Since there exists a single point of server failure in conventional centralized federated learning, the decentralized federated learning (DFL) framework has become increasingly popular in recent years. However, when a large number of edge devices participate in DFL, it requires frequent model interactions between edge devices and long convergence time. In this work, we combat the impact of device heterogeneity in the large-scale DFL framework. To optimize communication efficiency and reduce the network complexity in large-scale DFL framework, we propose a decentralized edge devices clustering (DEDC) approach which leverages dense connectivity as the foundation to group edge devices with similar data distributions into clusters, thereby forming a novel multi-cluster decentralized federated edge learning (MD-FEEL) framework. The clustering method is adaptive, meaning it can effectively work across various network topologies, as long as the network is connected. We propose an asynchronous algorithm in the formed MD-FEEL framework, which consists four steps, i.e., local stochastic gradient descent (SGD) update, gradient consensus, intra-cluster model aggregation and inter-cluster model aggregation. We prove the convergence of our proposed asynchronous MD-FEEL algorithm on a non-convex setting and elaborate on the effect of some hyperparameters. Empirically, we evaluate our proposed asynchronous MD-FEEL on the MNIST and CIFAR-10 datasets. The simulations show that our proposed asynchronous MD-FEEL can perform better in terms of convergence speed and generalization performance than some benchmark algorithms.

Semantic communication, as one of the key technologies for next-generation wireless communications, shows great potential for improving transmission effectiveness. In this paper, we investigate efficient resource allocation methods in the multi-user multiple-input single-output (MU-MISO) image semantic communication system. To this end, we first construct a function in closed-form for semantic similarity accuracy using a data regression approach. A multi-user weighted sum semantic similarity maximization problem is then formulated, by jointly optimizing the semantic downsampling module and transmit beamforming, following both the transmit power and transmission delay constraints. Due to the finite and dependent nature of the original problem, we decompose the problem into multiple transmit beamforming problems in parallel. For the transmit beamforming problem, we use the successive convex approximation (SCA) technique and transform the problem into a second-order cone programming. A special case study is further discussed to reduce the complexity of transmit beamforming problem as a linear programming. Extensive experiments demonstrate the superiority of the proposed algorithm compared to conventional beamforming methods.

Catastrophic forgetting remains a central challenge in continual learning (CL) with pre-trained models. While existing approaches typically freeze the backbone and fine-tune a small number of parameters to mitigate forgetting, they still rely on iterative error backpropagation and gradient-based optimization, which can be computationally intensive and less suitable for resource-constrained environments.To address this, we propose FoRo, a forward-only, gradient-free continual learning method. FoRo consists of a lightweight prompt tuning strategy and a novel knowledge encoding mechanism, both designed without modifying the pre-trained model. Specifically, prompt embeddings are inserted at the input layer and optimized using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), which mitigates distribution shifts and extracts high-quality task representations. Subsequently, task-specific knowledge is encoded into a knowledge encoding matrix via nonlinear random projection and recursive least squares, enabling incremental updates to the classifier without revisiting prior data.Experiments show that FoRo significantly reduces average forgetting and improves accuracy. Thanks to forward-only learning, FoRo reduces memory usage and run time while maintaining high knowledge retention across long task sequences. These results suggest that FoRo could serve as a promising direction for exploring continual learning with pre-trained models, especially in real-world multimedia applications where both efficiency and effectiveness are critical.

The rapid development of Unmanned Ground Vehicles (UGVs) and edge computing in urban environments has enhanced UGVs’ data acquisition and processing capabilities. Since critical information in urban roads is distributed across all segments, the UGV must actively traverse them to comprehensively acquire and process the data. To address this, we adopt a Chinese Postman Problem (CPP) framework to ensure complete data acquisition from all target arcs in the road network and incorporate a time-decaying reward collection model on each target arc to handle various time-sensitive data types. Leveraging both onboard and roadside unit computing resources for collaborative processing, we formulate a non-convex mixed-integer optimization problem to maximize the collected reward that integrates path planning, UGV speed control, offloading decisions, and energy management. We propose an iterative algorithm to solve this challenging problem effectively, in which an improved low-complexity Branch-and-Bound algorithm is developed. Numerical results demonstrate the superiority of our system design and the proposed algorithm.

The rapid advancement of unmanned aerial vehicles (UAVs) and mobile edge computing (MEC) has spurred growing interest in multi-UAV cooperative systems across various applications. However, effectively coordinating multiple UAVs for target point (TP) data processing, task offloading, and path planning remains challenging, particularly when aiming to minimize task completion time. To address these challenges, we propose a joint task assignment and resource allocation optimization (TARO) approach for efficient TP access in multi-UAV MEC systems. Specifically, we formulate an optimization problem that minimizes the task completion time while balancing task assignment through path planning, task offloading and bandwidth allocation. Given the mixed-integer nonlinear programming (MINLP) nature of the problem, we develop an efficient optimization algorithm by decomposing it into three subproblems. We propose an iterative algorithm to solve the challenging problem effectively. Simulation results demonstrate that the proposed TARO approach significantly reduces task completion time compared to traditional benchmarks.

Current multi-modal object re-identification approaches based on large-scale pre-trained backbones (i.e., ViT) have displayed remarkable progress and achieved excellent performance. However, these methods usually adopt the standard full fine-tuning paradigm, which requires the optimization of considerable backbone parameters, causing extensive computational and storage requirements. In this work, we propose an efficient prompt-tuning framework tailored for multi-modal object re-identification; dubbed DMPT, which freezes the main backbone and only optimizes several newly added decoupled modality-aware parameters. Specifically, we explicitly decouple the visual prompts into modality-specific prompts which leverage prior modality knowledge from a powerful text encoder and modality-independent semantic prompts which extract semantic information from multi-modal inputs, such as visible, near-infrared, and thermal-infrared. Built upon the extracted features, we further design a Prompt Inverse Bind (PromptIBind) strategy that employs bind prompts as a medium to connect the semantic prompt tokens of different modalities and facilitates the exchange of complementary multi-modal information, boosting final re-identification results. Experimental results on multiple common benchmarks demonstrate that our DMPT can achieve competitive results to existing state-of-the-art methods while requiring only 6.5% fine-tuning of the backbone parameters.

Integrating large language models (LLMs) with the Internet of Things (IoT) offer great potential for enhancing vehicle personalization and adaptability in autonomous driving (AD), particularly in open-world scenarios. However, the increasing scarcity of high-quality public data poses a challenge, which could hinder the progress of LLMs in AD. To address this, we propose a novel approach, federated instruction tuning (FIT), that leverages federated learning (FL) to enable collaborative training of a shared model across multiple data owners without sharing raw data, thereby preserving privacy and mitigating data scarcity. Complementing FIT, we introduce a feature diversity (FD) strategy that enriches visual and textual diversity and significantly expands AD data by generating new instruction-following data across key dimensions, such as time, weather, and occlusion. Extensive experiments using LLaMA-Adapter as the base model and four FL methods validate the effectiveness of the FIT framework and the FD strategy. Our analysis also compares LLMs ranging from 1.1 to 7B parameters, with results evaluated using GPT score, demonstrating the potential of FIT in AD. Our findings suggest that FIT and FD can support intelligent network operation and optimization in IoT, benefiting both the AD and artificial intelligence (AI) industries.

With the convergence of the social networks and the Internet of Things (IoT), social IoT (SIoT) has emerged as a promising application scenario of federated learning. Meanwhile, most centralized federated learning (CFL) algorithms encounter single-point-of-failure risks and high bandwidth pressure at the central server. Therefore, decentralized FL (DFL) has been widely studied in recent years. However, when a substantial number of social nodes participate in DFL, the model consensus process requires a significant amount of communication among social nodes. This incurs a high communication overhead and low training efficiency, especially for the SIoT with dynamic network topology. In this work, we propose a communication-effective DFL algorithm for a general dynamic SIoT network with a large number of social nodes. To improve the communication efficiency and simplify network complexity, we employ a limited label propagation algorithm (LLPA) to periodically cluster social nodes into a dynamic multi-cluster decentralized federated learning (DMC-DFL) framework. We design an effective algorithm in the formed DMC-DFL framework, which consists of three steps, i.e., local update, intra-cluster communication and inter-cluster communication. Empirically, we conduct extensive comparison and ablation experiments based on four datasets. The experiment results validate the feasibility of DMC-DFL algorithm in both static and dynamic SIoT networks and illustrate the superiority of DMC-DFL algorithm over some benchmark DFL algorithms.

Reality digital twin (DT) model needs to be updated periodically, then the physical entity can be maintained efficiently. Since some physical entities may not be able to upload the entire status information actively, and also considering the universality of practical applications, we propose to use unmanned ground vehicle (UGV) as an information collector to assist in updating the reality DTs, where the UGV iterates over each target point (TP) to gather information by on-board sensors. Furthermore, considering the large amount of updating data and the limited computing resources on the UGV, we leverage mobile edge computing (MEC) technology to collaboratively process the data. In addition, to quantify the freshness of DTs, we propose the concept Age of DTs (AoDTs) as a metric to quantify the freshness of DT model. Thus, an AoDT minimization problem is established, which jointly optimizes offloading decisions, UGV waypoints selections, and TPs’ visiting orders, while also taking the obstacle avoidance into account. Considering the difficulty of the problem, we propose a novel low-complexity iterative algorithm to solve it. During which, a modified traveling salesman problem (TSP) solution is also proposed by taking into consideration the additional distance required to bypass the obstacles on each interwaypoints path. Finally, extensive simulation results show that the proposed algorithm can effectively reduce the AoDT, comparing to the benchmark algorithms.

Data acquisition and processing (DAP) by uncrewed aerial vehicle (UAV) hold the key to a diverse range of practical applications. However, UAVs often face challenges in completing complex DAP tasks due to inadequate on-board resources. To ameliorate such difficulties, previous researches have explored the mobile edge computing (MEC)-assisted UAV DAP, but they do not prioritize the mobility of MEC servers, resulting in challenges when communication and computation resources are limited. To address this issue, this work proposes a novel design that incorporates a movable uncrewedground vehicle (UGV)-mounted MEC server to assist UAV DAP, which enables the UAV to thoroughly leverage edge resources. Under this design, we aim to maximize UAV DAP ability with constraints including real-time data processing, UGV obstacles avoidance and system motion. Efficient algorithms are proposed, with obstacles avoidance problem solved, from a systematic perspective. Extensive simulation and real-world experiment results demonstrate the effectiveness of our system design and algorithms.

Network slicing emerges as a promising solution for accommodating diverse services with varying quality of service (QoS) requirements. This paper delves into the problem of achieving energy-efficient network slicing in heterogeneous wireless networks. We first introduce a novel performance metric to evaluate the energy efficiency of slice. In order to reduce energy consumption and enhance energy efficiency, an adaptive slice switching mechanism is proposed, which enables slices to dynamically switch on and off based on real-time network traffic. Furthermore, we formulate a two-timescale optimization problem that jointly addresses slice switching and resource allocation to maximize long-term slice average energy efficiency while guaranteeing service delay requirements. To solve the joint problem, we decouple the problem into two subproblems in different timescales and develop a learning-based two-layer slice switching and resource allocation (SWEET) algorithm to make decisions in an online manner. Specifically, the long-timescale slice switching decisions are determined via a reinforcement learning algorithm in an outer layer, while the short-timescale resource allocation decisions are determined via leveraging coalition game and convex optimization in an inner layer. Extensive simulation results based on real-world datasets demonstrate that the SWEET algorithm yields an average improvement of 16.45% in slice energy efficiency, as well as adapts to network dynamics.

In conventional federated learning (FL), there exists a single point of failure in the central server. Thus the studies about decentralized federated learning (DFL) paradigm have become popular recently. In DFL, some clients with poor computation capacity may take a long time to train local models, therefore, the convergence speed of the global model is usually slow in existing synchronous algorithms. In this work, we consider a large-scale system with device heterogeneity. To reduce training time and fully utilize edge node computation capacity, we propose an asynchronous algorithm in a novel multi-cluster decentralized federated edge learning (MD-FEEL) framework, where there are many clusters and each cluster consists of some clients. Our proposed asynchronous MD-FEEL contains four steps, i.e., local stochastic gradient descent (SGD) update, gradient consensus, intra-cluster model aggregation and inter-cluster model aggregation. To measure the staleness of cluster model, we introduce age of update (AoU) in inter-cluster aggregation stage and theoretically prove the convergence of our proposed algorithm on a non-convex setting. We evaluate our asynchronous MD-FEEL on MNIST and CIFAR-10 datasets and the simulation results show it can aggregate to a global model with better accuracy performance and faster convergence speed than some existing synchronous algorithms.

Owing to the excellent mobility character, unmanned aerial vehicles (UAVs) show great potentials as a means of data collector in a wireless sensor network (WSN). However, limitations in communication resources require UAVs to collect data from large-scale WSNs with high spectral efficiency. In this work, we investigate a UAV-assisted data collection strategy for a WSN enabled by hybrid NOMA. Specifically, the WSN is organized into clusters and each cluster is further subdivided into multiple groups. The sensor nodes in each group transmit data to the UAV using NOMA scheme. To ensure fairness between sensor nodes, we aim to maximize the minimum throughput by jointly optimizing the sensor node pairing, decoding order, transmit power and the UAV waypoints, subject to transmit power and UAV trajectory length constraints. To address the formulated mixed-integer nonlinear programming problem, we first propose an optimization-based algorithm by applying block coordinate descent method where a closed-form solution of the transmit power is derived, and then design a low-complexity heuristic algorithm. In addition, we present an initialization scheme for the proposed algorithms and an implementation strategy for our system. Extensive simulations demonstrate that our proposed UAV-assisted data collection scheme with hybrid NOMA can enhance system performance efficiently.

The coexistence of reconfigurable intelligent surface (RIS) and dual-functional radar-communication base station (DFBS) in the same system has attracted increasing discussions. In this paper, we focus on a RIS-assisted target sensing and multi-user communications system. To improve the radar performance, we formulate a radar detection signal-to-noise ratio (SNR) maximization problem, by jointly considering the transmit beamforming and RIS phase vector design, while meeting predefined communication quality-of-service, transmit power budget, and phase-shift constraints. Since the variables are deeply coupled, we first utilize the penalty based method to decouple them. Then, we employ the block coordinate descent (BCD) method to split the problem into three sub-problems, and resort to semidefinite relaxation (SDR) and second order cone programming (SOCP) methods to solve subproblems. Finally, we illustrate numerically that the radar detection SNR is greatly enhanced via the proposed approach.

Industrial Internet of Things (IIoT) improving by leaps and bounds has brought new possibilities for industrial manufacturing. Meanwhile, it does bring some serious challenges with the increasing number of devices. In which massive connectivity and multiservice are two major challenges for IIoT, in order to address the two main issues, in this article, we jointly consider nonorthogonal multiple access (NOMA) and wireless network slicing scenario, where multiservice devices share the same communication resources. To connect devices as many as possible, we formulate the connectivity maximization problem with joint subcarrier association and power allocation as a mixed-integer nonlinear programming (MINLP) problem, under the constraints of limited communication resources. To solve the problem effectively, we first split the MINLP problem into two subproblems by introducing a power allocation weight. Then, we analyze the theoretical approach for a special case and propose the layered access (LA) algorithm for general cases. Furthermore, a bisection search (Bisearch) algorithm is devised to find out the optimal power allocation weight. Simulation results show that the proposed LA algorithm has better performance compared to other benchmark schemes.

Industrial Internet of Things (IIoT) is a technological revolution that is profoundly reshaping the visage of industry. Facing the explosively increasing number of multi-service devices, traditional industrial network technology is no longer applicable. The advent of the fifth-generation (5G) wireless networks brings unprecedented possibilities for deploying the anticipated IIoT. To address the two main issues, i.e., connection density and multi-service requirements, in 5G empowered IIoT, we consider the non-orthogonal network slicing in this work. In particular, we jointly utilize network slicing to incorporate different types of services and exploit non-orthogonal multiple access (NOMA) to enhance the connection density. We formulate the connectivity maximization problem with joint sub-carrier association and power allocation as a mixed-integer nonlinear programming (MINLP). To tackle the intractable MINLP, we first transform it into a mixed-integer linear programming (MILP) and then simplify the MILP into an integer linear programming (ILP) by developing a simple yet effective pairing guideline. In order to further reduce the computational complexity, we then propose the alternating selection best-effort pairing (AS-BEP) algorithm with low complexity to solve the ILP effectively. Our analyses are supplemented by comprehensive simulation results that illustrate the performance superiority of the proposed algorithms to the benchmark schemes.

Learning-based mechanical fault diagnosis (FD) methods have been widely investigated in recent years. To overcome the shortages of centralized learning techniques from the perspective of data privacy and high communication overhead, federated learning (FL) is emerging as a promising method for FD. However, a large number of labeled fault data is required for the FL technique, which is not accessible in real-world industrial Internet-of-Things (IIoT) scenarios. To address the data scarcity challenge (i.e., few-shot), we propose a collaborative learning method that incorporates meta-learning into the federated learning framework. Specifically, our approach learns an effectively global meta-learner, which can quickly adapt to a new machine or a newly encountered fault category with just a few labeled examples and training iterations. Further, we theoretically analyze the convergence of the proposed algorithm in a non-convex setting. We conduct an extensive empirical evaluation of two real-world fault diagnosis datasets and they demonstrate that our proposed method achieves significantly faster convergence and higher accuracy, compared with the existing approaches.

Making a timely and precise scheduling in Industrial Internet of Things (IIoT) is fundamental and critical. Recently, Age of Incorrect Information (AoII) is proposed and utilized to measure the timeliness and accuracy of monitoring. In this work, we investigate a multi-sensor update system and leverage AoII to quantify the information freshness. Our goal is to obtain an optimal scheduling policy to minimize the system-wide cost. We first model the source statuses monitored by sensors as Markov chains and the scheduling problem as a Markov decision process (MDP). Due to the heterogeneity of source statuses in IIoT, it is prohibitive to solve the formulated MDP problem by conventional methods. To this end, we make use of a deep reinforcement learning (DRL) algorithm to solve this scheduling problem. Extensive numerical results verify the effectiveness of the adopted DRL algorithm. In addition, comparing to the conventional Age of Information (AoI) oriented method, we find that the AoII oriented method is much more effective, from the perspective of system-wide cost.

Owing to the rapid development of unmanned aerial vehicles (UAVs), UAV-enabled data collection has emerged as a promising technology. However, for the scenarios where the UAVs are dispatched to gather surrounding information actively and dynamically, the existing data collection methods can hardly fulfill the corresponding demands. To this end, we aim to study the paradigm data acquisition, where the UAV dynamically gathers information by on-board sensors. In this paper, we consider the scenario where a UAV acquires data in real time, which needs to be timely processed with the assistance of a mobile edge computing server. To address the real-time data acquisition characteristics, we construct a novel data acquisition rate model, which is with respect to the UAV speed. Next, we formulate a UAV energy consumption minimization problem that jointly considers UAV trajectory, transmission power, and CPU frequency. To tackle the highly complex problem, we propose an efficient iterative algorithm, and rigorously derive a closed-form solution for the UAV transmission power and computation resources allocation. With the obtained numerical results, we further validate the superiority of the proposed system design and the effectiveness of our algorithm against benchmark schemes.

This article studies the freshness of information with the aid of Age of Information (AoI) in the Industrial Internet of Things (IIoT), which plays a vital role to ensure quality and timely delivery of data services. To reduce the AoI, we leverage mobile-edge computing (MEC) to partially offload information to the mobile edge server. Aiming to cope with the packet error in the setting of short packet communication (SPC) in IIoT, we consider the standard automatic repeat request (ARQ) protocol with two policies, i.e., either retransmitting an out-of-date packet (RO) or transmitting a freshest packet (TF), when a packet error occurs. We derive the closed form of average AoI under these two policies, respectively, and then formulate the average AoI minimization problem by jointly optimizing the short packet blocklength and MEC offloading ratio. Due to the nonconvexity nature of the problem, we tackle it by employing block coordinate descent (BCD) and successive convex approximation (SCA) methods and then prove their convergence. Our extensive numerical results show that the optimal average AoI yielded by our proposed approach is almost identical to that from the high-complexity exhaustive search method, and has significant improvement over the benchmark methods. From the AoI perspective, it is revealed that the optimal strategy tends to offload all information to mobile edge server when the computing capacity of local device is less than a threshold. Furthermore, it is found that the RO policy is suitable for the relatively small bandwidth and large local computing capability scenario, whilst the TF policy is better for the large bandwidth and small local computing capability case.

The emergence of 5G brings unprecedented possibilities for deploying the anticipated Industrial Internet of Things (IIoT). To achieve high density connectivity with multiple services in 5G empowered IIoT, we consider the non-orthogonal network slicing in this work. In particular, we jointly utilize network slicing to incorporate two different types of services and exploit non-orthogonal multiple access (NOMA) to maximize the number of total devices that can be accessed to the system. We formulate the connectivity maximization problem as a mixed-integer nonlinear programming (MINLP) by jointly optimizing the transmit power and device-subcarrier association. To tackle the intractable MINLP, we first transform it into a mixed-integer linear programming (MILP) and then reduce the MILP by devising a simple but effective transmit power allocation scheme. Thereafter, we propose a low-complexity best-effort pairing (BEP) algorithm to solve the reduced MILP. By comprehensive simulations, we find that our proposed BEP significantly outperforms the benchmark schemes.

In recent years, the rapid development of mobile edge computing (MEC) provides an efficient execution platform at the edge for Internet-of-Things (IoT) applications. Nevertheless, the MEC also provides optimal resources to different microservices, however, underlying network conditions and infrastructures inherently affect the execution process in MEC. Therefore, in the presence of varying network conditions, it is necessary to optimally execute the available task of end users while maximizing the energy efficiency in edge platform and we also need to provide fair Quality-of-Service (QoS). On the other hand, it is necessary to schedule the microservices dynamically to minimize the total network delay and network price. Thus, in this article, unlike most of the existing works, we propose a dynamic microservice scheduling scheme for MEC. We design the microservice scheduling framework mathematically and also discuss the computational complexity of the scheduling algorithm. Extensive simulation results show that the microservice scheduling framework significantly improves the performance metrics in terms of total network delay, average price, satisfaction level, energy consumption rate (ECR), failure rate, and network throughput over other existing baselines.

The fifth generation (5G) wireless systems aims to differentiate its services based on different application scenarios. Instead of constructing different physical networks to support each application, radio access network (RAN) slicing is deemed as a prospective solution to help operate multiple logical separated wireless networks in a single physical network. In this paper, we incorporate two typical 5G services, i.e., enhanced Mobile BroadBand (eMBB) and Ultra-Reliable Low-Latency Communications (URLLC), in a cloud RAN (C-RAN), which is suitable for RAN slicing due to its high flexibility. In particular, for eMBB, we make use of multicasting to improve the throughput, and for URLLC, we leverage finite blocklength capacity to capture the delay accurately. Our objective is to minimize the total power consumption, subject to the limited physical resource constraints. We formulate the problem as a nonconvex optimization problem and exploit efficient approaches to solve it, such as successive convex approximation and semidefinite relaxation. Simulation results show that our proposed algorithm saves system power consumption significantly.

The fifth generation (5G) wireless system aims to differentiate its services based on different application scenarios. Instead of constructing different physical networks to support each application, radio access network (RAN) slicing is deemed as a prospective solution to help operate multiple logical separated wireless networks in a single physical network. In this paper, we incorporate two typical 5G services, i.e., enhanced Mobile BroadBand (eMBB) and ultra-reliable low-latency communications (URLLC), in a cloud RAN (C-RAN), which is suitable for RAN slicing due to its high flexibility. In particular, for eMBB, we make use of multicasting to improve the throughput, and for URLLC, we leverage the finite blocklength capacity to capture the delay accurately. We envision that there will be many slice requests for each of these two services. Accepting a slice request means a certain amount of revenue (consists of long-term revenue and shot-term revenue) is earned by the C-RAN operator. Our objective is to maximize the C-RAN operator's revenue by properly admitting the slice requests, subject to the limited physical resource constraints. We formulate the revenue maximization problem as a mixed-integer nonlinear programming and exploit efficient approaches to solve it, such as successive convex approximation and semidefinite relaxation. Simulation results show that our proposed algorithm significantly saves system power consumption and receives the near-optimal revenue with an acceptable time complexity.

Recently, cloud radio access network (C-RAN) with caching as a service (CaaS) was proposed to merge the functionalities of communication, computing, and caching (CC&C) together. In this paper, we dissect the interactions of CC&C in C-RAN with CaaS from two dimensions: physical resource dimension and time dimension. In the physical resource dimension, we identify how to segment the baseband unit (BBU) pool resources (i.e., computation and storage) into different types of virtual machines (VMs). In the time dimension, we address how the long-term resource segmentation in the BBU pool impacts on the short-term transmit beamforming at the remote radio heads. We formulate the problem as a stochastic mixed-integer nonlinear programming (SMINLP) to minimize the system cost, including the server cost, VM cost and wireless transmission cost. After a series of approximation, including sample average approximation, successive convex approximation, and semidefinite relaxation, the SMINLP is approximated as a global consensus problem. The alternating direction method of multipliers (ADMM) is utilized to obtain the solution in a parallel fashion. Simulation results verify the convergence of our proposed algorithm, and also confirm that the proposed scheme is more cost-saving than that without considering the integration of CC&C.

Network slicing is a fundamental architectural technology for the fifth generation mobile network. It is challenging to design a robust end-to-end network slice spanning overall networks, where a slice is constituted by a set of virtual network functions (VNFs) and links. Bugs may accidentally occur in some VNFs, invalidating some slices, and triggering slice recovery processes. Besides, the traffic demands in each slice can be stochastic, and drastic changes of traffic demands may trigger slice reconfiguration. In this paper, we investigate robust network slicing mechanisms by addressing the slice recovery and reconfiguration in a unified framework. We first develop an optimal slice recovery mechanism for deterministic traffic demands. This optimal solution is used as a benchmark for evaluating other robust slicing algorithms. Then, we design an optimal joint slice recovery and reconfiguration algorithm for stochastic traffic demands by exploiting robust optimization. To tackle the slow convergence issue in the robust optimization algorithm, we propose a heuristic algorithm based on variable neighborhood search. Numerical results reveal that our proposed robust network slicing algorithms can provide adjustable tolerance of traffic uncertainties compared with the deterministic algorithm.

Supporting cooperative radio among remote radio heads (RRHs) and elastic cloud service in the baseband unit (BBU) pool, cloud radio access network (C-RAN) is perceived as a promising solution for the next mobile network. In C-RAN, cooperative radio can enhance power saving at RRHs, along with impact on computation effort and power saving in the BBU pool. In turn, power saving at RRHs, benefited from cooperative radio is restricted by the constrained computation capacity provisioned by processors in the BBU pool. Besides, limited fronthauls, which support baseband signal transfer between the BBU pool and RRHs, affect the cooperative radio design and power consumption at RRH as well. By jointly optimizing transmit beamforming among RRHs and processor sleeping in the BBU pool, we exploit such interplay between communication and computation for system power minimization in C-RAN with limited fronthaul and computation capacity. Specifically, we formulate this problem as a mixed-integer non-linear programming problem (MINLP), and then leverage the special structure of the MINLP to make a near optimal decision on the set of active processors and transmit beamforming vectors with high efficiency. Finally, extensive numerical results demonstrate that our proposed algorithms can enforce processor sleeping and reduce system power consumption significantly.

Network slicing is an emerging terminology that enables operators to partition a shared substrate network into multiple logically isolated and on- demand virtual networks to support diverse communication cases. However, the performance of network slices can be heavily deteriorated due to unexpected software or hardware malfunctions in the substrate network. Furthermore, the traffic demand is usually considered as a deterministic parameter during the network slice deployment, while it could be stochastically varied in reality, and such stochasticality may invalidate some network slices. Therefore, it is imperative to develop a robust network slicing algorithm. In this paper, we first formulate the failure recovery problem of network slicing as a mixed integer programming (MIP) and then model the robust MIP (RMIP) to capture the stochastic traffic demand. We solve the RMIP by using the robust optimization approach. Numerical results reveal that the proposed the robust network slicing algorithm can provide adjustable tolerance of traffic uncertainty in comparison with the nonrobust algorithm. In the meanwhile, the trade-off between robustness of requests and the load of substrate links can be hence efficiently managed and controlled.

Cloud radio access network (C-RAN) is perceived as an energy-efficient solution for the next mobile network. The cloud-based baseband unit (BBU) pool is capable of dynamically provisioning computational resources for mobile users to improve hardware utilization such that unused processors can be switched off for power saving in the BBU pool. Besides, cooperative radio among remote radio heads (RRHs) can optimize transmit beamforming to reduce power consumption at RRHs. Thus, to achieve more judicious system power saving, this is need to consider power consumption in the BBU pool and that at RRHs together. In this paper, we aim to minimize system power consumption by jointly exploiting transmit beamforming and processor sleeping. Specifically, we formulate a mixed integer non-linear system power minimization problem, which is hard to solve. For tractability purpose, we transform this problem to an equivalent clustering problem embedded with a series of transmit beamforming problems and processor sleeping problems. On this basis, we first focus on solving the embedded problems with the given clustering and then propose a low-complexity clustering algorithm to search out the optimal clustering with minimum system power consumption. Finally, simulation results show that our proposed algorithms can save system power significantly.

The energy consumption and applications' execution latency of mobile devices (MDs) can be improved by migrating application tasks to a nearby edge device. In this paper, we propose an optimization framework to investigate the scenario when a MD can offload tasks to multiple access points (APs) and scale its central process unit (CPU) frequency. Firstly, the optimal solution is derived from an exhaustive search based approach; and then a semidefinite relaxation (SDR) based approach is proposed to efficiently solve the problem. The obtained results from our simulation indicate that the SDR-based algorithm is able to achieve close-to-optimal performance. We also show that our proposed scheme can reduce the MD's energy consumption and tasks' execution latency, by taking advantage of having multiple APs and flexible CPU frequency.

In this paper, we propose an optimization framework of offloading from a single mobile device (MD) to multiple edge devices. We aim to minimize both total tasks' execution latency and the MD's energy consumption by jointly optimizing the task allocation decision and the MD's central process unit (CPU) frequency. This paper considers two cases for the MD, i.e., fixed CPU frequency and elastic CPU frequency. Since these problems are NP-hard, we propose a linear relaxation-based approach and a semidefinite relaxation (SDR)-based approach for the fixed CPU frequency case, and an exhaustive search-based approach and an SDR-based approach for the elastic CPU frequency case. Our simulation results show that the SDR-based algorithms achieve near optimal performance. Performance improvement can be obtained with the proposed scheme in terms of energy consumption and tasks' execution latency when multiple edge devices and elastic CPU frequency are considered. Finally, we show that the MD's flexible CPU range can have an impact on the task allocation.

Cloud radio access network (C-RAN) is emerging as a potential alternative for the next generation RAN by merging RAN and cloud computing together. In this paper, we consider the baseband unit (BBU) pool of C-RAN as a collection of virtual machines (VMs). We allow each user equipment (UE) to associate with multiple VMs in the BBU pool, and each remote radio head (RRH) can only serve a limited number of UEs. Under this model, we jointly optimize the VM activation in the BBU pool and sparse beamforming in the coordinated RRH cluster, which is constrained by limited fronthaul capacity, to minimize the system cost of C-RAN. We formulate this problem as a mixed-integer nonlinear programming problem, and then propose efficient methods to optimize the number of active VMs, as well as the sparse beamforming vectors. Moreover, we derive a closed-form solution for the beamforming vectors. Simulation results suggest that our proposed algorithms have better performance than the benchmark algorithms in terms of both system cost and robustness.

Cloud radio access network (C-RAN), regarded as a promising green network architecture, facilitates cooperative transmission among remote radio heads (RRHs) while enabling flexible computation provisioning in the virtualized baseband unit pool. By jointly optimizing cooperative transmission, i.e., transmit power allocation with zero-forcing precoding adopted, and computation provisioning, i.e., virtual machine assignment, this paper minimizes the system power consumption comprised of transmit power and processing power in downlink C-RAN. Specifically, subject to per-RRH power constraint (PRPC) and per-MU quality of service constraint, the system power consumption minimization problem is formulated as a mixed integer nonlinear programming (MINLP) problem. To solve the challenging MINLP, we reformulate the MINLP as a minimum weight perfect matching problem to get the initial solution without considering the PRPC. On this basis, a power-aware greedy algorithm is further devised to modify the solution such that the PRPC is satisfied. Finally, extensive simulations show the superiority of the proposed scheme on system power saving and the tradeoff between transmit power and processing power.

By migrating baseband processing functionalities into a centralized cloud-based baseband unit (BBU) pool, cloud radio access network (C-RAN) facilitates cooperative transmission among remote radio heads (RRHs) and enables flexible computation provisioning in the BBU pool. In C-RAN, due to the high amount of data transfer from the BBU pool to RRHs through fronthauls, limited fronthaul capacity becomes a key factor when designing cooperative transmission schemes among RRHs. Meanwhile, as computational resources are provisioned to mobile users (MUs) for baseband processing in the form of virtual machines (VMs) in the BBU pool, an effective VM assignment strategy is also with great significance. In this paper, we propose a holistic framework for green C-RAN under the constraint of limited fronthaul capacity, where we jointly optimize hybrid clustering and computation provisioning to appropriately provide a cluster of RRHs and a VM to each MU for cooperative transmission and baseband processing, aiming at minimizing the system power consumption. The system power minimization problem is formulated as an integer non-linear programming problem, which is hard to tackle. For tractability purpose, we transform this problem to an equivalent hybrid clustering problem embedded with a series of VM assignment problems. On this basis, we first achieve the optimal solution for system power minimization with high computational complexity, and then, a greedy algorithm is proposed to solve the hybrid clustering problem for practical implementation. Finally, the simulation results demonstrate that the proposed joint optimization of hybrid clustering and computation provisioning can significantly reduce the system power consumption.

Cloud radio access network (C-RAN) aims to improve spectrum and energy efficiency of wireless networks by migrating conventional distributed base station functionalities into a centralized cloud baseband unit (BBU) pool. We propose and investigate a cross-layer resource allocation model for C-RAN to minimize the overall system power consumption in the BBU pool, fiber links and the remote radio heads (RRHs). We characterize the cross-layer resource allocation problem as a mixed-integer nonlinear programming (MINLP), which jointly considers elastic service scaling, RRH selection, and joint beamforming. The MINLP is however a combinatorial optimization problem and NP-hard. We relax the original MINLP problem into an extended sum-utility maximization (ESUM) problem, and propose two different solution approaches. We also propose a low-complexity Shaping-and-Pruning (SP) algorithm to obtain a sparse solution for the active RRH set. Simulation results suggest that the average sparsity of the solution given by our SP algorithm is close to that obtained by a recently proposed greedy selection algorithm, which has higher computational complexity. Furthermore, our proposed cross-layer resource allocation is more energy efficient than the greedy selection and successive selection algorithms.

We consider the problem of optimally redirecting user requests in a cloud-centric media network (CCMN) to multiple destination Virtual Machines (VMs), which elastically scale their service capacities in order to minimize a cost function that includes service response times, computing costs, and routing costs. We also allow the request arrival process to switch between normal and flash crowd modes to model user requests to a CCMN. We quantify the trade-offs in flash crowd detection delay and false alarm frequency, request allocation rates, and service capacities at the VMs. We show that under each request arrival mode (normal or flash crowd), the optimal redirection policy can be found in terms of a price for each VM, which is a function of the VM's service cost, with requests redirected to VMs in order of nondecreasing prices, and no redirection to VMs with prices above a threshold price. Applying our proposed strategy to a YouTube request trace data set shows that our strategy outperforms various benchmark strategies. We also present simulation results when various arrival traffic characteristics are varied, which again suggest that our proposed strategy performs well under these conditions.

According to the new features of traffic distribution, a novel forwarding migration algorithm (FMA) is proposed. Programmable routers (PR) are used in FMA. One PR sets FMA bit for unbalanced traffic. Another PR forwards related traffic to the FMA link rather than to wait for matching the longest address prefix. Theoretic analysis shows that FMA can reduce the forwarding burden of core router. Simulation proves that FMA obtains a lower transmission delay and packet loss rate.