Federated reinforcement learning (FRL) has emerged as a promising paradigm, enabling multiple agents to collaborate and learn a shared policy adaptable across heterogeneous environments. Among the various reinforcement learning (RL) algorithms, the actor-critic (AC) algorithm stands out for its low variance and high sample efficiency. However, little to nothing is known theoretically about AC in a federated manner, especially each agent interacts with a potentially different environment. The lack of such results is attributed to various technical challenges: a two-level structure illustrating the coupling effect between the actor and the critic, heterogeneous environments, Markovian sampling and multiple local updates. In response, we study Single-Loop Federated Actor Critic (SFAC) where agents perform AC learning in a two-level federated manner while interacting with heterogeneous environments. We then provide bounds on the convergence error of SFAC. The results show that the convergence error asymptotically converges to a near-stationary point, with the extent proportional to environment heterogeneity. Moreover, the sample complexity exhibits a linear speed-up through the federation of agents. We evaluate the performance of SFAC through numerical experiments using common RL benchmarks, which demonstrate its effectiveness.

Federated Temporal Difference (FTD) learning has emerged as a promising framework for collaboratively evaluating policies without sharing raw data. Despite its potential, existing approaches often yield biased convergence results due to the inherent challenges of federated reinforcement learning, such as multiple local updates and environment heterogeneity. In response, we investigate federated temporal difference (TD) learning, focusing on collaborative policy evaluation with linear function approximation among agents operating in heterogeneous environments. We devise a heterogeneous federated temporal difference (HFTD) algorithm which iteratively aggregates agents' local stochastic gradients for TD learning. The HFTD algorithm involves two major novel contributions: 1) it aims to find the optimal value function model for the mixture environment which is the environment randomly drawn from agents' heterogeneous environments, using the local gradients of agents' mean squared Bellman errors (MSBEs) for their respective environments; 2) it allows agents to perform different numbers of local iterations for TD learning based on their heterogeneous computational capabilities. We analyze the finite-time convergence of the HFTD algorithm for the scenarios of IID sampling and Markovian sampling respectively. By characterizing bounds on the convergence error, we show that the HFTD algorithm can exactly converge to the optimal model and also achieves linear speedups as the number of agents increases.

Decentralized federated learning (DFL) can greatly reduce communication costs due to its decentralized communication structure compared to traditional centralized federated learning (FL). Existing works on FL with partial client participation often considered idealized scenarios (such as all clients participate in a round with the same probability), or required using clients' past gradient/model information which can be too costly to implement, or focused on centralized FL. In this paper, we study lightweight decentralized federated learning that does not use any client's past gradient/model information. We first present a novel sample-path-based cyclic convergence analysis for lightweight DFL with arbitrary client participation for the non-convex objectives case. The cyclic convergence analysis bounds clients' local model drifts due to partial participation over multiple rounds within a cycle and the cyclic consensus error via a per-cycle descent approach, while capturing the effect of client participation through a single unified term. By analyzing this term, we propose Cyclic Decentralized Federated Learning (CDFL), which enables general cyclic client participation by requiring only that each client performs the same total number of local updates per cycle. Our results show that CDFL achieves a convergence rate that matches existing benchmarks. We further propose a cyclic control framework that is both training-round and energy efficient to adaptively select participating clients and determine their number of local updates. Numerical experiments using real-world datasets verify our theoretical results and demonstrate the effectiveness of CDFL and the adaptive cyclic control framework.

In this paper, we explore distributed edge computation offloading (DECO) that offloads computation to distributed edge devices connected wirelessly, which perform the offloaded computation in parallel. By integrating edge computing with parallel computing, DECO can substantially reduce the total computation delay. In particular, we study the fundamental problem of minimizing the total completion time of DECO. We show that the time-sharing based communication resource allocation always outperforms the bandwidth-sharing scheme, so that it suffices to focus on the time-sharing based communication scheduling. Based on the time-sharing scheme, we first establish some structural properties of the optimal communication scheduling policy. Then, given these properties, we develop an efficient algorithm that finds the optimal allocation of computation workloads. Next, based on the optimal computation allocation, we characterize the optimal scheduling order of communications, which exhibits an elegant structure: the optimal order is in the non-decreasing order of the ratio between a device’s computation rate and its communication time. Last, based on the optimal computation allocation and communication scheduling, we show that the optimal device selection problem is a submodular minimization problem, so that it can be solved efficiently using some existing methods. We further extend the study to the setting where devices are subject to maximum computation workload constraints, and develop an efficient algorithm that finds the optimal computation workload allocation. Our results provide useful insights for the optimal computation-communication co-design for DECO. We evaluate the theoretical findings using extensive simulations in both practical settings and controlled settings, which demonstrate the performance of DECO in practice and also the efficiency of our proposed schemes and algorithms for DECO.

Federated learning (FL) has recently emerged as a promising paradigm that trains machine learning (ML) models on clients' devices in a distributed manner without the need of transmitting clients' data to the FL server. In many applications of ML (e.g., image classification), the labels of training data need to be generated manually by human agents (e.g., recognizing and annotating objects in an image), which are usually costly and error-prone. In this paper, we study FL with crowdsourced data labeling where the local data of each participating client of FL are labeled manually by the client. We consider the strategic behavior of clients who may not make desired effort in their local data labeling and local model computation (quantified by the mini-batch size used in the stochastic gradient computation), and may misreport their local models to the FL server. We first characterize the performance bounds on the training loss as a function of clients' data labeling effort, local computation effort, and reported local models, which reveal the impacts of these factors on the training loss. With these insights, we devise Labeling and Computation Effort and local Model Elicitation (LCEME) mechanisms which incentivize strategic clients to make truthful efforts as desired by the server in local data labeling and local model computation, and also report true local models to the server. The truthful design of the LCEME mechanism exploits the non-trivial dependence of the training loss on clients' hidden efforts and private local models, and overcomes the intricate coupling in the joint elicitation of clients' efforts and local models. Under the LCEME mechanism, we characterize the server’s optimal local computation effort assignments and analyze their performance. We evaluate the proposed FL algorithms with crowdsourced data labeling and the LCEME mechanism for the MNIST-based hand-written digit classification. The results corroborate the improved learning accuracy and cost-effectiveness of the proposed approaches.

The rapid advances in federated learning (FL) in the past few years have recently inspired a great deal of research on this emerging topic. Existing work on FL often assume that clients participate in the learning process with some particular pattern (such as balanced participation), and/or in a synchronous manner, and/or with the same number of local iterations, while these assumptions can be hard to hold in practice. In this paper, we propose AFL-DGA, an Anarchic Federated Learning algorithm with Delayed Gradient Averaging, which gives maximum freedom to clients. In particular, AFL-DGA allows clients to 1) participate in any rounds; 2) participate asynchronously; 3) participate with any number of local iterations; 4) perform gradient computations and gradient communications in parallel. The proposed AFL-DGA algorithm enables clients to participate in FL flexibly according to their heterogeneous and time-varying computation and communication capabilities, and also efficiently by improving utilization of their computation and communication resources. We characterize performance bounds on the learning loss of AFL-DGA as a function of clients' local iteration numbers, local model delays, and global model delays. Our results show that the AFL-DGA algorithm can achieve a convergence rate of [EQUATION] and also a linear convergence speedup, which matches that of existing benchmarks. The results also characterize the impacts of various system parameters on the learning loss, which provide useful insights. Numerical results demonstrate the efficiency of the proposed algorithm.

Distributed edge computation offloading makes use of distributed wireless edge devices to perform offloaded computation in parallel, which can substantially reduce the computation time. In this article, we explore distributed edge computation offloading where the computation workloads of edge devices are correlated with their communication workloads. In particular, we study the fundamental problem of computation workload allocation and communication scheduling for minimizing the total completion time of the computation offloading. To solve this problem, we need to tackle several challenges due to the precedence constraints of computations and communications, the interference constraints of wireless edge devices, and the correlation between computation and communication workloads. We consider preemptive, half-preemptive, and non-preemptive networks for the formulated problem, respectively. For each setting, we first develop a simplified problem of computation allocation, based on which we then devise an efficient and feasible policy that can arbitrarily approach the optimal policy. For half-preemptive and non-preemptive networks, we also characterize the optimal communication orders. Our results provides useful insights for the computation-communication co-design of distributed edge computation offloading. We evaluate the proposed algorithms using simulation results, which corroborate the advantages of the algorithms.

Crowdsourcing has found a wide variety of applications, including spectrum sensing, traffic monitoring, as well as data annotation for machine learning based data analytics. To improve data accuracy and cost-effectiveness, workers’ data quality can be learned from their data in an online manner, which can be used for task assignment and data aggregation. However, crowdsourcing is vulnerable to data poisoning attacks, where the attacker reports malicious data to reduce aggregated data accuracy. In this paper, we study malicious data attacks on dynamic crowdsourcing where tasks are assigned and performed sequentially, and we explore online quality learning as a defense mechanism against the attack by finding malicious workers with low quality. We first focus on the asymptotic setting where workers’ quality is accurately learned by the requester, based on which we then turn to the general non-asymptotic setting where the quality is estimated online with errors. For each setting, we first characterize the conditions under which the attack strategy can effectively reduce the aggregated data accuracy. Our results show that the malicious noise variance needs to be within a certain range for the attack to be effective. Then we analyze the harm of effective attack strategies. It reveals that the regret of the online quality learning algorithm can be substantially increased from$\mathcal {O}(\log ^2T)$(upper bound) to$\Omega (T)$(lower bound) due to effective attacks. To further mitigate the attack, we also study median and maximum influence of estimation based data aggregation as defense mechanisms. Our results provide useful insights on the impacts of data poisoning attacks when online quality learning is used to defend against the attack. We evaluate the proposed attacks and defenses via extensive simulation results based on real-world data, which demonstrate the effectiveness of the attacks and defenses.

The space-air-ground integrated network (SAGIN), one of the key technologies for next-generation mobile communication systems, can facilitate data transmission for users all over the world, especially in some remote areas where vast amounts of informative data are collected by Internet of remote things (IoRT) devices to support various data-driven artificial intelligence (AI) services. However, training AI models centrally with the assistance of SAGIN faces the challenges of highly constrained network topology, inefficient data transmission, and privacy issues. To tackle these challenges, we first propose a novel topology-aware federated learning framework for the SAGIN, namely Olive Branch Learning (OBL). Specifically, the IoRT devices in the ground layer leverage their private data to perform model training locally, while the air nodes in the air layer and the ring-structured low earth orbit (LEO) satellite constellation in the space layer are in charge of model aggregation (synchronization) at different scales. To further enhance communication efficiency and inference performance of OBL, an efficient Communication and Non-IID-aware Air node-Satellite Assignment (CNASA) algorithm is designed by taking the data class distribution of the air nodes as well as their geographic locations into account. Furthermore, we extend our OBL framework and CNASA algorithm to adapt to more complex multi-orbit satellite networks. We analyze the convergence of our OBL framework and conclude that the CNASA algorithm contributes to the fast convergence of the global model. Extensive experiments based on realistic datasets corroborate the superior performance of our algorithm over the benchmark policies.

Edge-computing-enhanced Internet of Vehicles (EC-IoV) enables ubiquitous data processing and content sharing among vehicles and terrestrial edge computing (TEC) infrastructures (e.g., 5G base stations and roadside units) with little or no human intervention, and plays a key role in the intelligent transportation systems. However, EC-IoV is heavily dependent on the connections and interactions between vehicles and TEC infrastructures, thus will break down in some remote areas where TEC infrastructures are unavailable (e.g., desert, isolated islands, and disaster-stricken areas). Driven by the ubiquitous connections and global-area coverage, space–air–ground-integrated networks (SAGINs) efficiently support seamless coverage and efficient resource management, and represent the next frontier for edge computing. In light of this, we first review the state-of-the-art edge computing research for SAGINs in this article. After discussing several existing orbital and aerial edge computing architectures, we propose a framework of edge computing-enabled SAGINs to support various Internet of Vehicles (EC-IoV) services for the vehicles in remote areas. The main objective of the framework is to minimize the task completion time and satellite resource usage. To this end, a preclassification scheme is presented to reduce the size of action space, and a deep imitation learning-driven offloading and caching algorithm is proposed to achieve real-time decision making. The simulation results show the effectiveness of our proposed scheme. Finally, we also discuss some technology challenges and future directions.

Data crowdsourcing is a promising paradigm that leverages the “wisdom” of a potentially large crowd of “workers” in many application domains. Quality-aware crowdsourcing is beneficial as it makes use of workers’ data quality to perform task allocation and data aggregation. However, a worker’s quality and data can be her private information that she may have incentive to misreport to the crowdsourcing requester. Moreover, a worker’s quality and data can depend on her sensitive information (e.g., location), which can be inferred from the outcomes of task allocation and data aggregation by an adversary. In this paper, we devise Privacy-preserving crowdsourcing mechanisms for truthful Data Quality Elicitation (PDQE). In these mechanisms, we design differentially private task allocation and data aggregation algorithms to prevent the inference of a worker’s quality and data from the outcomes of these algorithms. In the meantime, the mechanisms also incentivize workers to truthfully report their quality and data and make desired efforts. We first focus on the mechanisms for a single task (S-PDQE) and then extend it to the case of multiple tasks (M-PDQE). We further show that both the mechanisms achieve a bounded performance gap compared to the optimal strategy. We evaluate the proposed mechanisms using simulations based on real-world data, which corroborate their highly-desired properties on truthful data quality elicitation, data accuracy and privacy protection.

Incentive mechanism design and quality control are two key challenges in data crowdsourcing, because of the need for recruitment of crowd users and their limited capabilities. Without considering users’ social influences, existing mechanisms often result in low efficiency in terms of the platform’s cost. In this paper, we exploit social influences among users as incentives to motivate users’ participation, in order to reduce the cost of recruiting users. Based on social influences, we design incentive mechanisms with the goal of achieving high quality of crowdsourced data and low cost of incentivizing users’ participation. Specifically, we consider three scenarios. In the full information scenario, we design task assignment and user recruitment mechanisms to optimize the data quality while reducing the incentive cost. In the partial information scenario, users’ qualities and costs are unknown. We exploit the correlation between tasks to overcome the information asymmetry, for both cases of opportunistic crowdsourcing and participatory crowdsourcing. Further, in the dynamic social influence scenario, we investigate the dynamics of users’ social influences and design extra rewards for users to make full use of the social influence and achieve maximum cost saving. We evaluate the incentive mechanisms using numerical results, which demonstrate their effectiveness.

In wireless federated learning (WFL), machine learning (ML) models are trained distributively on wireless edge devices without the need of collecting data from the devices. In such a setting, the quality of a local model update heavily depends on the variance of the local stochastic gradient, determined by the mini-batch data size used to compute the update. In this paper, we explore quality-aware distributed computation for WFL where user devices share limited communication resources, using mini-batch size as a "knob" to control the quality of users’ local updates. In particular, we study joint mini-batch size design and communication scheduling, with the goal of minimizing the training loss as well as the training time of the FL algorithm. For the case of IID data, we first characterize the optimal communication scheduling and the optimal minibatch sizes. Then we develop a greedy algorithm that finds the optimal set of participating users with an approximation ratio. For the case of non-IID data, we first characterize the optimal communication structure and the optimal mini-batch sizes. Then we develop algorithms that find the optimal communication order for some special cases. Our findings provide useful insights for the computation-communication co-design for WFL. We evaluate the proposed mini-batch size design and communication scheduling using simulations, which corroborate improved learning accuracy and learning time.

Wireless federated learning (WFL) trains machine learning (ML) models on wireless edge devices in a distributed manner without the need of collecting data from users. In WFL, the quality of a local model update depends on the variance of the local stochastic gradient, determined by the mini-batch data size used to compute the update. In this paper, we study quality-aware distributed computation for WFL with non-convex problems and asynchronous algorithms, using mini-batch size as a "knob" to control the quality of users' local updates. We first characterize performance bounds on the training loss as a function of local updates' quality over the training process, for both non-convex and asynchronous settings. Our findings reveal that the impact of a local update's quality on the training loss 1) increases with the stepsize used for that local update for non- convex learning, and 2) increases when there are more other users' local updates which are coupled with that local update (depending on the update delays) for asynchronous learning. Based on these useful insights, we design channel-aware adaptive algorithms that determine users' mini-batch sizes over the training process, based on the impacts of local updates' quality on the training loss as well as users' wireless channel conditions (which determine the update delays) and computation costs. We evaluate the proposed quality- aware adaptive algorithms using simulations, which demonstrate improved learning accuracy and learning cost.

Recently, smart cities, healthcare system, and smart vehicles have raised challenges on the capability and connectivity of state-of-the-art Internet-of-Things (IoT) devices, especially for the devices in hotspots area. Multiaccess edge computing (MEC) can enhance the ability of emerging resource-intensive IoT applications and has attracted much attention. However, due to the time-varying network environments, as well as the heterogeneous resources of network devices, it is hard to achieve stable, reliable, and real-time interactions between edge devices and their serving edge servers, especially in the 5G ultradense network (UDN) scenarios. Ultradense edge computing (UDEC) has the potential to fill this gap, especially in the 5G era, but it still faces challenges in its current solutions, such as the lack of: 1) efficient utilization of multiple 5G resources (e.g., computation, communication, storage, and service resources); 2) low overhead offloading decision making and resource allocation strategies; and 3) privacy and security protection schemes. Thus, we first propose an intelligent UDEC (I-UDEC) framework, which integrates blockchain and artificial intelligence (AI) into 5G UDEC networks. Then, in order to achieve real-time and low overhead computation offloading decisions and resource allocation strategies, we design a novel two-timescale deep reinforcement learning (2Ts-DRL) approach, consisting of a fast-timescale and a slow-timescale learning process, respectively. The primary objective is to minimize the total offloading delay and network resource usage by jointly optimizing computation offloading, resource allocation, and service caching placement. We also leverage federated learning (FL) to train the 2Ts-DRL model in a distributed manner, aiming to protect the edge devices' data privacy. Simulation results corroborate the effectiveness of both the 2Ts-DRL and FL in the I-UDEC framework and prove that our proposed algorithm can reduce task execution time up to 31.87%.

By integrating edge computing with parallel computing, distributed edge computing (DEC) makes use of distributed devices in edge networks to perform computing in parallel, which can substantially reduce service delays. In this paper, we explore DEC that exploits distributed edge devices connected by a wireless network to perform a computation task offloaded from an end device. In particular, we study the fundamental problem of minimizing the delay of executing a distributed algorithm of the computation task. We first establish some structural properties of the optimal communication scheduling policy. Then, given these properties, we characterize the optimal computation allocation policy, which can be found by an efficient algorithm. Next, based on the optimal computation allocation, we characterize the optimal scheduling order of communications for some special cases, and develop an efficient algorithm with a finite approximation ratio to find it for the general case. Last, based on the optimal computation allocation and communication scheduling, we further show that the optimal selection of devices can be found efficiently for some special cases. Our results provide some useful insights for the optimal computation-communication codesign. We evaluate the performance of the theoretical findings using simulations.

The fusion of edge computing and artificial intelligence (AI) technology is a key enabler for the smart Internet of Things (IoT). However, these two emerging paradigms face many issues for their integration, such as data storage structure, model generation algorithms, and cloud-edge collaboration mechanisms. Moreover, edge computing is not ready for supporting AI and can be enabled to support AI via some basic network functions related to Quality of Experience (QoE), such as passive computation offloading and content caching. In this article, we present an intelligent cooperative edge (ICE) computing in IoT networks to achieve a complementary integration of AI and edge computing. The AI-related modules of edge computing are redesigned for distributing AI's core functions from the cloud to the edge. IoT-generated data are differentiated as user-private data preserved locally in IoT devices, edge-private data isolated on the edge and public data uploaded to the cloud. Therefore, a cloud-scale machine learning model can be generated, followed by privacy-preserving transfer learning running on each edge, which also has data updated more frequently that enables the model's incremental learning. The model distribution is accomplished through lightweight deployment pipelines consisting of cloud compression and edge reconstruction. Conversely, some key issues of edge computing, such as the computation offloading and content caching, achieve a better solution using the localized AI. We perform the prototype-based evaluation, which indicates that the ICE computing architecture enables a benign combination of AI and edge computing.

Crowdsensing has found a variety of applications (e.g., spectrum sensing, environmental monitoring) by leveraging the "wisdom" of a potentially large crowd of mobile users as "workers". The value of data collected in crowdsensing heavily depends on the quality of data provided by the workers participating in a crowdsensing task. In general, the quality of data varies for different workers. To fully exploit the potential of crowdsensing, it is important for the crowdsensing requester to know workers' data quality, based on which the requester allocates tasks to workers and aggregates data from workers. Such quality-aware crowdsensing can greatly improve the value and usefulness of data in crowdsensing. However, the quality of workers' data is often unknown to the requester (due to, e.g., workers' characteristics are unknown). In this paper, under a dynamic multi-task crowdsensing framework, we devise an online data quality learning algorithm that learns the data quality of workers from their data on the fly, while making use of the learned quality information to perform task allocation and data aggregation. Compared to prior online learning algorithms (such as those for the multi-armed bandit problems), our algorithm needs to overcome the challenge that the ground truth of the interested variable is unknown. We show that under some mild conditions, our algorithm converges to the offline optimal strategy over time, and have a regret in the order of O(logt) compared to the offline optimal strategy. We provide bounds on the regret for both the requester's utility and cost, and for both the simple average rule and the weighted average rule. We demonstrate the efficiency of the algorithm using simulation results.

Crowdsensing has found a broad range of applications (e.g., spectrum sensing, environmental monitoring) by leveraging the "wisdom" of a potentially large crowd of "workers" (i.e., mobile users). One important class of applications use the data collected from crowdsensing for data analytics via machine learning (e.g., for wireless indoor localization). To exploit the potential of crowdsensing for machine learning, it is beneficial for the crowdsensing requester to know and make use of the quality of worker's data. In this paper, based on a general linear regression model of machine learning, we devise truthful quality-aware crowdsensing mechanisms for quality and effort elicitation, which incentivize workers to truthfully report their private worker quality to the requester, and make effort as desired by the requester. The truthful design of the mechanisms overcomes the differences of ground truths of workers' tasks, and the coupling in the joint elicitation of workers' quality, effort, and data. Under the mechanisms, we investigated the socially optimal and the requester's optimal effort assignments, and analyze their performance. We show that the requester's optimal assignment is determined by the "virtual quality" rather than the highest quality among workers, which depends on the worker's quality and the quality's distribution. Simulation results are provided which demonstrate the truthfulness of the mechanisms and the performance of the optimal effort assignments.

Mobile crowdsensing has found a variety of applications (e.g., spectrum sensing, environmental monitoring) by leveraging the “wisdom” of a potentially large crowd of mobile users. An important metric of a crowdsensing task is data accuracy, which relies on the data quality of the participating users' data (e.g., users' received SNRs for measuring a transmitter's transmit signal strength). However, the quality of a user can be its private information (which, e.g., may depend on the user's location) that it can manipulate to its own advantage, which can mislead the crowdsensing requester about the knowledge of the data's accuracy. This issue is exacerbated by the fact that the user can also manipulate its effort made in the crowdsensing task, which is a hidden action that could result in the requester having incorrect knowledge of the data's accuracy. In this paper, we devise truthful crowdsensing mechanisms for Quality and Effort Elicitation (QEE), which incentivize strategic users to truthfully reveal their private quality and truthfully make efforts as desired by the requester. The QEE mechanisms achieve the truthful design by overcoming the intricate dependency of a user's data on its private quality and hidden effort. Under the QEE mechanisms, we show that the crowdsensing requester's optimal (RO) effort assignment assigns effort only to the best user that has the smallest “virtual valuation”, which depends on the user's quality and the quality's distribution. We also show that, as the number of users increases, the performance gap between the RO effort assignment and the socially optimal effort assignment decreases, and converges to 0 asymptotically. We further discuss some extensions of the QEE mechanisms. Simulation results demonstrate the truthfulness of the QEE mechanisms and the system efficiency of the RO effort assignment.

Mobile fog computing (MFC) is an emerging paradigm that extends cloud computing (CC) by adding a new layer between the cloud and its end users.With the cloud-aware MFC, the cloud can pre-push certain important resources to the fog to reduce the networking latency and release the traffic burden over the links.e end user then is able to perform offline computing on the fog layer so that only the important results need to be delivered to and stored in the cloud.Moreover, the dense geographical deployment of fog servers enables the system to be aware of the end user's location.erefore, some location-sensitive applications could be well supported by the fog-aided cloud systems.Note that the cloud-aware MFC is different from the mobile edge computing (MEC), another promising technology for overcoming the shortcomings of CC, since MFC is able to jointly work with the cloud, but MEC is usually defined by the exclusion of CC.Specifically, in MEC, computing applications, data, and services are pushed away from the centralized nodes to the network edge, which enables network edge to run in an isolated environment from the rest of the network and provides access to local resources and data.In contrast, MFC provides not only a systemlevel horizontal architecture but also a new way to distribute, orchestrate, and manage secure resources across the network rather than just performing computing at the network edge.How to design efficient system architectures, transmission strategies, and protocols for MFC and how to efficiently analyze and evaluate the system performance are very important and essential.ese topics have carved out a new area rich in research and innovation potential.is special issue aims to address all these topics and invite contributions from worldwide leading researchers.

Mobile data crowdsourcing has found a broad range of applications (e.g., spectrum sensing, environmental monitoring) by leveraging the "wisdom" of a potentially large crowd of "workers" (i.e., mobile users). A key metric of crowdsourcing is data accuracy, which relies on the quality of the participating workers' data (e.g., the probability that the data is equal to the ground truth). However, the data quality of a worker can be its own private information (which the worker learns, e.g., based on its location) that it may have incentive to misreport, which can in turn mislead the crowdsourcing requester about the accuracy of the data. This issue is further complicated by the fact that the worker can also manipulate its effort made in the crowdsourcing task and the data reported to the requester, which can also mislead the requester. In this paper, we devise truthful crowdsourcing mechanisms for Quality, Effort, and Data Elicitation (QEDE), which incentivize strategic workers to truthfully report their private worker quality and data to the requester, and make truthful effort as desired by the requester. The truthful design of the QEDE mechanisms overcomes the lack of ground truth and the coupling in the joint elicitation of worker quality, effort, and data. Under the QEDE mechanisms, we characterize the socially optimal and the requester's optimal task assignments, and analyze their performance. We show that the requester's optimal assignment is determined by the largest "virtual valuation" rather than the highest quality among workers, which depends on the worker's quality and the quality's distribution. We evaluate the QEDE mechanisms using simulations which demonstrate the truthfulness of the mechanisms and the performance of the optimal task assignments.

Mobile crowd sensing has found a variety of applications (e.g., spectrum sensing, environmental monitoring) by leveraging the "wisdom" of a potentially large crowd of mobile users. An important metric of a crowd sensing task is data accuracy, which relies on the qualities of the participating users' data (e.g., users' received SNRs for measuring a transmitter's transmit signal strength). However, the quality of a user can be its private information (which, e.g., may depend on the user's location) that it can manipulate to its own advantage, which can mislead the crowd sensing requester about the knowledge of the data's accuracy. This issue is exacerbated by the fact that the user can also manipulate its effort made in the crowd sensing task, which is a hidden action that could result in the requester having incorrect knowledge of the data's accuracy. In this paper, we devise truthful crowd sensing mechanisms for Quality and Effort Elicitation (QEE), which incentivize strategic users to truthfully reveal their private qualities and truthfully make efforts as desired by the requester. The QEE mechanisms achieve the truthful design by overcoming the intricate dependency of a user's data on its private quality and hidden effort. Under the QEE mechanisms, we show that the crowd sensing requester's optimal (CO) effort assignment assigns effort only to the best user that has the smallest "virtual valuation", which depends on the user's quality and the quality's distribution. We also show that, as the number of users increases, the performance gap between the CO effort assignment and the socially optimal effort assignment decreases, and converges to 0 asymptotically. We further show that while the requester's payoff and the social welfare attained by the CO effort assignment both increase as the number of users increases, interestingly, the corresponding users' payoffs can decrease. Simulation results demonstrate the truthfulness of the QEE mechanisms and the system efficiency of the CO effort assignment.

The rapid growth of online social networks has strengthened wireless users' social relationships, which in turn has resulted in more data traffic due to network effect in the social domain. Nevertheless, the boosted demand for wireless services may challenge the limited wireless capacity. To build a thorough understanding, we study mobile users' data usage behavior by jointly considering the network effect due to their social relationships in the social domain and the congestion effect in the physical wireless domain. Specifically, we develop a Stackelberg game for socially aware data usage: in Stage I, a wireless provider first decides the data pricing to all users in order to maximize its revenue, and then in Stage II, users decide their data usage, for the given price, subject to mutual interactions under both social network effect and congestion effect. We analyze the two-stage game via backward induction. In particular, for Stage II, we first provide conditions for the existence and the uniqueness of a user demand equilibrium (UDE). Then, we propose algorithms to find the UDE and for users to reach the UDE in a distributed manner. We further investigate the impact of different system parameters on the UDE. Next, for Stage I, we develop an optimal pricing algorithm to maximize the wireless provider's revenue. We numerically evaluate the performance of our proposed algorithms using real data, and thereby draw useful engineering insights for the operation of wireless providers: 1) when social network effect dominates congestion effect, the marginal gain of the total usage increases with the social ties and the number of users, or decreases with the congestion coefficient; in contrast, when congestion effect dominates social network effect, the marginal gain decreases (or increases, respectively) with these parameters and 2) when social network effect is strong, a lower price should be set to increase the total revenue; in contrast, when congestion effect is strong, a higher price is preferred.

Distributed spectrum access (DSA) is challenging, since an individual secondary user often has limited sensing capabilities only. One key insight is that channel recommendation among secondary users can help to take advantage of the inherent correlation structure of spectrum availability in both time and space, and enable users to obtain more informed spectrum opportunities. With this insight, we advocate to leverage the wisdom of crowds, and devise social recommendation aided DSA mechanisms to orient secondary users to make more intelligent spectrum access decisions, for both strong and weak network information cases. We start with the strong network information case where secondary users have the statistical information. To mitigate the difficulty due to the curse of dimensionality in the stochastic game approach, we take the one-step Nash approach and cast the social recommendation aided DSA decision making problem at each time slot as a strategic game. We show that it is a potential game, and then devise an algorithm to achieve the Nash equilibrium by exploiting its finite improvement property. For the weak information case where secondary users do not have the statistical information, we develop a distributed reinforcement learning mechanism for social recommendation aided DSA based on the local observations of secondary users only. Appealing to the maximum-norm contraction mapping, we also derive the conditions under which the distributed mechanism converges and characterize the equilibrium therein. Numerical results reveal that the proposed social recommendation aided DSA mechanisms can achieve a superior performance using real social data traces and its performance loss in the weak network information case is insignificant, compared with the strong network information case.

With increasing popularity of location-based services (LBSs), there have also been growing concerns for location privacy. To protect location privacy in an LBS, mobile users in physical proximity can work in concert to collectively change their pseudonyms, in order to hide spatial-temporal correlation in their location traces. In this paper, we leverage mobile users' social tie structure to motivate them to participate in pseudonym change. Drawing on a social group utility maximization framework, we cast users' decision making of whether to change pseudonyms as a socially aware pseudonym change game (SA-PCG). The SA-PCG further assumes a general anonymity model that allows a user to have its specific anonymity set for personalized location privacy. For the SA-PCG, we show that there exists a socially aware Nash equilibrium (SNE), and quantify the system efficiency of SNEs with respect to the optimal social welfare. Then, we develop a greedy algorithm that myopically determines users' strategies, based on the social group utility derived from only the users whose strategies have already been determined. We show that this algorithm efficiently finds an SNE that enjoys desirable properties: 1) it is socially aware coalition-proof, and thus is also Pareto-optimal; 2) it achieves higher social welfare than any SNE for the socially oblivious pseudonym change game. We further quantify the system efficiency of this SNE with respect to the optimal social welfare. We also show that this SNE can be achieved in a distributed manner. Numerical results using real data corroborate that social welfare can be significantly improved by exploiting social ties.

In spite of the pronounced benefit brought by crowdsensing, a user would not participate in sensing without adequate incentive, indicating that effective incentive design plays a critical role in making crowdsensing a reality. In this work, we examine the impact of two conflicting factors on incentives for users' participation: 1) the concern about privacy leakage and 2) the (positive) network effect from many sensing participants. The former factor hinders privacy- aware users from participating, whereas the latter encourages users' participation. Taking into consideration both factors, we devise a privacy-preserving crowdsensing scheme, in which a reverse `privacy' auction is first run by the crowdsensing platform to select users based on their privacy valuations and the network effect. Then the trusted platform carries out differentially private data aggregation over the collected data such that the released sensing result remains useful for the task agent, while all participants' data privacy is guaranteed. A natural objective here is then to maximize the profit of the task agent, i.e., the difference between its utility and the total reward to the participants. To this end, the platform utilizes a random-sampling based mechanism for the 'privacy' auction, followed by a Laplace mechanism for data aggregation. We show that this auction mechanism design is 4-competitive, and further it exhibits desirable properties, including individual rationality, truthfulness, computational efficiency. Simulation results corroborate the theoretical properties of the proposed privacy-preserving crowdsensing scheme.

We develop a social group utility maximization (SGUM) framework for cooperative wireless networking that takes into account both social relationships and physical coupling among users. Specifically, instead of maximizing its individual utility or the overall network utility, each user aims to maximize its social group utility that hinges heavily on its social tie structure with other users. We show that this framework provides rich modeling flexibility and spans the continuum between non-cooperative game and network utility maximization (NUM)-two traditionally disjoint paradigms for network optimization. Based on this framework, we study three important applications of SGUM, in database assisted spectrum access, power control, and random access control, respectively. For the case of database assisted spectrum access, we show that the SGUM game is a potential game and always admits a socially-aware Nash equilibrium (SNE). We also develop a distributed spectrum access algorithm that can converge to the SNE and also quantify the trade-off between the performance and convergence time of the algorithm. For the cases of power control and random access control, we show that there exists a unique SNE and the network performance improves as the strength of social ties increase. Numerical results corroborate that the SGUM solutions can achieve superior performance using real social data trace. Furthermore, we show that the SGUM framework can be generalized to take into account both positive and negative social ties among users, which can be a useful tool for studying network security problems.

By taking advantage of active sensing using radio waves, radar sensors can offer several advantages over passive sensors. Although much attention has been given to multistatic and multiple-input-multiple-output (MIMO) radar concepts, little has been paid to understanding radar networks (i.e., multiple individual radars working in concert). In this context, we study the coverage problem of a bistatic radar (BR) sensor network, which is very challenging due to the Cassini oval sensing region of a BR and the coupling of sensing regions across different BRs. In particular, we consider the problem of deploying a network of BRs in a region to maximize the worst-case intrusion detectability, which amounts to minimizing the vulnerability of a barrier. We show that it is optimal to place BRs on the shortest barrier if it is the shortest line segment that connects the left and right boundary of the region. Based on this, we study the optimal placement of BRs on a line segment to minimize its vulnerability, which is a nonconvex optimization problem. By exploiting certain specific structural properties pertaining to the problem (particularly an important structure of detectability), we characterize the optimal placement order and the optimal placement spacing of the BR nodes, both of which present elegant balanced structures. Our findings provide valuable insights into the placement of BRs for barrier coverage. To our best knowledge, this is the first work to explore the barrier coverage of a network of BRs.

In this paper, we study a privacy-preserving spectrum sharing system to protect secondary users' location privacy while enhancing spectrum access. The location privacy of secondary users can be compromised by an external adversary via the received signal strength (RSS)-based localization technique. To mitigate such privacy threat, we employ a random power perturbation approach that allows each secondary user to judiciously obfuscate the RSS captured by the adversary. While it can protect users' location privacy, the power perturbation approach would inevitably degrade the system performance and bring challenges to the design of the spectrum allocation algorithm. In this work, we adopt a socially-aware database assisted spectrum access system and cast the spectrum allocation under users' power perturbation as a stochastic channel selection game played among the users. To tackle the challenge brought by the privacy protection, we develop a two time-scale distributed learning algorithm, which is shown to converge almost surely to a socially-aware ε-Nash equilibrium. The numerical results show that the higher the privacy protection level is, the more significant the degradation of the network throughput would be.

With increasing popularity of location-based services (LBSs), there have been growing concerns for location privacy. To protect location privacy in a LBS, mobile users in physical proximity can work in concert to collectively change their pseudonyms, in order to hide spatial-temporal correlation in their location traces. In this study, we leverage the social tie structure among mobile users to motivate them to participate in pseudonym change. Drawing on a social group utility maximization (SGUM) framework, we cast users' decision making of whether to change pseudonyms as a socially-aware pseudonym change game (PCG). The PCG further assumes a general anonymity model that allows a user to have its specific anonymity set for personalized location privacy. For the SGUM-based PCG, we show that there exists a socially-aware Nash equilibrium (SNE), and quantify the system efficiency of the SNE with respect to the optimal social welfare. Then we develop a greedy algorithm that myopically determines users' strategies, based on the social group utility derived from only the users whose strategies have already been determined. It turns out that this algorithm can efficiently find a Pareto-optimal SNE with social welfare higher than that for the socially-oblivious PCG, pointing out the impact of exploiting social tie structure. We further show that the Pareto-optimal SNE can be achieved in a distributed manner.

The recent development of social services tightens wireless users' social relationships and encourages them to generate more data traffic under network effect. This boosts the demand for wireless services yet may challenge the limited wireless capacity. To fully exploit this opportunity, we study mobile users' data usage behaviors by jointly considering the network effect based on their social relationships in the social domain and the congestion effect in the physical wireless domain. Accordingly, we develop a Stackelberg game for problem formulation: In Stage I, a wireless provider first decides the data pricing to all users to maximize its revenue, and then in Stage II users observe the price and decide data usage subject to mutual interactions under both network and congestion effects. We analyze the two-stage game using backward induction. For Stage II, we first show the existence and uniqueness of a user demand equilibrium (UDE). Then we propose a distributed update algorithm for users to reach the UDE. Furthermore, we investigate the impacts of different parameters on the UDE. For Stage I, we develop an optimal pricing algorithm to maximize the wireless provider's revenue. We evaluate the performance of our proposed algorithms by numerical studies using real data, and thereby draw useful engineering insights for the operation of wireless providers.

Thanks to the convergence of pervasive mobile communications and fast-growing online social networking, mobile social networking is penetrating into our everyday life. Aiming to develop a systematic understanding of mobile social networks, in this paper we exploit social ties in human social networks to enhance cooperative device-to-device (D2D) communications. Specifically, as handheld devices are carried by human beings, we leverage two key social phenomena, namely social trust and social reciprocity, to promote efficient cooperation among devices. With this insight, we develop a coalitional game-theoretic framework to devise social-tie-based cooperation strategies for D2D communications. We also develop a network-assisted relay selection mechanism to implement the coalitional game solution, and show that the mechanism is immune to group deviations, individually rational, truthful, and computationally efficient. We evaluate the performance of the mechanism by using real social data traces. Simulation results corroborate that the proposed mechanism can achieve significant performance gain over the case without D2D cooperation.

In this paper, we develop a social group utility maximization (SGUM) framework for cooperative networking that takes into account both social relationships and physical coupling among users. Specifically, instead of maximizing its individual utility or the overall network utility, each user aims to maximize its social group utility that hinges heavily on its social ties with other users. We show that this framework provides rich modeling flexibility and spans the continuum space between non-cooperative game and network utility maximization (NUM) - two traditionally disjoint paradigms for network optimization. Based on this framework, we study an important application in database assisted spectrum access. We formulate the distributed spectrum access problem among white-space users with social ties as a SGUM game. We show that the game is a potential game and always admits a social-aware Nash equilibrium. We also design a distributed spectrum access algorithm that can achieve the social-aware Nash equilibrium of the game and quantify its performance gap. We evaluate the performance of the SGUM solution using real social data traces. Numerical results demonstrate that the performance gap between the SGUM solution and the NUM (social welfare optimal) solution is at most 15%.

This paper studies deterministic sensor deployment for barrier coverage in wireless sensor networks. Most of existing works focused on line-based deployment, ignoring a wide spectrum of potential curve-based solutions. We, for the first time, extensively study the sensor deployment under a general setting. We first present a condition under which the line-based deployment is suboptimal, revealing the advantage of curve-based deployment. By constructing a contracting mapping, we identify the characteristics for a deployment curve to be optimal. Based on the optimal deployment curve, we design sensor deployment algorithms by introducing a new notion of distance-continuous. Our findings show that i) when the deployment curve is distance-continuous, the proposed algorithm is optimal in terms of the vulnerability corresponding to the deployment, and ii) when the deployment curve is not distance-continuous, the approximation ratio of the vulnerability corresponding to the deployment by the proposed algorithm to the optimal one is upper bounded by min (π, ||ÃB̃||/||ÃG̃B̃|| 2n+√2-1/2n ), where ||ÃB̃|| and ||ÃG̃B̃|| are some constants, and n is the number of sensors. We generalize the study to the heterogeneous sensing model, and show that the proposed algorithm can provide close-to-optimal performance. Extensive numerical results corroborate our analysis.

Radar sensors, which actively transmit radio waves and collect RF energy scattered by objects in the environment, offer a number of advantages over purely passive sensors. An important issue in radar is that the transmitted energy may be scattered by objects that are not of interest as well as objects of interest (e.g., targets). The detection performance of radar systems is affected by such clutter as well as noise. Further, in many applications, clutter can be substantially stronger than the signals of interest. To combat the effect of clutter, a popular method is to take advantage of the Doppler frequency shift (DFS) extracted from the echo signal due to the relative motion of a target with respect to the radar. Unfortunately, a sensor coverage model that only depends on the distance to a target would fail to capture the DFS. In this paper, we set forth the concept of Doppler coverage for a network of spatially distributed radars. Specifically, a target is said to be Doppler-covered if, regardless of its direction of motion, there exists some radar in the network whose signalto-noise ratio (SNR) is sufficiently high and the DFS at that radar is sufficiently large. Based on the Doppler coverage model, we first propose an efficient method to characterize Dopplercovered regions for arbitrarily deployed radars. Then we design an algorithm for deriving the minimum radar density required to achieve Doppler coverage in a region under any polygonal deployment pattern, and further apply it to investigate the regular triangle based deployment.

This paper studies deterministic sensor deployment to ensure barrier coverage in wireless sensor networks. Most of existing work focused on line-based deployment, ignoring a wide spectrum of potential curve-based solutions. We, for the first time, extensively study the sensor deployment under general settings. We first present a condition under which line-based deployment is suboptimal, pointing to the advantage of curve-based deployment. By constructing a contracting mapping, we identify the characteristics for a deployment curve to be optimal. We then design sensor deployment algorithms for the optimal deployment curve by introducing a new notion of distance-continuous. Our findings show that i) when the deployment curve is distance-continuous, the proposed algorithm is optimal in terms of the vulnerability corresponding to the deployment, and ii) when the deployment curve is not distance-continuous, the approximation ratio of the vulnerability corresponding to the deployment by the proposed algorithm to the optimal one is upper bounded by min (π, ||AB||/||AGB|| 2n+√(2-1)/2n), where ||AB||, ||AGB|| and n are constants. Extensive numerical results corroborate our analysis.

Thanks to the convergence of pervasive mobile communications and fast-growing online social networking, mobile social networking is penetrating into our everyday life. Aiming to develop a systematic understanding of the interplay between social structure and mobile communications, in this paper we exploit social ties in human social networks to enhance cooperative device-to-device communications. Specifically, as hand-held devices are carried by human beings, we leverage two key social phenomena, namely social trust and social reciprocity, to promote efficient cooperation among devices. With this insight, we develop a coalitional game theoretic framework to devise social-tie based cooperation strategies for device-to-device communications. We also develop a network assisted relay selection mechanism to implement the coalitional game solution, and show that the mechanism is immune to group deviations, individually rational, and truthful. We evaluate the performance of the mechanism by using real social data traces. Numerical results show that the proposed mechanism can achieve up-to 122% performance gain over the case without D2D cooperation.

By taking advantage of active sensing using radio waves, radar sensors can offer several advantages over passive sensors. Although much recent attention has been given to multistatic and MIMO radar concepts, little has been paid to understanding the performance of radar networks (i.e., multiple individual radars working in concert). In this context, we study the optimal placement of a bistatic radar (BR) sensor network for barrier coverage. The coverage problem in a bistatic radar network (BRN) is challenging because: 1) in contrast to the disk sensing model of a traditional passive sensor, the sensing region of a BR depends on the locations of both the BR transmitter and receiver, and is characterized by a Cassini oval; 2) since a BR transmitter (or receiver) can potentially form multiple BRs with different BR transmitters (or receivers, respectively), the sensing regions of different BRs are coupled, making the coverage of a BRN highly non-trivial. This paper considers the problem of deploying a network of BRs in a region for maximizing the worst-case intrusion detectability, which amounts to minimizing the vulnerability of a barrier. We show that the shortest barrier-based placement is optimal if the shortest barrier is also the shortest line segment connecting the region's two boundaries. Based on this observation, we study the optimal placement of the BRs on a line segment for minimizing its vulnerability, which is a non-convex optimization problem. By exploiting some specific structural properties pertaining to the problem (particularly an important structure of detectability), we find the optimal placement order and the optimal placement spacing of the BR nodes, both of which exhibit elegant balanced structures. Our findings give valuable insight for the placement of BRs for barrier coverage. To our best knowledge, this is the first work to explore the coverage of a network of BRs.

We consider a bistatic radar network that consists of multiple separated radar transmitters and receivers, which are deployed to detect potential attacks at some points of interest (PoIs). To better defend these PoIs, the design of the bistatic radar network is investigated in two stages. First, we study the problem of optimally placing a number of radar transmitters and receivers in the sense of minimizing the maximum distance product between a PoI and its closest transmitter-receiver pair. For this problem, we propose a randomized Voronoi algorithm. Next, given the radars' locations, assuming that the transmitters use fixed and orthogonal frequencies to illuminate signals for interference avoidance, we study the problem of frequency selection for the receivers. Since an intelligent attacker can adaptively change the PoI to attack, the receivers should dynamically adapt their frequencies to cover different subsets of the PoIs. Accordingly, we model the dynamic interactions between the bistatic radar network and the attacker as a repeated security game. Based on their respective information, we propose two learning algorithms for each of them, respectively. We show that if both players follow the modified-regret-matching procedures, the empirical distributions of their actions converge to the set of correlated equilibria.

This paper considers opportunistic cooperative networking (OCN) in wireless ad hoc networks, with a focus on characterizing the desired tradeoff between the probing cost for establishing cooperative relaying and hence higher throughput via opportunistic cooperative networking. Specifically, opportunistic cooperative networking is treated as an optimal stopping problem with two-levels of incomplete information. Cases with or without dedicated relays are considered, and the existence of the optimal strategies for both cases are established. Then, it is shown that for the case with dedicated relays, the optimal strategy exhibits a threshold structure, in which it is optimal to probe the dedicated relay when the signal-to-noise ratio (SNR) of the source-relay link exceeds some threshold. For the case without dedicated relays, under more restrictive conditions, the optimal strategy is also threshold-based, in the sense that it is optimal to probe potential relays when the SNR of the source-destination link lies between two thresholds. Furthermore, these strategies can be implemented in a distributed manner.

This paper considers distributed opportunistic scheduling (DOS) with cooperative relaying in wireless ad hoc networks, with a focus on characterizing the desired tradeoff between the probing cost for establishing cooperative relaying and the higher throughput via opportunistic cooperative networking. Specifically, distributed scheduling and probing for cooperative relaying is treated as an optimal stopping problem with two levels of incomplete information. Cases with or without dedicated relays are considered, and the existence of the optimal strategies for both cases are established. Then, it is shown that for the case with dedicated relays, the optimal strategy exhibits a threshold structure, in which it is optimal to probe the dedicated relay when the signal-to-noise ratio (SNR) of the source-relay link exceeds some threshold. For the case without dedicated relays, under more restrictive conditions, the optimal strategy is also threshold-based, in the sense that it is optimal to probe potential relays when the SNR of the source-destination link lies between two thresholds. Furthermore, these strategies can be implemented in a distributed manner.

A problem of joint optimal bandwidth and power allocation in cognitive networks under fading channels is considered. It is assumed that multiple secondary users (SUs) share the spectrum of a primary user (PU) using frequency division multiple access. The bandwidth and power are allocated so as to maximize the sum ergodic capacity of all SUs under the total bandwidth constraint of the licensed spectrum as well as different combinations of the peak/average transmit power constraints at the SUs and the peak/average interference power constraint imposed by the PU. Although the optimization problem is convex, its dimension and, thus, complexity may be high. Therefore, computationally efficient ways of solving the problem are of importance and are investigated here by finding structures of the optimal solutions to the problem under different combinations of the constraints.

The resource allocation problem in wireless multi-user decode-and-forward (DF) relay networks is considered. The conventional resource allocation schemes based on the equal distribution of bandwidth and/or power may not be efficient for the networks with constrained/limited power and bandwidth resources at both sources and relays. Therefore, joint bandwidth and power allocation schemes are proposed based on (i) the maximization of the sum capacity of all users (source-destination pairs); (ii) the maximization of the worst user capacity; (iii) the minimization of the total power consumptions for all users. It is shown that the proposed problem formulations can be transformed to equivalent convex optimization problems. Therefore, the joint bandwidth and power allocation problems can be efficiently solved. The performance improvements offered by the proposed schemes are demonstrated by simulations.

In cognitive radio networks, secondary users (SUs) opportunistically exploit the spectrum unutilized by primary users (PUs). In this paper, we study the secondary communication where secondary transmitters and receivers have different available spectrum. Considering the spectrum diversity and the space distance between different PUs, we introduce cognitive relay node into the secondary communication and propose a novel Cooperative Relay Scheme (CRS) to increase the SINR at secondary receivers. A novel Opportunistic Sharing Scheme (OSS) is also proposed for the secondary transmitters to share the spectrum of relay nodes. We model it with a non-cooperative game, and study the performance of competition of SUs. The Nash equilibrium and Pareto efficiency of this game is presented. Simulations show that CRS can increase SINR at secondary receivers under proper configurations.