demonstrate that TFSC can significantly mitigate the error accumulation effect and improve the efficiency of Spinal codes, which achieves 1.1x to 4.4x overall performance improvement when compared with Spinal codes over the mobile fading channel between UAVs.

Data collection is key issue in vehicular networks since it is vital for supporting many applications in vehicular environments. With the explosive growth of sensing data in urban area, however, strategies for efficient collection of big data in vehicular networks are still far from being well studied. In this paper, we focus on studying this issue and accordingly propose a Software Defined Vehicular Networks (SDVN) architecture. On this architecture, a predictive data collection algorithm is proposed. In this algorithm, packet delivery is fulfilled by cooperative cellular and ad hoc network interfaces, in which collections of big data always adopts ad hoc based multi-hop relaying whenever applicable to forward packets to Road Side Units (RSUs). Cellular networks are used for data uploading only when no multi-hop relaying opportunity is available. Our proposed SDVN architecture enables such efficient cooperative communications, in which predictive routing decisions are made based on real-time network status other than empirical knowledge. Simulation results demonstrate that our algorithm outperforms existing algorithms in terms of packet delivery ratio and transmit efficiency.

Abstract Topology control is an efficient strategy for improving the performance of wireless ad hoc and sensor networks by building network topologies with desirable features. In this process, location information of nodes can be used to improve the performance of a topology control algorithm and also ease its operations. Many location‐based topology control algorithms have been proposed. In this paper, we propose two location‐assisted grid‐based topology control (GBP) algorithms. The design objective of our algorithm is to effectively reduce the number of active nodes required to keep global network connectivity. In grid‐based topology control, a network is divided into equally spaced squares (called grids). We accordingly design cross‐sectional topology control algorithm and diagonal topology control algorithm based on different network parameter settings. The key idea is to build near‐minimal connected dominating set for the network at the grid level. Analytical and simulation results demonstrate that our designed algorithms outperform existing work. Furthermore, the diagonal algorithm outperforms the cross‐sectional algorithm. Copyright © 2016 John Wiley & Sons, Ltd.

Much previous work had demonstrated the remarkable performance of backpressure based routing and scheduling algorithms in wireless sensor networks (WSNs). However, the absence of consideration on energy use efficiency in the design of existing backpressure based algorithms makes them difficult to be deployed in resource-limited WSNs. In this paper, we study how to improve the energy use efficiency of backpressure based algorithm. For this purpose, we propose an energy efficient backpressure routing and scheduling algorithm (EBP) for WSNs. In EBP, a new link weight calculation method is designed, based on which nodal energy status is considered when making decisions on backpressure based transmission scheduling. In EBP, packets are encouraged to be forwarded to nodes with more residual energy while the throughput-optimality of backpressure based algorithm is still preserved. Simulation results show that EBP can obtain significant performance improvements in terms of energy use efficiency, network throughput, and packet delivery ratio as compared with existing work.

In this paper, we study how to effectively reduce the average end-to-end (E2E) packet delay in backpressure based scheduling in wireless multi-hop networks. We accordingly propose a virtual gradient based back-pressure scheduling algorithm, referred to as VBR. In VBR, intentional virtual queue, whose length (called virtual gradient) depends on the distance to destination, is first built at nodes in a network in the network configuration phase. In this way, virtual gradient is established at nodes in the network. In the network operation phase, the scheduling decision at each node needs to jointly consider both real queue length and virtual queue length. Simulation results show that VBR can obtain significant performance improvement on back-pressure based routing and scheduling, in terms of packet delivery ratio and average E2E delay.

In this paper, we propose a location-based friend-assisted coding-aware routing protocol (LFCR) for wireless multihop networks. To achieve improved network throughout, LFCR performs inter-flow network coding based routing with the assistance of location information. Specifically, LFCR combines friend-assisted path discovery and coding-aware routing. Further, when making decision on next hop selection, LFCR takes into account both coding opportunities and forwarding progress in next hop selection and attempts to make a good tradeoff between them. Simulation results show that LFCR significantly outperforms existing work in terms of network throughput, packet delivery ratio, and coding frequency.

Recently, much effort has been made for implementation of back-pressure scheduling in wireless networks. In this paper, we explore the implementation of back-pressure-based forwarding in wireless sensor networks. For this purpose, we propose Gradient-pressure, a practical Gradient-assisted anycast-based back-pressure framework for wireless sensor networks. Gradient-pressure introduces gradient information to assist transmission scheduling and realizes distributed anycast-based back-pressure scheduling on top of IEEE 802.11. Simulation results demonstrate that Gradient-pressure has high performance in terms of energy-use efficiency and goodput.

ABSTRACT Achieving high data quality and efficient network resource utilization is two major design objectives of wireless sensor networks (WSNs). However, these two objectives are often conflictive. By allowing sensors to report sampled data at high rates, fine‐grained data quality can be obtained. However, the limited resources of a WSN make it difficult to support very high traffic rate. Therefore, the capability of adaptively adjusting sensor nodes' traffic‐generating rates on the basis of the availability of network resources and application requirements is critical. This issue has attracted much attention recently, and some work has been carried out. To achieve high data quality and improved utilization of network resources, in this paper, we propose rate‐based adaptive precision setting (RAPS) protocol, which works in a way that each sensor can adaptively adjust its traffic‐generating rate on the basis of the current network resources availability and application requirements. RAPS introduces the following two key factors into its design: application's precision requirement and packet arrival rate. Analytical and simulation results show that RAPS can achieve improved data quality while reducing packet delivery latency. Copyright © 2012 John Wiley & Sons, Ltd.

In this paper, we propose an efficient network coding based back-pressure algorithm (NBP). NBP introduces the interflow network coding to improve the performance of the backpressure algorithm (a famous throughput-optimal cross-layer scheduling algorithm) for scheduling the transmissions of packets and also higher transmission efficiency. We theoretically prove that NBP can stabilize such networks. Simulation results demonstrate that NBP significantly outperforms traditional back-pressure algorithm in terms of packet delivery delay and average forwarding queue length.