The Internet of Drones (IoD) integrates drone technology with the Internet of Things, enabling efficient data collection and communication applications. Federated learning (FL) in IoD networks facilitates collaborative model training while preserving data privacy but imposes significant computational and communication demands on resource-constrained drones. Federated knowledge distillation learning (FedKD) addresses this challenge by training both a large teacher model and a smaller student model locally but only updating the smaller student model, thereby reducing communication overhead. This article tackles the resource allocation problem in FedKD within IoD networks, focusing on optimizing CPU computing resource, wireless transmission power, and bandwidth allocation to minimize overall drone energy consumption. We formulate this as an optimization problem, considering constraints on latency, computing resource, bandwidth, and power. To effectively address this problem, we design a low-complexity algorithm. Extensive simulations validate our approach, showing it reduces energy consumption by an average of 85% compared to FedKD and 94% compared to FedAvg (a standard FL algorithm).

Generative Adversarial Networks (GANs) are deep learning models that learn and generate new samples similar to existing ones. Traditionally, GANs are trained in centralized data centers, raising data privacy concerns due to the need for clients to upload their data. To address this, Federated Learning (FL) integrates with GANs, allowing collaborative training without sharing local data. However, this integration is complex because GANs involve two interdependent models—the generator and the discriminator—while FL typically handles a single model over distributed datasets. In this article, we propose a novel asynchronous FL framework for GANs, called AsyncFedGAN, designed to efficiently and distributively train both models tailored for molecule generation. AsyncFedGAN addresses the challenges of training interactive models, resolves the straggler issue in synchronous FL, reduces model staleness in asynchronous FL, and lowers client energy consumption. Our extensive simulations for molecular discovery show that AsyncFedGAN achieves convergence with proper settings, outperforms baseline methods, and balances model performance with client energy usage.

In wireless Internet of Things (IoT) networks, Federated Reinforcement Learning (FRL) has emerged as a decentralized strategy for data-driven decision-making, enabling devices to learn directly from real-time environmental interactions, sidestepping the need for labeled data. This method promises enhanced data privacy and finds practical applications in autonomous driving, smart grids, and industrial automation. However, the integrity of FRL can be compromised by malicious clients injecting false data, underlining the need for a fault-tolerant mechanism to sustain the robustness and accuracy of the learning phase. Moreover, the inherent client heterogeneity within IoT networks propels the demand for judicious client selection, optimizing computational and communication resources. This paper investigates client selection problem within a fault-tolerant FRL framework for wireless IoT networks. Our objective is to explore the tradeoff between maximizing client participation and minimizing energy consumption of IoT devices. We formulate our problem as a mixed-integer linear programming (MILP) model and design an efficient algorithm with low computational complexity to address it. Extensive simulations are conducted to demonstrate the superiority of our proposed algorithm.

Recent advances in deep learning have accelerated its use in various applications, such as cellular image analysis and molecular discovery. In molecular discovery, a generative adversarial network (GAN), which comprises a discriminator to distinguish generated molecules from existing molecules and a generator to generate new molecules, is one of the premier technologies due to its ability to learn from a large molecular data set efficiently and generate novel molecules that preserve similar properties. However, different pharmaceutical companies may be unwilling or unable to share their local data sets due to the geo-distributed and sensitive nature of molecular data sets, making it impossible to train GANs in a centralized manner. In this paper, we propose aGraphconvolutional network inGenerativeAdversarialNetworks viaFederated learning (GraphGANFed) framework, which integrates graph convolutional neural Network (GCN), GAN, and federated learning (FL) as a whole system to generate novel molecules without sharing local data sets. In GraphGANFed, the discriminator is implemented as a GCN to better capture features from molecules represented as molecular graphs, and FL is used to train both the discriminator and generator in a distributive manner to preserve data privacy. Extensive simulations are conducted based on the three benchmark data sets to demonstrate the feasibility and effectiveness of GraphGANFed. The molecules generated by GraphGANFed can achieve high novelty$(\approx 100 )$and diversity$(\gt 0.9)$. The simulation results also indicate that 1) a lower complexity discriminator model can better avoid mode collapse for a smaller data set, 2) there is a tradeoff among different evaluation metrics, and 3) having the right dropout ratio of the generator and discriminator can avoid mode collapse.

Internet of drones (IoD), where drones act as the Internet of things (IoT) devices, has attracted much attention for its application in traffic surveillance, object tracking, and disaster rescue. These applications rely heavily on machine learning (ML) techniques for image classification. The traditional method of training ML models in IoD networks involves sending all data to a centralized ground base station (BS), which can result in privacy and security concerns. Split learning, an approach that separates the training model into a client model and a server model, can mitigate these concerns by allowing the client model to reside in the drones while the server model is in the BS, preserving data privacy without sharing raw data. In this study, we explore the application of split learning in IoD networks and conduct simulations to evaluate our designed algorithm. Simulation results demonstrate that different separations do not significantly impact split learning accuracy, increasing the number of layers on clients can lead to longer training times, communication overhead is a significant bottleneck in split learning for IoD networks, client numbers do not significantly affect accuracy, and training time slightly increases with an increasing number of clients.

Internet of drones (IoD), where drones act as the Internet of things (IoT) devices, makes IoT networks much more flexible and responsive because of high mobility of drones. Machine learning (ML) techniques can be applied in IoD to facilitate multiple applications such as object tracking and traffic surveillance, where ML data samples are collected and analyzed in the edge servers at the ground base station (BS). However, aggregating all data samples incurs huge wireless network traffic and potential data privacy leakage. Federated learning (FL) is then proposed to address these challenges by performing local training in drones and aggregating model parameters at the BS without sharing raw data samples. The FL performance in IoD networks is greatly affected by limited drone batteries which power FL local training, wireless data transmission, and drones’ movements. This paper hence investigates the energy-efficient FL in IoD networks to optimize CPU frequencies of drones’ on-board computing units such that total energy consumption of all the drones in the FL process can be minimized, while satisfying the FL training time requirement. We formulate the problem as a non-linear programming problem and then design an algorithm with polynomial time complexity to derive the optimum solution. Extensive simulations are conducted to demonstrate the performance of our proposed algorithm.

To reap the benefit of big data generated by the massive number of Internet of Things (IoT) devices while preserving data privacy, federated learning (FL) has been proposed to enable IoT devices to train machine learning models locally. That is, instead of sharing the local data sets, different clients in terms of IoT devices only need to upload their local models to a centralized FL server. Client selection in FL is critical to maximize the number of qualified clients, who can successfully upload their local models to the FL server before the predefined deadline. Normally, client selection is coupled with wireless resource management owing to the fact that different clients need to share the same spectrum to upload their local models. The existing solutions of joint optimizing client selection and resource management are designed based on frequency-division multiple access (FDMA) or time-division multiple access (TDMA), which do not consider the dynamics of the clients and lead to low bandwidth utilization. In this article, we propose the Nonorthogonal-Multiple-Access (NOMA)-based resource allocation for client selection in FL to dynamically and jointly optimize client selection for each global iteration as well as the transmission power of each selected client in each time slot within a global iteration. We design the deep-reinforcement-learn-based client selection in NOMA-based federated learning (DREAM-FL) algorithm to solve the problem. Extensive simulations are conducted to demonstrate that DREAM-FL can select more qualified clients and has higher model accuracy than FDMA and TDMA-based solutions.

Internet of Drones (IoD), where drones act as the Internet of Things (IoT) devices, provides multiple services, such as object recognition, traffic monitoring, and disaster rescue. The service response time is limited by drones’ onboard computing power, hence greatly affecting the Quality of Service (QoS). Machine learning [(ML), e.g., image processing] task offloading to the fog node attached to the ground base station can help reduce workload from drones and hence decrease service time. The communication latency between drones and the fog node during task offloading also affects the service time and hence the communication efficiency needs to be considered. Therefore, in this work, we consider the joint optimization of ML task offloading and power control in IoD to determine each drone’s number of offloaded images and wireless transmission power. We analyze the energy model of the commonly used convolutional neural network (CNN) for object recognition, and formulate the joint optimization problem as a mixed-integer nonlinear programming (MINLP) problem to minimize drones’ average service time constrained by drones’ energy budgets. An approximation algorithm with low computational complexity is then designed to address the problem and its performances are compared with the lower bounds and demonstrated via extensive simulations.

The sensing service of Internet-of-Things (IoT) networks enables IoT sensors to sense the environment information (e.g., temperature and traffic conditions) and send them through the IoT gateway to the users who request those information. The explosive growth of IoT users and sensors injects massive traffic into IoT networks and easily depletes the battery of IoT sensors. Caching at the IoT gateway is hence a promising solution to mitigate this problem by storing popular IoT data at the IoT gateway and sending them directly to the users instead of activating IoT sensors to transmit the data. In our work, we investigate the content placement problem, which determines data to be cached at each time epoch in dynamic IoT networks with the objective to minimize the average data transmission delay constrained by the cache storage capacity and IoT data freshness. We formulate our problem as an integer linear programming (ILP) problem and then model it as a Markov decision process (MDP). A deep reinforcement learning algorithm is proposed to solve this problem and its performances are demonstrated via extensive simulations.

Machine learning models have been built in fog nodes in fog-aided Internet-of-Things (IoT) networks to provision future events prediction and image classification by training data collected from IoT devices. However, sending massive data from all devices to a fog node incurs huge network traffic in wireless links in between. Federated learning is proposed to address the challenge by training models locally in IoT devices and only sharing model parameters in the fog node. In this article, we investigate both the CPU frequency control and wireless transmission power control of all IoT devices to balance the tradeoff between the device energy consumption and federated learning time (consisting of both the computation and communication latencies) in fog-aided IoT networks. We formulate the joint optimization of CPU and power control as a nonlinear programming (NLP) problem with the objective to minimize the energy consumption of all IoT devices constrained by the federated learning time requirement. An alternative direction algorithm, which alternatively optimizes the CPU frequency and wireless transmission power until convergence, is hence designed to solve this problem and its performance is demonstrated via extensive simulations.

Internet of Drones (IoD) employs drones as the internet of things (IoT) devices to provision applications such as traffic surveillance and object tracking. Data collection service is a typical application where multiple drones are deployed to collect information from the ground and send them to the IoT gateway for further processing. The performance of IoD networks is constrained by drones' battery capacities, and hence we utilize both energy harvesting technologies and power control to address this limitation. Specifically, we optimize drones' wireless transmission power at each time epoch in energy harvesting aided time-varying IoD networks for the data collection service with the objective to minimize the average system energy cost. We then formulate a Markov Decision Process (MDP) model to characterize the power control process in dynamic IoD networks, which is then solved by our proposed model-free deep actor-critic reinforcement learning algorithm. The performance of our algorithm is demonstrated via extensive simulations.

The new generation video compression standard of high efficiency video coding (HEVC) improves compression performance by 50% compared with the previous video coding standard, while also increases the coding complexity greatly. In order to reduce the complexity and further improve system efficiency, a fast intra prediction algorithm with simplified prediction modes is proposed. Based on the correlation between rough mode decision (RMD) and the best prediction mode in the intra prediction mode selection, and the utilization rates of 35 intra prediction modes, the proposed algorithm removes some prediction modes in the candidate list and reduces the computation time of the rate distortion (RD) cost. Experimental results show that compared with HEVC intra prediction coding algorithm, the proposed fast intra prediction algorithm increases efficiency by about 20% with no significant changes for compression bit rates.

Internet of drones (IoD), employing drones as the internet of things (IoT) devices, brings flexibility to IoT networks and has been used to provision several applications (e.g., object tracking and traffic surveillance). The explosive growth of users and IoD applications injects massive traffic into IoD networks, hence causing congestions and reducing the quality of service (QoS). In order to improve the QoS, caching at IoD gateways is a promising solution which stores popular IoD data and sends them directly to the users instead of activating drones to transmit the data; this reduces the traffic in IoD networks. In order to fully utilize the storage-limited caches, appropriate content placement decisions should be made to determine which data should be cached. On the other hand, appropriate drone association strategies, which determine the serving IoD gateway for each drone, help distribute the network traffic properly and hence improve the QoS. In our work, we consider a joint optimization of drone association and content placement problem aimed at maximizing the average data transfer rate. This problem is formulated as an integer linear programming (ILP) problem. We then design the Drone Association and Content Placement (DACP) algorithm to solve this problem with low computational complexity. Extensive simulations demonstrate the performance of DACP.

Unmanned aerial vehicles (UAVs) have attracted attention from both the academic and industry because of its highly controllable mobility. The UAV has hence become a potential alternative for a large amount of geographically distributed sensors in provisioning sensing service where the information of different locations (e.g., temperature, humidity, pollutant level and traffic condition) are sensed and sent to the ground station (GS). However, the UAV on-board battery is usually limited due to the size and weight constraints, and greatly affects the UAV performance. Practically, the UAV usually needs to return to the GS for recharging before the battery exhaustion. The trajectory routes, therefore, should be well designed to meet the battery capacity constraint and improve the quality of service (QoS). In this paper, we investigate the trajectory optimization of rechargeable UAV for sensing service to minimize the task completion latency. We formulate this problem as a mixed integer linear programming (MILP) model. A Clone Searching Algorithm (CSA), which clones the rechargeable UAV into several non-rechargeable virtual UAVs and simultaneously search trajectory routes for each virtual UAV, is then designed to reduce the computational complexity of MILP. Numerical results demonstrate the performance of our proposed algorithm.

Fog-aided Internet of Things (IoT) networks provide low latency IoT services by offloading computational intensive and delay sensitive tasks to the fog nodes, which are deployed close to the IoT devices. Mobile IoT relies on battery limited mobile IoT devices (e.g., wearable devices and smartphones) to provision networks with enhanced flexibility. Mobile IoT faces the challenges of varying wireless channel conditions and hence may degrade the quality of service (QoS). We investigate the task allocation, which intelligently distributes tasks to different fog nodes and adapts to IoT varying mobile environment, such that the average task completion latency, constrained by QoS requirements and mobile IoT device battery capacity, is minimized. An integer linear programming (ILP) problem is then formulated to solve this problem. However, it is difficult to obtain the user mobility patterns (i.e., future locations where tasks are offloaded) and user side information (i.e., task length and computing intensity). Therefore, we propose an online learning algorithm to engineer task allocation decisions and then demonstrate its performances by extensive simulations.

It is meaningful to use data mining technology to predict the type of lawsuit which a company may receive so that enterprises can avoid lawsuit risks. So we propose a corporation lawsuit prediction algorithm based on guiding learning and collaborative filtering recommendation. Firstly, we use the adaptive synthetic sampling approach (ADASYN) to generate more synthetic data for different minority classes according to their different level of difficulty in learning, so that the training would focus on these minority classes that are difficulty to learn and reduce the learning bias introduced by the imbalance of data distribution. Secondly, for the sake of solving the problem that the insufficient samples make it difficult for the model to learn enough knowledge resulting in a large fluctuation of final scores during the training and poor model stability, we use guiding learning to integrate the basic knowledge of all types of lawsuit a company may receive in the future obtained by the multi-label classification model into the training process of TOP-1 and TOP-2 predictive models. Finally, in order to further improve the prediction accuracy, we use the collaborative filtering recommendation algorithm (CFRA) to select the most similar sample with each test sample from the training set, and the lawsuit type of the selected sample is directly used as the predicted lawsuit type of the corresponding test sample, thereby improving the total prediction accuracy. The experimental results show that the proposed algorithm can effectively predict the most probable lawsuit types of the Top2 for corporations.

Internet of People (IoP), which focuses on personal information collection by a wide range of the mobile applications, is the next frontier for Internet of Things. Nowadays, people become more and more dependent on the Internet, increasingly receiving and sending information on social networks (e.g., Twitter, etc.); thus social networks play a decisive role in IoP. Therefore, community discovery has emerged as one of the most challenging problems in social networks analysis. To this end, many algorithms have been proposed to detect communities in static networks. However, microblogging social networks are extremely dynamic in both content distribution and topological structure. In this paper, we propose a model for efficient evolutionary user interest community discovery which employs a nature-inspired genetic algorithm to improve the quality of community discovery. Specifically, a preprocessing method based on hypertext induced topic search improves the quality of initial users and posts, and a label propagation method is used to restrict the conditions of the mutation process to further improve the efficiency and effectiveness of user interest community detection. Finally, the experiments on the real datasets validate the effectiveness of the proposed model.

The Internet of Things (IoT) sensing service allows systems and users to monitor environment states by transmitting the content sensed by a variety of sensors. Owing to billions of sensors and devices deployed in the IoT system, a huge amount of data (big data) are generated, thus injecting tremendous traffic into the network. Cloud radio access network (C-RAN) is a promising wireless network architecture to accommodate the fast growing IoT traffic and improve the performance of IoT services. Caching in C-RAN, which brings content to the edges, not only alleviates the network traffic, thus improving the end-to-end user quality of service (QoS), but also avoids activating the sensors too frequently, thus reducing their energy consumption. The content placement problem determines what and where to cache in C-RAN. However, the caching performance is highly related to the caching storages. The storage allocation problem determines the storage capacities of network entities. In this paper, we jointly optimize the storage allocation problem and content placement problem in a hierarchical cache-enabled C-RAN architecture for IoT sensing service. We formulate the joint problem as an integer linear programming (ILP) model with the objective to minimize the total network traffic cost. The storage allocation problem and content placement problem are constrained by caching storage budgets and cache capacities, respectively. Two heuristic algorithms are proposed in order to reduce the computational complexity of ILP. Extensive simulations have been conducted to demonstrate that the performances of our proposed algorithms approximate the optimal solutions.

The Internet of Things (IoT) sensing service enables users to monitor the ambient environment by fetching data from IoT sensors. The explosive growth of mobile users and IoT applications injects massive traffic to the IoT network and also speeds up the drainage of sensor batteries. Caching at the IoT gateway (GW), which stores the IoT data and directly send them to the users, can avoid activating sensors too frequently, hence reducing the traffic in the IoT network as well as the energy consumption of sensors. To overcome the limited energy capacity of sensors, energy transmitters (ETs) are deployed to charge them. Practically, the GW and ETs may be owned by different operators, and the GW operator needs to incentivize ETs to provision the charging service. In this paper, we formulate a Stackelberg game in the cache-enabled energy harvesting aided IoT framework to improve the user quality of service. Caching strategies, incentive strategies, and ET transmission power strategies are jointly optimized to find the Stackelberg equilibrium by our proposed alternative direction approach. Simulation results elicit the benefits of our framework and demonstrate the performances of our proposed algorithm.

To provide a better quality of service (QoS), cloud computing paradigm in Internet of Things (IoT) networks has shifted toward the edge. Fog-aided IoT networks deploy fog nodes, which are equipped with computing and storage resources, at the network edge to take over the deadline-driven computing tasks from IoT devices. In the fog node, where multiple virtual machines (VMs) can be rented to process the tasks, fog provisioning is to determine which VM should be rented and how to distribute different tasks to VMs in order to minimize the system cost (i.e., VM rentals). On the other hand, VMs may fail and lead to QoS degradation. Hence, reliability of VMs should also be considered when addressing the fog resource provisioning problem. To improve reliability, more VMs should be rented to satisfy the QoS requirement; this leads to higher system cost. Therefore, there is a tradeoff between reliability and the system cost. In this paper, we investigate the tradeoff of maximizing the reliability and minimizing the system cost for fog resource provisioning in IoT networks. An integer linear programming (ILP) problem is formulated but suffers from a high computational complexity. We then design an alternative algorithm to achieve suboptimal solutions with better time efficiency. The simulation results demonstrate the performances of our proposed algorithm.

Fog-aided Internet of Things (IoT) addresses the resource limitations of IoT devices in terms of computing and energy capacities, and enables computational intensive and delay-sensitive tasks to be offloaded to the fog nodes attached to the IoT gateways. A fog node, utilizing the cloud technologies, can lease and release virtual machines (VMs) in an on-demand fashion. For the power-limited mobile IoT devices (e.g., wearable devices and smart phones), their quality of service may be degraded owing to the varying wireless channel conditions. Power control helps maintain the wireless transmission rate and hence the quality of service (QoS). The QoS (i.e., task completion time) is affected by both the fog processing and wireless transmission; it is thus important to jointly optimize fog resource provisioning (i.e., decisions on the number of VMs to rent) and power control. This paper addresses this joint optimization problem to minimize the system cost (VM rentals) while guaranteeing QoS requirements, formulated as a mixed integer nonlinear programming problem. An approximation algorithm is then proposed to solve the problem. Simulation results demonstrate the performance of our proposed algorithm.

Rapid growth of Internet of Things (IoT) services with different service level agreement (SLA) requirements has called for offloading computing tasks from the resource constrained IoT devices to the remote cloud. The growing number of IoT devices is exacerbating the links between IoT devices and the cloud. Furthermore, many services (e.g., health and disaster response) require fast service response and immediate data analytics, thus posing a great challenge to the remote cloud. Fog computing provides the solution by bringing computing resources to the edge of the network to assume substantial computing tasks. However, the resource failures during service processing should not be overlooked because they can greatly degrade service performance. In our work, we investigate the fog resource provisioning problem for the deadline-driven IoT services to minimize the system cost considering the probability of resource failures. We formulate the problem as an integer linear programming (ILP) model, and then design a Weighted Best Fit Decreasing (WBFD) algorithm with low computational complexity. Simulation results validate that our proposed algorithm performs close to the optimal solutions of ILP.

Caching popular contents closer to users has been proposed to alleviate wireless network traffic and improve user quality of experience (QoE). Decisions on where and what to cache is of great importance. In this paper, we propose a hierarchical cache-enabled cloud radio access network (C-RAN) architecture where joint caching is considered in both remote radio heads (RRHs) and baseband units (BBUs) with the constraints of backhaul and fronthaul links. We formulate the content placement problem as an integer linear programming (ILP) model with the objective of minimizing the average content download time. A heuristic algorithm is proposed in order to reduce the time complexity. Simulation results of the average download delay are analyzed from different aspects including caching locations, total file lengths, cache sizes and file popularities, and they demonstrate that the performance of the proposed popularity-based algorithm approximates ILP solutions closely but with high time efficiency.

We optimize the disaster backup in multi-datacenter (multi-DC) cloud systems and design disaster-aware algorithms to realize rapid backup with the objective of minimizing the backup window for all the DCs in the network. A mixed integer linear programming (MILP) model is first formulated to optimize the backup processes of all production DCs jointly. We then develop three heuristics that use the one-step or two-step approaches for the selection of backup DCs and the calculation of backup routing paths. Simulation results show that the Two-Step algorithm can achieve the shortest backup window with the lowest operation complexity.