By reducing the size of transmitted data between device-side and edge-side machine learning model parts, intermediate activation (IA) compression can alleviate communication overhead, lower latency, and conserve energy, thus enhancing the model partitioning in edge-device collaborative inference scenarios. However, existing studies lack refined resource allocation and fail to jointly optimize edge-device communication and computing resources, especially in terms of IA compression rates, leading to sub-optimal inference performance. To this end, we propose an adaptive model partitioning, intermediate activation compression, and resource allocation scheme for efficient edge-device collaborative inference. We jointly optimize the model partitioning point selection, IA compression rates control, computing resource allocation for both edge and devices, and device transmission power allocation. Our goal is to minimize the weighted sum of inference accuracy loss, inference latency, and device energy consumption. To solve the high-complexity optimization problem efficiently, we design a DRL-based algorithm, which decouples the problem into sub-problems firstly, and then employs an SD3 (Softmax Deep Double Deterministic Policy Gradients)-based DRL method to solve the partitioning point and IA compression rates sub-problem, and utilizes various numerical methods to solve the sub-problems of local and edge computing resource allocation and transmission power control. Extensive comparative simulations with four different schemes under different environmental parameters demonstrate the superiority and robustness of our approach.

With 6G communication technology advancing, the demand for Radio Environment Maps (REMs) has increased due to their critical role in network optimization, resource management, and signal coverage in complex environments. However, due to the high cost of sensing, only a small number of discrete radio sampling results can be obtained, which limits their application to specific tasks. To address this problem, we propose a novel method for constructing fine-grained REMs in complex environments based on a generative diffusion model. This method leverages the spatial correlations between sparse data points while incorporating conditional information based on global spatial correlations and geographic relationships. The goal is to construct fine-grained radio environment maps from sparse coarse-grained data. Experimental results demonstrate that our model performs significantly well, achieving an absolute error of only 2.79 dB even when the sampling rate is as low as 10%.

Aquatic information monitoring is crucial for the sustainable management of water environments. Conventional interpolation methods commonly hinge on assumptions of spatial proximity or temporal similarity. However, they often fall short of capturing the intricate spatiotemporal correlations present in water quality sequences, affecting our understanding of the spatial patterns of regional water quality conditions. In this study, we propose a framework for river basin information fine-grained sensing based on deep learning, which includes a global sensing model (SGM) and a static deployment model. Inside the SGM, we adopt a multidimensional convolutional neural network (CNN) to extract spatiotemporal features and an attention mechanism to fuse these features, to infer water quality variable information on unmonitored points. Since the inference outcomes could be affected by the locations of the sensors, to minimize the inference error of the SGM, the static deployment model was designed to aid the deployment of sensors into strategic locations of a river basin to obtain optimum spatial-temporal data samples. The research results not only revealed the spatial distribution patterns of total nitrogen (TN) concentrations but also showed that the proposed method could yield a better inference performance compared to traditional interpolation methods.

Hierarchical federated learning shows excellent potential for communication-computation trade-offs and reliable data privacy protection by introducing edge-cloud collaboration. Considering non-independent and identically distributed data distribution among devices and edges, this article aims to minimize the final loss function under time and energy budget constraints by optimizing the aggregation frequency and resource allocation jointly. Although there is no closed-form expression relating the final loss function to optimization variables, we divide the hierarchical federated learning process into multiple cloud intervals and analyze the convergence bound for each cloud interval. Then, we transform the initial problem into one that can be adaptively optimized in each cloud interval. We propose an adaptive hierarchical federated learning process, termed as AHFLP, where we determine edge and cloud aggregation frequency for each cloud interval based on estimated parameters, and then the CPU frequency of devices and wireless channel bandwidth allocation can be optimized in each edge. Simulations are conducted under different models, datasets and data distributions, and the results demonstrate the superiority of our proposed AHFLP compared with existing schemes.

Vehicular Edge Computing (VEC) offers a promising framework for providing vehicles with low-latency and highly reliable services. By leveraging the underutilized computational resources of parked and moving vehicles commonly found in urban areas, a VEC system can enhance the performance of surrounding user devices and alleviate the loads on its edge servers. In this study, a resource orchestration scheme is introduced for a multi-device, multi-vehicle, and multi-edge scenario. Tasks from a device can be offloaded to its associated edge server, a neighboring edge server, a parked vehicle, or a moving vehicle. Our goal is to achieve the total task processing cost (comprising task processing latency and energy consumption) minimization across all devices through making strategies for task offloading and computational and communication resource allocation. We decompose the optimization problem and propose a Twin Delayed Deep Deterministic Policy Gradient (TD3)-based Deep Reinforcement Learning (DRL) algorithm. Furthermore, to accelerate the convergence speed of the algorithm, we optimize the uplink transmit power allocation sub-problem separately by designing a numerical algorithm. We analyze the complexity of the algorithm and assess its convergence. Through extensive simulations across 5 different scenarios, our proposed scheme outperforms 4 reference schemes, showcasing reductions in total task processing costs ranging from 15.13% to 38.59%.

Split Federated Learning (SFL) has emerged as a promising paradigm to enhance FL by partitioning the Machine Learning (ML) model into parts and deploying them across clients and servers, effectively mitigating the workload on resource-constrained devices and preserving privacy. Compared to cloud-device-based and edge-device-based SFL, cloud-edge-device collaborative SFL offers both lower communication latency and wider network coverage. However, existing works adopt a uniform model partitioning strategy for different devices, ignoring the heterogeneous nature of device resources. This oversight leads to severe straggler problems, making the training process inefficient. Moreover, they do not consider joint optimization of model aggregation control and computing and communication resource allocation, and lack distributed algorithm design. To address these issues, we propose a joint resource management scheme for cloud-edge-device collaborative SFL to optimize the training latency and energy consumption of all devices. In our scheme, the partitioning strategy is optimized for each device based on resource heterogeneity. Meanwhile, we jointly optimize the aggregation frequency of ML models, computing resource allocation for all devices and edge servers, and transmit power allocation for all devices. We formulate a coordination game among all edge servers and then design a distributed optimization algorithm employing partially observable Multi-Agent Deep Reinforcement Learning (MADRL) with integrated numerical methods. Extensive experiments are conducted to validate the convergence of our algorithm and demonstrate the superiority of our scheme via evaluations under multiple scenarios and in comparison with four reference schemes.

Distinguished from terrestrial edge intelligence, satellite edge intelligence has unique characteristics, including the rapid mobility of satellites, limitations in computing and energy resources, and differences in the artificial intelligence models deployed on user devices, satellites, and ground cloud servers. In this paper, we propose a Deep Reinforcement Learning (DRL)-based resource management scheme for task inference in Low Earth Orbit (LEO)-based satellite-ground collaborative networks. In our approach, the task of a user can be inferred by the user device itself, the edge server of the current satellite via user-to-satellite transmission, the edge server of a neighboring satellite via satellite-to-satellite transmission, or a ground cloud server via satellite-to-cloud transmission. Our scheme jointly optimizes task offloading, computing resource allocation, and communication resource allocation to minimize the total system cost, which encompasses trade-offs among the task inference delays for all tasks, the energy consumption of system, and the task inference accuracies for all tasks, while ensuring that the transmit power budgets of all satellites and the satellite coverage time constraints for each user are met. A DRL-based algorithm combining the Softmax Deep Double Deterministic Policy Gradients (SD3) algorithm and two numerical methods is designed to solve the optimization problem efficiently. We prove the convergence of our algorithm and demonstrate the superiority of our scheme by performing extensive simulations in 4 scenarios with 4 reference schemes.

Vehicular edge intelligence, distinct from traditional edge intelligence, exhibits unique characteristics, including the mobility of vehicles, uneven spatial and temporal distribution of vehicles, and variability in the AI models deployed on vehicles, Roadside Units (RSUs), and edge servers (ESs). In this paper, we propose a Deep Reinforcement Learning (DRL)-based resource orchestration scheme for task inference in vehicle-RSU-edge collaborative networks. In our approach, vehicles' inference tasks can be processed on the vehicles, RSUs, or ESs, encompassing a total of 9 possible scenarios based on the cross-RSU mobility of vehicles. The scheme jointly optimizes task processing decision-making, transmission power allocation, computational resource allocation, and transmission rate allocation. The objective is to minimize the total cost, which involves a trade-off between task processing latency, energy consumption and inference error rate across all vehicle tasks. We design a DRL algorithm that decomposes the original optimization problem into sub-problems and efficiently solves them by combining the Softmax Deep Double Deterministic Policy Gradients (SD3) algorithm with multiple numerical methods. We analyzed the complexity and convergence of the algorithm. Specifically, we demonstrated its low complexity and fast, stable convergence, which prove its effectiveness in solving the problem. And we demonstrate the superiority of our scheme by comparing it with 5 benchmark schemes across 6 different scenarios.

We study the image cropping problem under the condition of determining the aspect ratio. We use two advanced saliency detection models to generate the initial region and then generate random cropping candidates around this region. These candidates are evaluated by an aesthetic scoring model. Next, Bayesian optimization is used to determine the next set of candidates that are likely to have better scores. These candidates are evaluated again by the aesthetic model, and after several iterations, we obtain a relatively optimal cropping result.

The combination of edge computing (EC) and blockchain can enhance task processing while ensuring security and credibility. To maximize system performance and avoid resource waste in task offloading, it is essential to match the resource allocation of the task computing and the blockchain consensus process. However, the existing works treated the above two processes as two independent processes and optimized them separately and ignored the above matching problem. In this article, we propose a resource management scheme for blockchain-assisted EC networks consisting of multiple devices, multiple base stations equipped with edge servers, a cloud server, and a network controller deployed on the edge layer. To minimize the total task processing delay and energy consumption of the devices, we formulate a joint task processing problem incorporating task scheduling, transmit power control, and computing resource allocation. To match the computing delay and consensus delay of each task, we balance the computing resources allocated for the two processes. We design a deep reinforcement learning (DRL) algorithm that utilizes the twin-delayed deep deterministic policy gradient (TD3) technology embedded with a fast numerical method, which effectively reduces the training complexity of the DRL model. Extensive experiments are conducted by varying four crucial parameters. The superiority of our scheme is demonstrated in comparison with three other reference schemes. The performance of our scheme is about 18.3%–24.1% higher than that of other schemes.

Vehicle edge computing (VEC) acts as an enhancement to provide low latency and low energy consumption for internet of vehicles (IoV) applications. Mobility of vehicles and load difference of roadside units (RSUs) are two important issues in VEC. The former results in task result reception failures owing to vehicles moving out of the coverage of their current RSUs; the latter leads to system performance degradation owing to load imbalance among the RSUs. They can be well solved by exploiting flexible RSU-RSU cooperation, which has not been fully studied by existing works. In this paper, we propose a novel resource management scheme for joint task offloading, computing resource allocation for vehicles and RSUs, vehicle-to-RSU transmit power allocation, and RSU-to-RSU transmission rate allocation. In our scheme, a task result can be transferred to the RSU where the vehicle is currently located, and a task can be further offloaded from a high-load RSU to a low-load RSU. To minimize the total task processing delay and energy consumption of all the vehicles, we design a twin delayed deep deterministic policy gradient (TD3)-based deep reinforcement learning (DRL) algorithm, where we embed an optimization subroutine to solve 2 sub-problems via numerical methods, thus reducing the training complexity of the algorithm. Extensive simulations are conducted in 6 different scenarios. Compared with 4 reference schemes, our scheme can reduce the total task processing cost by 17.3%-28.4%.

Enhanced by blockchain and cloud-edge-end cooperation, an edge computing network is capable to provide IoT (Internet of Things) devices higher task processing performance and better security and privacy guarantee. However, the joint resource management for both the task offloading and the blockchain services was less fully studied by existing works. To this end, in this paper, we focus on the task processing delay and energy consumption optimization problem in a multi-device and multi-base-station cloud-edge-end cooperative network. The task offloading, transmit power allocation, transmission rate allocation, and computing resource allocation are jointly optimized to minimize the long-term average total task processing delay of the tasks of all the devices while keeping the stability of the energy consumption of the devices and guaranteeing that the block mining speed matches the task offloading processes. We transform the optimization problem based on the Lyapunov optimization theory, and then design a hybrid deep reinforcement learning (DRL)-based algorithm. We decompose the problem into multiple sub-problems, and then embed multiple fast numerical methods into the twin delayed deep deterministic policy gradient (TD3) architecture as optimization subroutines to improve the learning performance of the DRL model. We also design a distributed deployment scheme for the algorithm and analyze the algorithm complexity. We demonstrate the superior performance of our algorithm in comparison with 5 reference schemes via extensive experiments in 7 scenarios.

In D2D (Device to Device)-assisted cloud-edge-device collaborative networks, the tasks of a busy device can be processed locally, offloaded to an idle device through D2D transmission, offloaded to the ES (Edge Server) of the affiliated BS (Base Station), further to the ES of another BS through ES-ES transmission, or further to the CS (Cloud Server) through ES-CS transmission. However, existing works did not fully consider both the D2D task offloading and cloud-edge-device collaboration. Moreover, the very high complexity of the joint resource optimization problem makes it extremely challenging to be solved efficiently by relying solely on numerical methods or machine-learning-based methods. In this paper, we propose a task offloading scheme to minimize the total system cost considering the time and energy consumption of all the devices. The task offloading decision, transmission power allocation, transmission rate allocation, and computational resource allocation are jointly optimized. We design a DRL (deep reinforcement learning)-based algorithm to solve the optimization problem efficiently through a hybrid approach, which decomposes the problem into sub-problems, and then jointly leverages an SD3 (Softmax Deep Double Deterministic Policy Gradients)-based DRL method to handle the task offloading sub-problem and uses multiple numerical methods to handle the other small-scale sub-problems. Extensive simulations are conducted in 7 scenarios. The superiority of our scheme is demonstrated in comparison with 4 reference schemes.

In time-slotted edge computing systems, task scheduling is conducted at the end of each time slot to make task offloading decisions and resource allocation for all the tasks pending for scheduling during the time slot. However, the existing works omitted the task scheduling delay, which is a period that a task has to wait from the task generation time point to the end of the current time slot. Such simplification is impractical in real scenarios because the task scheduling delay is a non-negligible part of the task processing delay, which was understood by existing works as the sum of only the task transmission and computing delays. In this paper, a novel time-slotted task offloading and resource allocation scheme for cloud-edge-end cooperative computing networks is proposed to realize the total task processing delay minimization for all the devices under the energy consumption constraint of each device. Our scheme makes task-offloading decision for each device from local processing, offloading to its affiliated base station (BS), to another BS, and to the cloud server. Besides, transmit power allocation, transmission rate allocation, and computing resource allocation are also jointly optimized in our optimization problem. We consider the impact of the task scheduling delay and design a two-stage distributed algorithm to decrease the negative impact by dividing the algorithm into a device-side part and a network-side part. The advantages of our scheme are validated by extensive simulations, where 4 reference schemes are compared in 8 different scenarios.

In an edge-cloud cooperative computing network, the task offloading performance can be further improved by the edge-cloud and edge-edge cooperation, in which the tasks can be offloaded from an edge server to the cloud server or another edge server. Such edge-cloud cooperative task offloading can jointly utilize the resources of all the edge servers and the cloud server. This paper proposes a collaborative service placement, task scheduling, computing resource allocation, and transmission rate allocation scheme for a multi-task and multi-service scenario with edge-cloud cooperation. The objective of our optimization problem is to minimize the total task processing delay while guaranteeing long-term task queuing stability. Considering the high complexity of the original optimization problem, we transform the problem into a deterministic problem for each time slot based on the Lyapunov optimization. Then, we design an iterative algorithm to obtain the whole solution to the problem efficiently based on a hybrid method using multiple numerical techniques. Further, considering the inherent difference in the optimization periods of the service placement, resource allocation, and task scheduling sub-problems, we design a multi-timescale algorithm to solve the sub-problems with different optimization periods. The complexity of the proposed algorithms is analyzed, and extensive simulations are conducted by varying multiple crucial parameters. The superiority of our scheme is demonstrated in comparison with 4 other schemes.

In multi-access edge computing (MEC) networks, network slicing enables the MEC network service provider (MEC-NSP) to provide customizable MEC services for user devices (UDs) with diverse QoS (Quality of Service) demands. In MEC network slicing, slice pricing and network resource allocation for slices are two core problems, which have not been jointly considered by existing works. To this end, we propose a two-stage slice pricing scheme to achieve balanced slice pricing and optimal network resource allocation. The goal of our scheme is to reduce the resource costs of the MEC-NSP and ensure its profit while meeting different user QoS requirements. At the first stage, we jointly optimize the computing, cache and communication resource allocation for all the slices by using problem decomposition. Then, we formulate a slice pricing problem based the Stackelberg game, prove the Nash equilibrium existence of the problem, and design an iterative algorithm based on the optimal response function. Extensive simulations are conducted in 4 scenarios, where our scheme is compared with 4 reference schemes. The simulation results demonstrate the superiority of our scheme in all the scenarios. The profit of the MEC-NSP optimized by our scheme is 17.64%-24.39% higher than those by the comparative works.

Multiaccess edge computing (MEC) is a promising approach to enhancing IoT devices running AI-based services. Especially, the edge–cloud architecture acts as a strong supporter of the resource-limited IoT devices. How to optimize the system resources efficiently to improve the service performance is the key issue in this scenario. Motivated by this, in this article, we focus on a multi-base station (BS) and multiservice edge–cloud-assisted IoT environment, where both the BSs (with edge servers deployed) and the cloud can assist the IoT devices to process multitype deep learning (DL) tasks via task offloading. DNN partition mechanism and both the communication and computing resources allocation are utilized to enable a collaborative optimization to minimize the processing delay of all the DL tasks in the system. Due to the mixed-integer nonlinear programming (MINLP) characteristic of our optimization problem, we propose an algorithm that decomposes the original problem into two subproblems, solves them separately, and then obtains the near-optimal solution efficiently. Extensive simulations are conducted by varying five different crucial parameters. The superiority of our scheme is demonstrated in comparisons with several other schemes proposed by existing works. Our scheme can achieve a notable 28.3% delay reduction on average.

Machine learning (ML) plays a key role in Intelligent Industrial Internet of Things (IIoT) applications. Processing of the computation-intensive ML tasks can be largely enhanced by applying edge computing (EC) to traditional cloud-based schemes. System optimizations in the existing works always ignore the inference accuracy of ML models with different complexities, and their impacts on error task inference. In this article, we propose a joint task offloading and resource allocation scheme for accuracy-aware machine-learning-based IIoT applications in an edge–cloud-based network architecture. We aim at minimizing the long-term average system cost affected by the task offloading, computing resource allocation, and inference accuracy of the ML models deployed on the sensors, edge server, and cloud server. The Lyapunov optimization technique is applied to convert the long-term stochastic optimization problem into a short-term deterministic problem. An optimal algorithm based on the general Benders decomposition (GBD) technology and a heuristic algorithm based on proportional computing resource allocation and task offloading strategy comparison are proposed to efficiently solve the problem, respectively. The performance of our scheme is proved by theoretical analysis and evaluated by extensive simulations conducted in multiple scenarios. Simulation results demonstrate the effectiveness and superiority of our two algorithms in comparison with several other schemes proposed by the existing works.

Mobile edge computing (MEC) shows prominent application prospects in the Industrial Internet of Things (IIoT) by allowing resource-restricted IIoT mobile devices (MDs) to offload their tasks to geographical proximity edge clouds. An efficient incentive mechanism should be designed jointly addressing resource allocation and pricing to incentivize MDs (i.e., buyers) and edge clouds (i.e., sellers) to participate in offloading service trading. This article aims to solve the social welfare maximization problem of a personalized MEC computation offloading service market where each edge cloud can allocate different computing and wireless resources to each MD according to the MDs’ delay and energy consumption constraints, and each MD submits bids to edge clouds differently based on the resource allocation of the edge clouds. We propose a truthful combinatorial auction (TCA) mechanism which involves three phases of resource allocation, buyer-seller matching, and payment determination. It should be highlighted that our proposed buyer-seller matching algorithm combines optimal matching and heuristic matching, so it greatly improves the auction effect while ensuring computational efficiency. Considerable theoretical analysis and experimental results prove that the performance of the proposed TCA mechanism is significantly superior to that of other auction mechanisms while holding the desirable properties.

Joint task inference, which fully utilizes end edge cloud cooperation, can effectively enhance the performance of deep neural network (DNN) inference services in the industrial internet of things (IIoT) applications. In this paper, we propose a novel joint resource management scheme for a multi task and multi service scenario consisting of multiple sensors, a cloud server, and a base station equipped with an edge server . A time slotted system model is proposed, incorporating DNN deployment, data size control, task offloading, computing resource allocation, and wireless channel allocation. Among them, the DNN deployment is to deploy proper DNNs on the edge server under its total resource constraint, and the data size control is to make trade off between task inference accuracy and task transmission delay through changing task da ta size. Our goal is to minimize the total cost including total task processing delay and total error inference penalty while guaranteeing long term task queue stability and all task inference accuracy requirements. Leveraging the Lyapunov optimization, we first transform the optimization problem into a deterministic problem for each time slot. Then, a deep deterministic policy gradient (DDPG) based deep reinforcement learning (DRL) algorithm is designed to provide the near optimal solution. We further desi gn a fast numerical method for the data size control sub problem to reduce the training complexity of the DRL model, and design a penalty mechanism to prevent frequent optimizations of DNN deployment. Extensive experiments are conducted by varying differen t crucial parameters. The superiority of our scheme is demonstrated in comparison with 3 other schemes.

In an internet of vehicle (IoV) scenario, vehicular edge computing (VEC) exploits the computing capabilities of the vehicles and roadside unit (RSU) to enhance the task processing capabilities of the vehicles. Resource management is essential to the performance improvement of the VEC system. In this paper, we propose a joint task offloading and resource allocation scheme to minimize the total task processing delay of all the vehicles through task scheduling, channel allocation, and computing resource allocation for the vehicles and RSU. Different from the existing works, our scheme: 1) considers task diversity by profiling the tasks of the vehicles by multiple attributes including data size, computation amount, delay tolerance, and task type; 2) considers vehicle classification by dividing the vehicles into 4 sets according to whether they have task offloading requirements or provide task processing services; 3) considers task processing flexibility by deciding for each vehicle to process its tasks locally, to offload the tasks to the RSU via V2I (Vehicle to Infrastructure) connections, or to the other vehicles via V2V (Vehicle to Vehicle) connections. An algorithm based on the Generalized Benders Decomposition (GBD) and Reformulation Linearization (RL) methods is designed to optimally solve the optimization problem. A heuristic algorithm is also designed to provide the sub-optimal solution with low computational complexity. We analyze the convergence and complexity of the proposed algorithms and conduct extensive simulations in 6 scenarios. The simulation results demonstrate the superiority of our scheme in comparison with 4 other schemes.

With the development of intelligent transport systems in the Internet of Things (IoT) smart cities, the bike-sharing system provides an environment-friendly choice for short-distance commuting, and it is employed extensively in major cities around the world. However, the issue of sharing bikes imbalance in various bike-sharing stations (BSS) constantly exists. Therefore, planning an effective route for rebalancing the bike-sharing system becomes a crucial task. In this article, based on a novel rebalancing problem of bike-sharing systems, which is to maximize the total allocated bikes at different stations under the constrained scheduling resources, we propose a meta-learning algorithm named ALRL to effectively allocate the sharing bikes under realistic constraints. Experimental results on real data sets and case studies demonstrate the effectiveness of our proposed approach which is better than the traditional methods.

Enabled by Multi-access Edge Computing (MEC) in a WiFi-cellular heterogeneous network, the tasks of mobile terminals (MTs) can be offloaded via the cellular network to the MEC servers or cloud server, or via the WiFi network to alleviate transmission congestion of the cellular network. The MEC also enables service caching to cache the programs/libraries/databases of the tasks to avoid repeated input data uploading. Existing research works lack joint optimization on the task offloading and service caching for MEC in the WiFi-cellular heterogeneous network. In this paper, a novel resource management scheme for joint task offloading and service caching is proposed to maximize the energy consumption benefits of all the MTs covered by a WiFi-cellular heterogeneous network while guaranteeing the task processing delay tolerance of each MT. We consider the constraints on limited computing and storage resources of the MEC servers equipped on the cellular base station and the WiFi access point, and we also consider cellular channel allocation for the task offloading. We design an iterative algorithm based on the alternating optimization technique to solve the proposed mixed integer nonlinear programming problem efficiently. Extensive simulations are conducted in multiple scenarios by varying different crucial parameters. The numerical results demonstrate that our scheme can largely improve the system performance in all the scenarios, and energy consumption reduction optimized by our scheme is 16.24%-43.09% higher than those by the comparative works.

The widely used Internet of Things (IoT) mobile devices (MDs) require fast processing capability to handle a large volume of computing tasks. Mobile-edge computing (MEC) can augment the capability of IoT MDs through offloading their computing tasks to the MEC-enabled base station (MEC-BS) that covers them. Most of the existing research works only focus on the computation offloading problems for a single MEC-BS. However, the load of a MEC-BS will rise as the increase of the scale of the offloaded tasks, especially during rush hours, and further it will result in deterioration of system performance. In this article, we propose a game-based multitype task offloading scheme among MEC-BSs. The tasks offloaded from IoT MDs can be further offloaded among MEC-BSs to alleviate high-load MEC-BSs. Aiming at balancing the computing delays of the tasks on each MEC-BS, a noncooperative game is formulated to model the computation offloading for the tasks with different types, indicated by computation amount, data size, and delay tolerance. The existence and convergence of the Nash equilibrium of the game are first proved using the variational inequality and regularization techniques. Then, we design a distributed iterative algorithm to efficiently solve the game problem. Simulation results show the fast convergence of our algorithm. The reduction of total computing delay optimized by our scheme can reach 45%–50% on average in multiple scenarios, and the superiority of our scheme is also demonstrated in comparisons with reference schemes.

For significant characteristics such as high efficiency and environmental protection, electric vehicles (EVs) have become a new technology trend in the intelligent transportation system (ITS). Data like real time traffic situation and charge point occupation in ITS shows significant real-time characteristics. Distributed storage system, an effective technology to ensure reliable sharing of dynamic data, is first adopted to storage allocation for delay-sensitive data in this article. In order to improve the recovery probability of delay-sensitive data within its timeliness, we establish an access queuing delay model based on the characteristics of sensitive data. Then, we find the optimal storage allocation strategy across distributed storage nodes for data with different delay threshold in terms of the recovery probability. The analysis shows that for data with low real-time performance and delay sensitivity, the maximal symmetric allocation is more excellent. For real-time data with low delay threshold, the minimal allocation is better under the condition of limited storage budget.

Electric vehicles (EVs) are viewed as an attractive option to reduce carbon emission and fuel consumption, but the popularization of EVs has been hindered by the cruising range limitation and the inconvenient charging process. In public charging stations, EVs usually spend a lot of time on queuing especially during peak hours of charging. Therefore, building an effective charging planning system has become a crucial task to reduce the total charging time for EVs. In this paper, we first introduce EVs charging scheduling problem and prove the NP-hardness of the problem. Then, we formalize the scheduling problem of EV charging as a Markov Decision Process and propose deep reinforcement learning algorithms to address it. The objective of the proposed algorithms is to minimize the total charging time of EVs and maximal reduction in the origin-destination distance. Finally, we experiment on real-world data and compare with two baseline algorithms to demonstrate the effectiveness of our approach. It shows that the proposed algorithms can significantly reduce the charging time of EVs compared to EST and NNCR algorithms.

Content-based sensor search is a core application in the Internet of Things (IoT), where target sensors can be quickly found by predicting the current output of the sensors. Due to the lack of update mechanism, the established prediction model in existing search architectures will gradually become unavailable in the highly dynamic IoT environment. Hence, a dual prediction structure based on edge intelligence is proposed in this article, to maintain the performance of the prediction model with minimum communication cost in long-term prediction. To implement our architecture, a more effective online learning algorithm is proposed to update prediction models online combined with our proposed adaptive window pattern clustering (AWPC) algorithm. Meanwhile, based on the dual prediction scheme (DPS), a mechanism is deployed on edge sensors to achieve transfer decision making in our architecture, where edge computing is performed to achieve selectively reporting data. With the designed architecture, about 76.56% of the communication energy consumption could be saved while achieving a 95.47% average prediction accuracy in continuous long-term prediction.

In terms of preventing traffic accidents, improving traffic efficiency and ensuring personal safety, the research on vehicular ad hoc networks (VANETs) is of great significance. Message authentication is an important security foundation for VANETs. With the rapid growth of the number of access terminals, the existing computing power of the VANETs will not be able to meet the fast message verification service load of large-scale dynamic networks. This article proposes a novel distributed collaborative authentication method. By selecting a reasonable number of assistance verification terminals in the VANETs system and cooperating with the roadside unit (RSU) to jointly undertake the task of network message verification, the purpose is to reduce the verification delay and achieve fast message verification. The simulation results show that in a large-scale connected vehicle with a large number of system terminals, the system message verification delay of our scheme is shortened to one tenth of the centralized verification system message verification delay.

In this letter, we propose a novel scheme, called Compensational Computation Offloading, which aims at efficiently maximizing the lifetime of edge networks. The main idea of the scheme is to offload the computing tasks of the nodes with low residual energies to the ones with high residual energies. At the same time, the energy consumptions cost by the transmitting and receiving tasks during the computation offloading are considered. IN order to balance the residual energies of the network, we design a high-efficiency scheduling algorithm which runs iteratively to search for near-optimal probabilities of each node offloading its tasks to its neighbor nodes. Finally, we obtain a set of probability values to form the overall scheduling algorithm. Simulation results demonstrate our scheme can efficiently increase the network's lifetime several times under different network topologies. The algorithm is very reliable and can approach approximate equality for each node in an edge network.

Event matching is vital in multi-dimensional content-based publish/subscribe services, which are widely employed for data dissemination in various scenarios. Existing mechanisms suffer from performance degradation in high-dynamic large-scale systems. To this end, we present GEM (Geometrical Event Matching), an analytic geometrical approach to fast event matching. GEM offers a very high event matching speed, and it also has low costs for subscription insertion/deletion operations and memory usage. In GEM, subscriptions are organized efficiently by a triangle-like index structure. A graph partitioning matching method and a selection matching method are jointly used for single-dimensional matching (SDM). Optimized by a decision algorithm for each incoming event, the event matching process is carried out in a pipeline consisting the SDM for each dimension. The search space shrinks continuously as the process goes, so that the event matching performance is promoted adaptively. A cache method is also designed to boost the first SDM in the pipeline. We implement extensive experiments to evaluate the performance of GEM in comparison with 3 state-of-the-art reference algorithms (TAMA, H-TREE and REIN). The results show that, the event matching time, subscription insertion/deletion time and memory consumption of GEM is on average 53.9%, 42.3%/49.5% and 31.8% lower than the best in other 3 algorithms, respectively. The event matching time of GEM is reduced efficiently via the cache method. The superiority of GEM appears more significantly as system scale and dynamic grow, and its performance also maintains in a high stability.