This paper addresses the challenges of unmanned aerial vehicle (UAV) assisted mobile edge computing (MEC) in high-density user mobility scenarios, a field that has not been extensively explored in current research. We introduce a novel deep reinforcement learning (DRL) framework, named “j-PPO+EN-ConvNTM”, specifically designed to optimize MEC performance in urban environments with user mobility. The framework integrates spatiotemporal data modeling and spatial transformer network (STN) through an enhance Convolution Neural Turing Machine (EN-ConvNTM) module, which includes a three-dimensional external memory. It also features a joint continuous and discrete action decision-making module, termed joint proximal policy optimization (j-PPO). This design enables effective handling of the dynamic and complex nature of urban mobility patterns. The proposed approach extends the PPO technique to accommodate joint continuous and discrete action decisions, thereby enhancing UAV adaptability and efficiency in providing MEC services. Extensive simulations demonstrate significant improvements over all baseline models, particularly in terms of equilibrium efficiency and service continuity in high-density scenarios. Our research addresses a critical gap in existing UAV-assisted MEC studies, which primarily focus on static or low-mobility user scenarios, and supports the development of more robust and efficient smart city applications, meeting the real-world demands of modern urban infrastructures.

Due to data heterogeneity and parameter transmission failure, federated learning (FL) usually suffers performance degradation in wireless networks. To address this issue, a novel federated Learning via reusing historical Information (FedRe) is proposed for convergence acceleration and performance enhancement without any additional communication cost. Specifically, an analytical model is first built to characterize the impacts of data heterogeneity and parameter transmission failure on FL performance. Based on this, a local gradient correction method and a statistical aggregation correction method are proposed to deal with data heterogeneity and transmission unreliability problems respectively. Additionally, the convergence of the proposed FedRe is proved and analyzed theoretically. Finally, extensive numerical results are presented with experiments on two public datasets to validate the effectiveness of FedRe with comparisons of baselines.

This paper proposes a novel FL method called FedFNS, which integrates feature norm regularization and statistical aggregation. Specifically, a local model correction strategy is proposed to reduce client bias through incorporating a regularization term into the loss function. To mitigate the negative effects of unreliable communications on global model performance, a statistical weighted aggregation strategy is proposed, which leverages the transmission success probability of each client. The effectiveness of FedFNS is validated through extensive experiments, demonstrating its superiority over several classic FL methods in terms of accuracy.

In order to fully explore the potential of intelligent reflecting surfaces (IRSs) in millimeter-wave (mmWave) communication systems and maximize system sum-rate performance within a user-centric framework, this paper investigates the joint optimization problem of user multiple association, transmit beamforming, and cooperative IRS passive beamforming. Meanwhile, the impact of IRS location on user association is also studied. Due to the deep coupling of multiple variables, the modeled problem is a complex non-convex optimization problem. To address it, an efficient alternating iterative optimization algorithm based on the Lagrangian dual decomposition and fractional programming techniques is proposed. Simulation results show that compared with traditional methods, the proposed algorithm significantly improves the system sum rate, validating its effectiveness.

The development of intelligent transportation has generated many ultra reliable low latency communication (URLLC) tasks, which require sufficient communication and computation resources for task offloading and processing. Although mobile edge computing (MEC) provides a promising solution, its efficiency is subject to the limited knowledge and analysis capability on the physical networks. Therefore, in this paper, we propose a digital twin (DT) empowered MEC framework to strengthen the MEC task offloading efficiency in cellular vehicle-to-everything (C-V2X) networks. Our proposed DT is constructed through a hybrid data-driven and model-driven approach to capture the realistic transportation network features. Then, DT leverages the metric of time to collision to predict vehicular safety levels and estimates the corresponding URLLC task requirements of future time slots. The prediction results are further utilized to make decisions on the URLLC resource reservation. Different from conventional studies, we consider the influence of DT’s inaccurate predictions (i.e., the prediction with error) on the resource allocations. Specifically, the inaccurate DT prediction results are considered as uncertain constraints of the resource reservation problem. A robust parameter from the robust optimization is adopted to adjust the tradeoff between the problem uncertainty and solution optimality degree. Further, we leverage the optimized resource reservation results to construct the task offloading problem. The problem is decoupled into two sub-problems of channel resource allocation and computation resource allocation, respectively. And a two-stage matching algorithm is developed to solve each sub-problem based on the resource reservation constraints. Finally, realistic road information is mapped into DT for simulations. Simulation results validate the advantages of our proposed approach by comparing with existing schemes.

This paper proposes a lightweight video semantic communication (LVSC) method to enable end-to-end wireless video transmissions. The proposed LVSC method cuts down the model parameters through designing a spatial pyramid structure for the residual codec and a one-dimensional CNN-based channel attention model for the joint source-channel coding (JSCC) codec, respectively. Additionally, the Swin Transformer is introduced to the JSCC codec to accelerate the semantic coding. Extensive experiments validate that the proposed LVSC performs better than the baseline methods in terms of PSNR and MS-SSIM while inferring much faster at the same time.

This paper investigates the Virtual Reality (VR) service delay performance for an edge-terminal cooperative system. A convolutional neural network (CNN)-based user interest analysis method is first proposed to characterize the content requesting behavior. Based on this, a service delay minimization problem is formulated with consideration of performance fairness among users. Thereafter, the caching and computing scheme is derived for each user with given communication resources, while a bisection-based communication resource allocation scheme is derived with given content caching and offloading information. With alternative optimization, a joint caching, computing and communication scheme is proposed. The effectiveness of the proposed scheme is finally validated by simulation results.

This paper presents a study on peak age of information (AoI), focusing on its distribution that is more important for AoI guarantees than the mean. Specifically, the relation of peak AoI to the underlying information generation and transmission processes is explicitly formulated. Based on this formulation and by exploring the independence information between the information generation and transmission processes, a general bound on the distribution of peak AoI is derived. To showcase the use of the derived bound, it is applied to two representative cases, which are characterized by the M/M/1 and D/M/1 queuing models. Numerical results obtained from the proposed bound analysis are finally introduced and discussed in comparison with exact results, validating the bound.

Federated learning (FL) enables distributed clients to cooperatively train a common machine learning (ML) model for intelligent inference without raw data sharing. However, problems in practical networks, such as non-independent-and-identically-distributed (non-iid) raw data and limited network resources, lead to slow and unstable convergence of the FL training process. To address these issues, this paper proposes a new FL method that can mitigate statistical heterogeneity through the depersonalization mechanism. Specifically, we decouple the global and local optimization objectives by alternating stochastic gradient descent, thus reducing the accumulated variance in local update phases to accelerate the FL convergence. Furthermore, the proposed FL method is analyzed in detail and proved to converge at a sublinear speed under the general non-convex setting. Finally, extensive experiments are conducted on public datasets to verify the effectiveness of the proposed method with comparisons of other representative baseline methods.

Screen content image/video has been widely applied into cloud game, AR/VR, online education etc., while the transmission and storage of screen content data is limited, and existing compression methods tend to lose details when encoding screen content with low bitrate. To address this challenge, a low bitrate screen content compression method based on super-resolution is proposed in this paper. The method combines super-resolution (SR) and down-sampling techniques to compress screen content while preserving its details. A novel super-resolution network is designed, which enhances detail preservation during feature fusion and eliminates compression distortion artifacts. The proposed method was evaluated against the state-of-the-art HEVC-SCC and can save bit rate by 19.77% and 26.18% on the SCID and SIQAD datasets, respectively while maintaining similar subjective quality.

Versatile Video Coding (VVC/H.266) greatly improves the compression performance at the cost of extremely high computational complexity compared with High Efficiency Video Coding (HEVC/H.265). Within the context of mobile devices with limited power and computational capabilities, reductions on encoding complexity are important; particularly to encode new data formats such as screen content sequences. Therefore, in this paper, aimed at two brand-new techniques with high complexity, matrix-weighted intra-prediction (MIP) and intra-sub-partition (ISP), we propose a fast intra-prediction scheme for VVC screen content coding based on ultra-lightweight convolutional neural network (CNN). Firstly, an ultra-lightweight CNN is designed to segment the image into the natural content region (NCR) and the screen content region (SCR). Then, an adaptive intra-coding mode pruning scheme based on the ultra-lightweight CNN is proposed to accelerate the VVC screen content coding process. Finally, our proposed method was implemented into VTM-10.0 and experimental results show that our method can save averagely 7.68% and maximum to 13.62% of the encoding time without encoding performance loss.

With the 5G commercial deployments rapidly proceeding, operating mobile networks efficiently has become a great challenge. Self-organizing networks have been proposed to focus on automatically monitoring, analyzing and optimizing networks. Regarding the growing scale and complexity, 5G self-organizing networks confront the challenge to handle the massive data. Therefore, AI models are urgently needed to enable end-to-end network automation. In this demonstration, benefiting from the open data interfaces of the standardized NWDAF within the 5GC network, we implement a containerized network data analytic framework embedding Docker-based AI model containers into 5G networks. Furthermore, we simulate a use case automatically monitoring and optimizing user-level QoE in real time and simulation results are presented.

With the massive and rapid deployment of Internet of Things (IoT) devices, the number of IoT devices in the cloud radio access network (C-RAN) has increased dramatically. It is intractable to provide deterministic Quality-of-Service (QoS) guarantee for task processing due to the task burstiness and the time-varying channels. In order to guarantee latency requirements of energy-limited IoT devices in C-RAN, this article proposes a statistical latency guarantee scheme for task computation. With the assistance of densely distributed remote radio heads (RRHs) providing wireless power transfer to the IoT devices, a wireless-powered edge computing C-RAN model is constructed. Then, the problem of minimizing task latency violation probability is formulated. With the help of effective capacity theory, the problem is decoupled into multiple subproblems, which are QoS parameter optimization, task offloading optimization, wireless power transfer, and energy allocation. Thereafter, low-complexity schemes are designed to jointly optimize the wireless power transfer, energy allocation, and task offloading. The effectiveness of the proposed statistical latency guarantee scheme is finally validated by extensive simulations.

This article focuses on the service freshness guarantee for cooperative sensing in the Internet of Everything. Specifically, the peak Age of Information (AoI) is first introduced to evaluate the information and service freshness. An analytical model is then constructed to decouple the components of peak AoI into the interarrival time and transmission time. With the knowledge of identical increments of update arrival and transmission process, a close bound of peak AoI violation probability is derived based on martingale theory. Furthermore, the impact of source node parameter configuration on the peak AoI violation probability are analyzed and a task allocation scheme is proposed to guide cooperative sensing. Numerical analysis validates the tightness of the peak AoI violation probability bound and the effectiveness of the proposed scheme in service freshness guarantee.

Edge caching enables low-delay and high-quality data services for the Industrial Internet. However, traditional popularity-based edge caching ignores the diversity and evolution of user interest, especially among user groups, and therefore has the limited quality of experience guarantees for users. In this regard, a recommendation-aided edge caching approach is proposed to leverage the time-varying user interest. Specifically, a dynamic interest capture model was proposed to mine the individual user interest, based on which, a group interest aggregation algorithm is then studied to determine the content caching strategies for edge nodes. Thereafter, an edge content recommendation is further proposed to optimize the cache hit ratio while ensuring a satisfying recommendation hit ratio based on the personalized user interest and given caching decision. The effectiveness of the proposed approach is finally validated by comparing it with other baseline approaches.

In order to make full use of the network data and guarantee user privacy simultaneously, federated learning (FL) is proposed to enable distributed intelligence for local nodes without sharing data with each other. However, in practice, due to resource limitations, traditional FL suffers from node scheduling and parameter transmission failure, which not only affects the final performance but also further reduces the fairness of the participating nodes. This article addresses the challenge and proposes an FL method to enhance the performance of FL on the basis of guaranteeing the fairness of the local nodes in a resource-constrained Internet of Things (IoT) network. Specifically, an analytical model is first constructed to characterize the performance of FL with joint considerations of node fairness, unreliable parameter transmissions as well as resource limitations. Thereafter, a statistically reweighted aggregation (SRA) scheme is proposed for parameter aggregation and the corresponding model is proved to be unbiased to that based on ideal parameter transmissions. With the knowledge of time dependency of the global model, we further extend SRA and propose a reliable SRA (RSRA) scheme. Additionally, we prove RSRA is able to achieve higher stability performance than SRA in model training. Furthermore, the convergence bound of the proposed RSRA is derived analytically, based on which an adaptive local training scheme is proposed under a given resource budget. Finally, extensive experiments are carried out with a public data set to validate the effectiveness of the proposed scheme with comparisons of other baseline schemes.

In Industrial Internet application scenarios, due to the ubiquitous connection requirements of the massive Internet of Things (IoT) devices in the edge layer. The data transmission rate is reduced, and the transmission delay increases, increasing the transmission energy consumption per bit. So a low-energy transmission strategy based on real-time edge layer traffic sensing is proposed. First, a mixed-integer modeling method for low-energy transmission of the IoT is proposed. This method aims to optimize the overall energy consumption of the system. The low-energy transmission task of the IoT is modeled as a mixed-integer linear programming problem. Second, a traffic prediction method for the estimation of the number of access packets is designed. Solve the problem of fog access point (F-AP) state change caused by the real-time change of network load. Finally, an energy-driven mapping strategy is designed. The transmission task can be dynamically mapped to the appropriate F-AP. The simulation results show that the strategy proposed in this article can effectively reduce the transmission energy consumption of IoT devices and the overall energy consumption of the system in the massive device access scenario.

To guarantee heterogeneous performance requirements of diverse vehicular services, it is necessary to design a full cooperative policy for both vehicle to infrastructure (V2I) links and vehicle to vehicle (V2V) links. This paper investigates how to improve the quality of experience (QoE) of the V2I users for video services while satisfying the delay requirements of both V2I and V2V links. In specific, a QoE maximization problem is formulated with consideration of vehicular collaboration where task offloading decision, channel reuse decision and power allocation of V2V users are all included. A multi-agent reinforcement learning (MARL) framework is then designed, where a new reward function is proposed to evaluate the utility of the considered network. Thereafter, a proximal policy optimization approach is proposed to enable each V2V user to learn policy individually with the shared global network reward. The effectiveness of the proposed approach is finally validated with comparison of other baseline approaches through extensive simulation experiments.

How to improve the content caching efficiency and user coverage rate at the same time is a fundamental challenge in edge caching networks. This paper studies an edge caching scheme based on user interest to address this issue. Specifically, a group interest aware caching framework is first developed. An individual interest prediction scheme is then proposed by merging factorization machine (FM) model and multi-layer perceptron (MLP) model, where both low-order and high-order features can be well learned simultaneously. Thereafter, the group interest is represented by a weighted average approach, based on which a caching scheme is further proposed. Moreover, the effectiveness of the proposed method is validated by extensive experiments with a real-world dataset.

With the tremendous growth of traffic demand caused by the conventional service and emerging Internet-of-Things (IoT) applications, it is vitally challenging to support seamlessly access for billions of IoT devices along with existing Internet service. Fiber-Wireless (FiWi) access networks is an ideal solution for the next-generation access networks to support both IoT and conventional service. One of the main challenges faced by the design of an FiWi access network is the high energy consumption due to low utilization of optical network units (ONUs) and high control overhead of data transmission. In this article, An adaptive frame aggregation scheme with load transfer is proposed to reduce the energy consumption in FiWi. By evaluating the quality of wireless channel, the proposed scheme adaptively adjusts the length of the aggregated frame to reduce the energy consumption caused by the frequent preemption of wireless channel and the numerous retransmission of data frames resulted by poor channel quality. In the optical backhaul, according to the delay requirements of services with different priorities, the proposed scheme calculates the optimal frame length for each queue and performs load transfer among ONUs to dynamically distribute network load and maximize the sleep rate of ONUs for energy saving. Simulation results and theoretical analysis both unanimously show that the proposed scheme achieves significant amounts of energy saving, while preserving delay performance of various services.

In this paper, a video service enhancement strategy is investigated under an edge-cloud collaboration framework, where video caching and delivery decisions are made at the cloud and edge respectively. We aim to guarantee the user fairness in terms of video coding rate under statistical delay constraint and edge caching capacity constraint. A hybrid human-artificial intelligence approach is developed to improve the user hit rate for video caching. Specifically, individual user interest is first characterized by merging factorization machine (FM) model and multi-layer perceptron (MLP) model, where both low-order and high-order features can be well learned simultaneously. Thereafter, a social aware similarity model is constructed to transfer individual user interest to group interest, based on which, videos can be selected to cache at the network edge. Furthermore, a dual bisection exploration scheme is proposed to optimize wireless resource allocation and video coding rate. The effectiveness of the proposed video caching and delivery scheme is finally validated by extensive experiments with a real-world dataset.

Internet of things (IoT) offers advanced and intelligent services for our life. However, smart IoT devices also bring various security vulnerabilities. Traditionally, attacks are solved by conventional authentication and authorization schemes, requiring extensive time and computational resources. In addition, it is possible to exploit artificial intelligence (AI) to provide countermeasures while enabling lightweight authentication. In this paper, we explore a solution on modelling a spoofing detection system based on machine learning and we propose a deep learning method using Auto-Extractor/Classifier Neural Network. Our scheme operates on the physical layer without causing computational overhead. Therefore, the lightweight authentication can be achieved and spoofing attacks are well- controlled in IoT scenarios.

In this paper, a video collaborative distribution mechanism is studied in Cloud Radio Access Networks (C-RANs) with object to guarantee the quality of experience (QoE) for different users. Specifically, a framework which enables the remote radio head (RRH)-to-device (R2D) technology to cooperate with the device-to-device (D2D) technology is constructed to transmit video traffic efficiently. Besides, a new QoE evaluation model is built in terms of the transition characteristics of video quality version and the interruption characteristics of video transmissions. Then, the optimal choice of video quality version is studied to achieve a good tradeoff among the video quality, interruption and smoothness for a target user. Moreover, a resource allocation policy is proposed to reduce the mean latency caused by video interruption of the whole network. Simulation results verify that the proposed mechanism performs better than other existing ones when the latency jointly caused by the quality version transition and the transmission interruption is sensitive to the users.

In this paper, the buffer-constrained throughput performance of a multi-user wireless-powered communication (WPC) system is investigated, where energy harvesting follows a non-linear model. The investigation focuses on the buffer overflow performance of sending data in the downlink (DL) from the access point (AP) node to each user equipment (UE) node and that in the uplink from each UE node to the AP node, based on which the throughput performance on both directions when a buffer constraint is enforced is studied. Specifically, the buffer overflow probability at each node is analyzed, based on which the buffer-constrained throughput is studied. In addition, to ensure the throughput performance under the buffer constraint, the DL transmission power allocation policy and the required energy storage capacity at each UE are investigated. Also, the optimal channel time allocation policy is studied with the objective of maximizing the minimum buffer-constrained throughput guaranteed to each UE at the same time. To this aim, an optimization problem is first formulated and then a dichotomy-based time allocation algorithm combined with a one-dimensional search is proposed to solve this problem. The analysis and results, explicitly relating the throughput to the buffer constraint in addition to WPC characteristics, shed new light on the design and performance analysis of WPC systems.

As one of the crucial application scenarios, URLLC (Ultra-Reliable and Low Latency Communication) has been put on the agenda with the development of the 5G communication technology. This paper focuses on the downlink evaluation of shortened TTI (Transmission Time Interval), which is the pivotal characteristic of URLLC. In order to describe the varying control overhead of shortened TTI, we propose a model basing on the traffic arriving rate and the TTI length. What's more, we design a downlink scheduling algorithm to take the impact of shortened TTI into consideration. The system-level simulation verifies that shortened TTI is able to achieve high throughput as well as low latency due to low transmission delay at low system load. However, benefit brought by shortened TTI decreases as the load increases since its higher relative control overhead may lead to non-negligible queuing delay facing high system load. In this situation, system should choose relatively long shortened TTI configurations to reduce the influence of queueing delay.

In this paper, we investigate the delay and delay-constrained throughput performance of a point-to-point wireless-powered communication system, where one node, e.g. a user equipment (UE), is powered by the wireless energy transferred from the other node, e.g. an access point (AP), and uses the harvested wireless energy to send data to the other node. Our focus is on the delay performance of sending data over the uplink from the UE node to the AP node, and on its throughput performance when a delay constraint is enforced. Two representative time allocation schemes in using the link for the AP node to transfer energy (maybe together with data) and for the UE node to send data are considered. In particular, a lower bound on the cumulative capacity of the uplink is derived. In addition, an upper bound on the delay distribution is obtained for stochastic traffic arrivals, based on which, the delay-constrained throughput performance is further analyzed. Moreover, the accuracy of the analysis is validated by comparison with extensive simulation results. The analysis and results shed new light on the performance of such a wireless-powered communication system.

Studies of multi-server networks are usually conducted on the assumptions of independent servers and specific arrivals, which may not capture the characteristics of realistic networks. In this paper, we propose a novel stochastic network calculus approach to perform delay analysis for a general multi-server network. The stochastic network service curves are derived in two ways with different assumptions. After that, we use these service curves to derive queueing delay bounds including delay bound distribution and mean delay bound for a specific scenario of which the classical M/M/N model is a subset. Compared with previous studies, it is worth highlighting that our analysis contains the cases where the arrival and service processes as well as the servers can either be independent or correlated. In addition, the accuracy of the analytical delay bounds are verified by comparing with the queueing theory results in an M/M/N model.

Energy evaluation of wireless transmission for data service under different QoS (quality of service) constraints is a hot topic in green communications. In this paper, we employ the theory of the stochastic network calculus to investigate the relationship between the traffic arrival rate and energy harvesting rate under the delay and energy storage constraints. We first construct a wireless system model which works only by consuming the harvested renewable energy. Also, stochastic traffic arrivals, packet size as well as a two-state Markov chain interference on energy harvesting are jointly considered to ensure the accuracy of the analysis. We derive the minimum energy harvesting rate needed to satisfy the delay and energy storage constraints under a given average arrival rate. The efficiency of the harvested energy is also studied and shown to be convex in the numerical simulation. Numerical results illustrate that while the minimum energy harvesting rate is positively correlated with the packet size, it is negatively related to the delay requirement, under a given arrival rate. Additionally, the probabilistic bound of energy insufficiency is shown to be positively correlated with the transition cycle of the interference.