Federated Edge Learning (FEL) has emerged as a promising approach for enabling edge devices to collaboratively train machine learning models while preserving data privacy. Despite its advantages, practical FEL deployment faces significant challenges related to device constraints and device-server interactions, necessitating heterogeneous, user-adaptive model training with limited and uncertain communication. While knowledge cache-driven federated learning offers a promising FEL solution for demanding edge environments, its logits-based interaction design provides poor richness of exchanged information for on-device model optimization. To tackle this issue, we introduce DistilCacheFL, a novel personalized FEL architecture that enhances the exchange of optimization insights while delivering state-of-the-art performance with efficient communication. DistilCacheFL incorporates the benefits of both dataset distillation and knowledge cache-driven federated learning by storing and organizing distilled data as knowledge in the server-side knowledge cache, allowing devices to periodically download and utilize personalized knowledge for local model optimization. Moreover, a device-centric cache sampling strategy is introduced to tailor transferred knowledge for individual devices within controlled communication bandwidth. Extensive experiments on five datasets covering image recognition, audio understanding, and mobile sensor data mining tasks demonstrate that (1) DistilCacheFL significantly outperforms state-of-the-art methods regardless of model structures, data distributions, and modalities. (2) DistilCacheFL can train splendid personalized on-device models with at least 28.6 improvement in communication efficiency.

In federated learning (FL), although the original intention of “available but not visible” data is to allay data privacy concerns, it potentially brings new security threats, particularly poisoning attacks that target such “not visible” local data. Intuitively, such data poisoning attacks have great potential in stealthily degrading global FL outcomes, and are expected to be even stealthier if being enhanced by generative models like generative adversarial networks (GANs). However, existing defense methods have not been thoroughly challenged in this regard and generally fail to be aware of a local generation of seemingly legitimate poisoned data. With a growing concern on potentially stealthier attacks, in this paper, a cost-effective defense mechanism named Model Consistency-Based Defense (MCD) is proposed, which offers a comprehensive examination of available local models across multiple feature dimensions, providing an indirect yet effective means of identifying hidden data poisoning attackers. To push the limit of MCD against stealthier attacks, we propose a new GAN-based data poisoning attack model named VagueGAN and an unsupervised variant of it, which can be flexibly deployed to generate seemingly legitimate but noisy poisoned data. The consistency of GAN outputs revealed by VagueGAN helps strengthen MCD to work against stealthier GAN-based attacks as well as other mainstream ones. Extensive experiments on multiple open datasets (MNIST, Fashion-MNIST, CIFAR-10, CIFAR-100, and Mini-Imagenet) indicate that our attack method better balances the trade-off between attack effectiveness and stealthiness with low complexity. More importantly, our defense mechanism is shown to be more competent in identifying a variety of poisoned data, particularly stealthier GAN-poisoned ones.

Heterogeneous device-edge-cloud computing infrastructures have become the backbone of modern telecommunication operators and Wide Area Networks (WANs), providing multi-tier computational support for emerging intelligent applications. With the rapid proliferation of Large Language Model (LLM) services, efficiently coordinating inference tasks and reducing communication burden within these multi-tier network architectures becomes a critical deployment challenge. Current LLM serving paradigms exhibit significant limitations: on-device deployment restricts service to lightweight LLMs due to hardware constraints, while cloud-centric deployment encounters resource congestion and considerable prompt communication overhead during peak periods. Model-cascading inference, though better suited for multi-tier networks, depends on static, manually-tuned thresholds that cannot adapt to dynamic network conditions or varying task complexities. To address these challenges, we propose RecServe, a recursive offloading framework tailored for LLM serving in multi-tier networks. RecServe introduces a task-specific hierarchical confidence evaluation mechanism that guides offloading decisions based on inferred task complexity in progressively scaled LLMs across device, edge, and cloud tiers. To further enable intelligent task routing across tiers, RecServe employs a sliding-window-based dynamic offloading strategy with quantile interpolation, enabling real-time tracking of historical confidence distributions and adaptive offloading threshold adjustments. This design allows inference tasks to be recursively offloaded to higher tiers only when necessary, optimizing heterogeneous resource utilization while reducing cross-tier communication with little compromise on service quality. Theoretical analysis provides distinct conditions under which RecServe is expected to achieve reduced communication burden and computational costs. Experiments on eight datasets demonstrate that RecServe outperforms CasServe in both service quality and communication efficiency, and reduces the communication burden by over 50% compared to centralized cloud-based serving.

Federated Distillation (FD) is an innovative distributed machine learning paradigm that enables efficient and flexible cross-device knowledge transfer through knowledge distillation, without the need to upload large-scale model parameters to a central server. Although FD has attracted increasing attention in recent years, its security aspects remain relatively underexplored. Existing attack methods targeting traditional federated learning mainly focus on the transmission of model parameters and gradients, while attacks specifically designed for the unnormalized outputs (logits) in the emerging FD paradigm are still lacking. To fill this research gap and contribute to the enhancement of FD’s security, we previously proposed the Federated Distillation Logits Attack (FDLA), which manipulates the logits transmitted during communication to mislead and degrade the performance of client models. However, FDLA has limitations in controlling its impact on participants with different roles or identities and lacks a systematic investigation into the effects of malicious interventions at various stages of knowledge transfer. To overcome these limitations, we propose a more advanced and controllable logits poisoning method—Peak-Controlled Federated Distillation Logits Attack (PCFDLA). PCFDLA enhances the effectiveness of FDLA by precisely controlling the peak values of logits to adjust the intensity of the attack. This method generates highly misleading perturbations that achieve stronger attack performance while maintaining a similar level of stealthiness to FDLA when detection is based on differences in model parameters. Moreover, we introduce a novel evaluation metric to more comprehensively assess the performance of such attacks. Experimental results show that PCFDLA significantly increases the destructive impact on victim models while maintaining high stealth. It consistently achieves superior performance across multiple datasets, highlighting its potential threat to the security of federated distillation systems.

In the typical radio frequency (RF)-based wireless power transfer (WPT) system, the wireless power station (WPS) connected to the grid transmits energy to charge low-power sensors via radio signals. Such a system may not be green and also difficult to deploy in some special areas including deserts and mountainous areas, because it depends on the grid. To achieve a green RF WPT system design, this paper considers that the WPS is powered by natural energy sources rather than the grid. To explore the maximal total amount of the energy that can be harvested by the sensors, we focus on the offline setting, so similar to many existing works on offline optimization, we assume that the WPS knows prior knowledge about energy arrivals and channel changes, and then formulate an optimization problem to maximize the total harvested energy via optimizing the WPS’s time-domain transmit power subject to multiple constraints, including the finite battery capacity at the WPS, the causal relationship between the natural energy harvesting and the WPT, and the transmit power budget of the WPS, where for practicality, the nonlinear energy harvesting (EH) model is also taken into account. To solve this non-convex problem, we first equivalently transform it by using the epigraph reformulation and the variable substitution, and then use the first-order Taylor expansion to get an approximate convex version. Then, we present a successive convex approximation (SCA)-based algorithm to improve the accuracy of the obtained solution for approaching the optimal one. For the special case with a single sensor, we further propose a branch and bound (BB)-based algorithm that is able to get a more accurate solution with lower complexity than the SCA-based one. Numerical results demonstrate that the proposed algorithms are able to achieve the near-global optimal solution. As the average recharge rate increases, compared with the other two baselines, i.e., the greedy power (GP) policy and the constant power (CP) policy, the total harvested energy achieved by the SCA-based algorithm is up to about 2.48 times and 1.37 times that of the baselines respectively. For the single-sensor case, the BB-based algorithm always outperforms the SCA-based one in terms of the total harvested energy while reducing the running time required for solving by about 90% on average.

Federated Learning (FL) enables training Artificial Intelligence (AI) models over end devices without compromising their privacy. As computing tasks are increasingly performed by a combination of cloud, edge, and end devices, FL can benefit from this End-Edge-Cloud Collaboration (EECC) paradigm to achieve collaborative device-scale expansion with real-time access. Although Hierarchical Federated Learning (HFL) supports multitier model aggregation suitable for EECC, prior works assume the same model structure on all computing nodes, constraining the model scale by the weakest end devices. To address this issue, we propose Agglomerative Federated Learning (FedAgg), which is a novel EECC-empowered FL framework that allows the trained models from end, edge, to cloud to grow larger in size and stronger in generalization ability. FedAgg recursively organizes computing nodes among all tiers based on Bridge Sample Based Online Distillation Protocol (BSBODP), which enables every pair of parent-child computing nodes to mutually transfer and distill knowledge extracted from generated bridge samples. This design enhances the performance by exploiting the potential of larger models, with privacy constraints of FL and flexibility requirements of EECC both satisfied. Experiments under various settings demonstrate that FedAgg outperforms state-of-the-art methods by an average of 4.53% accuracy gains and remarkable improvements in convergence rate. Our code is available at https://github.com/wuzhiyuan2000/FedAgg.

This article investigates rate-splitting multiple access (RSMA)-based visible light communication (VLC) networks with simultaneous lightwave information and power transfer (SLIPT). To effectively enhance the fairness among information decoding users (IDUs), we formulate an optimization problem to maximize the minimum data rate by optimizing the direct current bias vector, the common message rates of RSMA, and the transmit precoding vectors. In the problem, the IDUs’ minimum energy harvesting (EH) requirements, the total power budget of the light-emitting diode (LED) transmitters, and the linear operation region of LEDs are also considered as the system constrains. To solve the formulated nonconvex problem, epigraph reformulation is first employed to transform the nonconvex objective function. Then, a series of transformations is proposed and the semi-definite relaxation (SDR) method is adopted to address the rank-one precoding matrix constraint. After that, an iterative algorithm is proposed to obtain an effective suboptimal solution by applying the successive convex approximation. Extensive simulations show that the max–min rate (MMR) is inversely proportional to the number of IDUs and it decreases as the minimum EH requirement becomes more stringent, especially in the high-EH region. Moreover, the value of the maximum drive current imposes a significant impact on the system performance, particularly, the MMR becomes saturated for a given maximum drive current even if the total power budget is sufficient. Besides, RSMA can contribute to both spectral efficiency and energy efficiency greatly in comparison to the traditional multiple access scheme.

This article investigates simultaneous transmitting and reflecting reconfigurable intelligent surface (STAR-RIS)-assisted downlink multiuser multiple-input–single-output (MU-MISO) networks with the rate splitting multiple access (RSMA) scheme. A base station (BS) desires to simultaneously transmit messages to multiple users with the assistance of an STAR-RIS to enhance communication quality as well as extend the coverage of users. An optimization problem is formulated to maximize the achievable sum rate of the networks on the premise of satisfying the constraints on power budget at the BS, total common-stream rate of users, and individual users’ minimum rate requirements, via jointly optimizing the beamforming vectors, the common-stream rate allocation vector, and the transmission and reflection coefficients (TARCs) matrix. Due to the dynamic changes of communication links and the coupling of multiple variables, it is challenging to solve such a nonconvex optimization problem by utilizing traditional methods. Therefore, a proximal policy optimization (PPO)-based deep reinforcement learning (DRL) algorithm is proposed, where the reward function, the action space and the state space are designed properly. A constraint-satisfaction-processing (CSP) method is employed to further adjust the optimized transmit power to make sure that the obtained optimized results satisfy the power budget constraint. Simulation results show that the proposed PPO-based DRL algorithm converges well and achieves much better performance than several baselines, such as the soft actor–critic (SAC), the deep deterministic policy gradient (DDPG), the genetic algorithm (GA), the maximum ratio transmission (MRT), the zero-forcing (ZF), and the random methods. Moreover, it demonstrates that deploying STAR-RIS greatly enhances the system sum rate and user fairness compared to deploying traditional reflecting-only RIS (RO-RIS) and without RIS. Besides, it also shows that adopting the RSMA scheme achieves more notable performance gains than the nonorthogonal multiple access (NOMA) scheme in such a network.

Traffic flow forecasting (TFF) is of great importance to the construction of Intelligent Transportation Systems. To mitigate communication burden and tackle with the problem of privacy leakage aroused by centralized forecasting methods, Federated Learning (FL) has been applied to TFF. However, existing FL-based approaches employ batch learning manner, which makes the pre-trained models inapplicable to subsequent traffic data, thus exhibiting subpar prediction performance. In this paper, we perform the first study of forecasting traffic flow adopting online learning manner in FL framework and then propose a novel prediction method named Online Spatio-Temporal Correlation-based Federated Learning (FedOSTC), aiming to guarantee performance gains regardless of traffic fluctuation. Specifically, clients employ Gated Recurrent Unit (GRU)-based encoders to obtain the internal temporal patterns inside traffic data sequences. Then, the central server evaluates spatial correlation among clients via Graph Attention Network (GAT), catering to the dynamic changes of spatial closeness caused by traffic fluctuation. Furthermore, to improve the generalization of the global model for upcoming traffic data, a period-aware aggregation mechanism is proposed to aggregate the local models which are optimized using Online Gradient Descent (OGD) algorithm at clients. We perform comprehensive experiments on two real-world datasets to validate the efficiency and effectiveness of our proposed method and the numerical results demonstrate the superiority of FedOSTC.

This paper investigates the energy-efficient coordinated beamforming design for multi-pair multiple-input single-output (MISO) networks with passive eavesdroppers. To be practical, it is assumed that only channel distribution information (CDI) of the network is known by the transmitters/sources, and the dynamic energy consumption model (DECM) is employed. In order to achieve a green network design, an energy efficiency (EE) maximization problem is formulated subjecting to the individual available power constraints, the rate outage probability constraints, and the information leakage probability constraints. To solve the formulated non-convex problem, semidefinite relaxation (SDR) and first-order lower bound are applied to transform the problem, and then an efficient algorithm is proposed based on successive convex approximation (SCA) and Dinkelbach's approaches. The proposed algorithm is theoretically proved to converge to a stationary point of the considered problem. Further, a distributed version of the proposed algorithm is designed, with which each transmitter is able to optimize its own beamforming vector with local CDI. Moreover, the computational complexities and the signaling overheads of the two developed algorithms are analyzed and compared. Simulation results show that both algorithms achieve good EE performance, and the EE performance achieved by the distributed algorithm is very similar to that achieved by the centralized one. Additionally, it is shown that similar to the conventional scenarios without eavesdroppers, the achieved system EE also has a saturation point w.r.t. the available power of the transmitters, and by employing our proposed algorithms, the network security is significantly enhanced.

Edge Intelligence (EI) allows Artificial Intelligence (AI) applications to run at the edge, where data analysis and decision-making can be performed in real-time and close to data sources. To protect data privacy and unify data silos distributed among end devices in EI, Federated Learning (FL) is proposed for collaborative training of shared AI models across multiple devices without compromising data privacy. However, the prevailing FL approaches cannot guarantee model generalization and adaptation on heterogeneous clients. Recently, Personalized Federated Learning (PFL) has drawn growing awareness in EI, as it enables a productive balance between local-specific training requirements inherent in devices and global-generalized optimization objectives for satisfactory performance. However, most existing PFL methods are based on the Parameters Interaction-based Architecture (PIA) represented by FedAvg, which suffers from unaffordable communication burdens due to large-scale parameters transmission between devices and the edge server. In contrast, Logits Interaction-based Architecture (LIA) allows to update model parameters with logits transfer and gains the advantages of communication lightweight and heterogeneous on-device model allowance compared to PIA. Nevertheless, previous LIA methods attempt to achieve satisfactory performance either relying on unrealistic public datasets or increasing communication overhead for additional information transmission other than logits. To tackle this dilemma, we propose a knowledge cache-driven PFL architecture, named FedCache, which reserves a knowledge cache on the server for fetching personalized knowledge from the samples with similar hashes to each given on-device sample. During the training phase, ensemble distillation is applied to on-device models for constructive optimization with personalized knowledge transferred from the server-side knowledge cache. Empirical experiments on four datasets demonstrate that FedCache achieves comparable performance with state-of-art PFL approaches, with more than two orders of magnitude improvements in communication efficiency. Our code and DEMO are available athttps://github.com/wuzhiyuan2000/FedCache.

The growing interest in intelligent services and privacy protection for mobile devices has given rise to the widespread application of federated learning in Multi-access Edge Computing (MEC). Diverse user behaviors call for personalized services with heterogeneous Machine Learning (ML) models on different devices. Federated Multi-task Learning (FMTL) is proposed to train related but personalized ML models for different devices, whereas previous works suffer from excessive communication overhead during training and neglect the model heterogeneity among devices in MEC. Introducing knowledge distillation into FMTL can simultaneously enable efficient communication and model heterogeneity among clients, whereas existing methods rely on a public dataset, which is impractical in reality. To tackle this dilemma,Federated MultI-task Distillation for Multi-access EdgeCompuTing (FedICT) is proposed. FedICT direct local-global knowledge aloof during bi-directional distillation processes between clients and the server, aiming to enable multi-task clients while alleviating client drift derived from divergent optimization directions of client-side local models. Specifically, FedICT includes Federated Prior Knowledge Distillation (FPKD) and Local Knowledge Adjustment (LKA). FPKD is proposed to reinforce the clients' fitting of local data by introducing prior knowledge of local data distributions. Moreover, LKA is proposed to correct the distillation loss of the server, making the transferred local knowledge better match the generalized representation. Extensive experiments on three datasets demonstrate that FedICT significantly outperforms all compared benchmarks in various data heterogeneous and model architecture settings, achieving improved accuracy with less than 1.2% training communication overhead compared with FedAvg and no more than 75% training communication round compared with FedGKT in all considered scenarios.

Federated learning (FL) is a privacy-preserving distributed learning paradigm relying on but without directly accessing privately owned datasets. However, the ‘‘available but not visible’’ nature of training data in FL leads to security risks. In particular, ‘‘not visible’’ local data can easily become the best targets of poisoning attacks. Although existing data poisoning methods may successfully attack FL systems, they mostly lead to significant data statistical changes and thus can be not hard to detect. In this paper, we propose VagueGAN, a new data poisoning attack model that unconventionally leverages the power of generative adversarial network (GAN) to generate seemingly legitimate vague data with appropriate amounts of poisonous noise. The quality of such vague data can be controlled on demand to achieve a balanced trade-off between attack effectiveness and stealthiness. Extensive experiments show that data poisoning attacks enhanced by our VagueGAN not only better degrade FL outcomes with low efforts but also are generally much less detectable.

The emergence of edge computing (EC) and artificial intelligence (AI) is driving the rapid growth of industrial internet of things (IIoT). However, few works comprehensively consider the impact of resource reallocation and number of reallocation on the system delay in dynamic industrial scenarios with time-varying geographic location characteristics. This paper takes the dynamic resource reallocation problem between the physical layer and edge layer within a time-varying factory scenario into account, proposes a reallocation-decision variable and reduces the computational stress on edge nodes caused by frequent reallocation. An optimization problem with the objective of minimizing the system average delay is established and a proximal policy optimization (PPO) based dynamic resource reallocation (P-DRR) algorithm is proposed for the problem solving. Experimental results show that P-DRR algorithm can effectively reduce average delay compared to the baseline algorithms without causing large computational pressure on edge nodes.

The participatory nature of federated learning (FL) makes it attractive for fingerprinting-based indoor localization in multibuilding and multifloor environments. A group of sensing clients can collaboratively leverage their private, local fingerprint data to help their edge server update a location prediction model. However, it is challenging to jointly handle the two involved issues, i.e., building-floor classification (BFC) and latitude–longitude regression (LLR), in a wide 3-D space through enabling FL on decentralized yet heterogeneous data and over an imperfect wireless network. In this article, we confront these challenges and propose an FL framework, FedLoc3D, for both BFC and LLR. Specifically, the former issue is addressed by an FedDSC-BFC approach, which generates a multilabel classification model based on a convolutional neural network with depthwise separable convolutions. The latter issue is addressed by an FedADA-LLR approach, which develops a multitarget regression model based on a deep neural network with autoencoder and data augmentation. Extensive experiments on a real-world data set of WiFi fingerprints are carried out, and our approaches with enhanced capabilities of feature extraction, generalization, and convergence are validated to improve both localization accuracy and learning efficiency under data heterogeneity and network instability.

This paper investigates a multi-user wireless powered fog computing (FC) network, where multiple energy-limited wireless sensor devices (WSDs) first harvest energy from a nearby hybrid access point (HAP), and then compute their tasks locally (i.e., the local computing (LC) mode) or offload the tasks to the HAP (i.e., the FC mode) via a binary offloading policy. In order to pursue the green computing network design, an optimization problem is formulated to minimize the transmit power at the HAP by jointly optimizing the time allocation ratio and the computing mode selection vector, under the energy causality constraints and the WSDs’ computing rate requirements constraints. To efficiently solve the formulated non-convex problem in a distributed manner, it is first transformed into an approximate form, and then an alternating direction method of multipliers (ADMM)-based algorithm is designed to solve the transformed problem, based on which the successive convex approximation (SCA) is adopted to improve the approximating precision in an iterative way. With the proposed ADMM-based distributed algorithm, each WSD is able to optimize its computing mode and offloading time with local channel state information (CSI), which thus is more suitable for large-scale networks. For comparison, a channel-sorting-based (CSB) centralized algorithm with global CSI is also presented, and the computational complexities of the proposed ADMM-based algorithm and the CSB algorithm are analyzed. Simulation results show that the proposed distributed algorithm achieves a comparable performance with the CSB centralized algorithm and the exhaustive search method. It is also observed that to minimize the transmit power at the HAP, the WSDs with the better channel quality are inclined to select the LC mode, which is much different from traditional sum-computation-rate maximization design.

In multiservice systems, multiple different Age of Information (AoI) penalty functions and corresponding algorithms are required to be deployed, which may result in high deployment complexity. Motivated by this, we propose a universal function$f(t)=\beta e^{\alpha t} -\beta $called$\alpha $-$\beta $AoI penaltyfunction to characterize different nonlinear forms of AoI penalty. With the presented$\alpha $-$\beta $AoI penalty function, we analyze the performance of wireless-powered communication networks (WPCNs), where a sensor first harvests energy from a wireless power station (WPS) and then transmits the generated update to its data collector. The sensor is equipped with a battery of limited energy capacity. When the battery of the sensor node is fully charged, the sensor generates a status update and uses all available energy to transmit it. A closed-form expression of the system average$\alpha $-$\beta $AoI penalty is derived by using some limit methods. In order to minimize the average$\alpha $-$\beta $AoI penalty of the system, an optimization problem is formulated to optimize the battery capacity. Simulation results demonstrate the correctness of our theoretical analysis results and show that there is a unique optimal battery capacity that optimizes the system AoI performance. Moreover, when the system is with the exponential-shape AoI penalty function ($\beta >0$and$\alpha >0$), with the increment of$\alpha $and$\beta $increase, the average$\alpha $-$\beta $AoI penalty also increases. Differently, when the system is with the logarithmic-shape AoI penalty function ($\beta < 0$and$\alpha < 0$), with the increment of$\alpha $and$\beta $, the average$\alpha $-$\beta $AoI penalty decreases.

This paper proposes a multi-agent double deep Q network (DDQN)-based approach to jointly optimize the beamforming vectors and power splitting (PS) ratio in multi-user multiple-input single-output (MU-MISO) simultaneous wireless information and power transfer (SWIPT)-enabled heterogeneous networks (HetNets), where a macro base station (MBS) and several femto base stations (FBSs) serve multiple macro user equipments (MUEs) and femto user equipments (FUEs). The PS receiver architecture is deployed at FUEs. An optimization problem is formulated to maximize the achievable sum information rate of FUEs under the constraints of the achievable information rate requirements of MUEs and FUEs and the energy harvesting (EH) requirements of FUEs. Since the optimization problem is challenging to handle due to the high dimension and time-varying environment, an efficient multi-agent DDQN-based algorithm is presented, which is trained in a centralized manner and runs in a distributed manner, where two sets of deep neural network parameters are jointly updated and trained to tackle the problem and avoid overestimation. To facilitate the presented multi-agent DDQN-based algorithm, the action space, the state space and the reward function are designed, where the codebook matrix is employed to deal with the complex transmit beamforming vectors. Simulation results validate the proposed algorithm. Notable performance gains are achieved by the proposed algorithm due to considering the beam directions in the action space and the adaptability to the Doppler frequency shifts. Besides, the proposed algorithm is shown to be superior to other benchmark ones numerically.

On-demand deployment of small cells plays a key role in augmenting macro-cell coverage for outdoor hotspots, where user devices are brought together and intensively upload self-generated data. In this paper, we study spectrum sharing among rapidly deployable small cells in the uplink, even without a priori global knowledge. We propose a hybrid multi-agent approach, which allows a leading macro-cell base station (MBS) and multiple following small base stations (SBSs) to take part in a user-centric, online joint optimization of small cell deployment and uplink resource allocation. Specifically, we propose a centralized mechanism for the MBS to solve the first subproblem of small cell deployment stage by stage, based on an adversarial bandit model. Furthermore, we propose a distributed mechanism for the group of SBSs to collectively solve the second subproblem of uplink resource allocation stage by stage, based on a stochastic game model. We prove that our approach is guaranteed to produce a joint strategy, which is built upon a mixed strategy with bounded regret on the first tier and an equilibrium solution on the second tier. Our approach is validated by simulations on the aspects of convergence behavior, strategy correctness, power consumption, and spectral efficiency.

The legacy concept of exclusion zones (EZs) is inept at enabling efficient utilization of fallow spectrum by secondary users (SUs), since legacy EZs are static and overly-conservative. The notion of a static EZ implies that it has to protect incumbent users (IUs) from the union of likely interference scenarios, leading to a worst-case, conservative solution. In this paper, we propose the concept of dynamic, multi-tier EZs, which takes advantage of participatory spectrum sensing carried out by SUs to support efficient database-driven spectrum sharing while protecting IUs against SU-induced aggregate interference. Specifically, the database directly incentivizes SUs to participate in spectrum sensing, which augments geolocation database by defining smaller EZs with dynamic boundaries and creating additional spectrum access opportunities for SUs. We propose an incentive mechanism based on a two-level game-theoretic model, in which the database conducts dynamic pricing in a first-level Stackelberg game in the presence of SUs who strategically contribute to spectrum sensing in a second-level stochastic game. The existence of an equilibrium solution is proven. According to our findings, the proposed incentive mechanism for the concept of dynamic, multi-tier EZs is effective to improve spectrum utilization efficiency while guaranteeing incumbent protection.

The coexistence of cognitive radio (CR) networks in the same swath of spectrum has become an increasingly important problem, which is especially challenging when coexisting networks are heterogeneous (i.e., use different air interface standards), such as the case in TV white spaces. In this paper, we propose a credit-token-based spectrum etiquette framework that enables spectrum sharing among distributed heterogeneous CR networks with equal priority. Specifically, we propose a game-auction coexistence framework. Each network acts as either an offerer or a requester, and coexists with other networks via a non-cooperative game and a truthful multi-winner auction. The framework addresses the trade-offs among social welfare and offerer's revenue in the auction and requester's utility in the game. We prove that the framework guarantees system stability. Our simulation results show that the proposed coexistence framework always converges to a near-optimal distributed solution and improves coexistence fairness and spectrum utilization.

In database-driven opportunistic spectrum access, location information of secondary users plays an important role. In a database query-and-update procedure, a secondary user reports to the geolocation database of its location information, so that the updated knowledgebase facilitates location-aided incumbent protection and network coexistence. However, such database-driven spectrum sharing becomes very challenging when the secondary users are mobile. In this paper, we propose a probabilistic coexistence framework that supports mobile users by incorporating the solutions to solve two core problems: (i) white space allocation (WSA) at the database and (ii) location update control (LUC) at the users. We frame the two problems such that they interact through dynamic control of the users' location uncertainty levels. For WSA, we derive a centralized real-time solution to mitigate mutual interference among secondary users and protect primary users against harmful interference. For LUC, we design a local two-level strategy to enable both movement-driven and interference-driven control of location uncertainty. This strategy makes an appropriate trade-off between the effectiveness of interference mitigation and the cost of database queries. To evaluate our algorithms, we have carried out both theoretical model-driven and real-world trace-driven simulation experiments. Our simulation results show that the proposed framework can determine and adapt the database query intervals of mobile users to achieve near-optimal interference mitigation with minimal location updates.

The depletion of usable radio frequency spectrum has stimulated increasing interest in dynamic spectrum access technologies, such as cognitive radio (CR). In a scenario where multiple co-located CR networks operate in the same swath of white-space (or unlicensed) spectrum with little or no direct coordination, co-channel self-coexistence is a challenging problem. In this paper, we focus on the problem of spectrum sharing among coexisting CR networks that employ orthogonal frequency division multiple access (OFDMA) in their uplink and do not rely on inter-network coordination. An uplink soft frequency reuse (USFR) technique is proposed to enable globally power-efficient and locally fair spectrum sharing. We frame the self-coexistence problem as a non-cooperative game. In each network cell, uplink resource allocation (URA) problem is decoupled into two subproblems: subchannel allocation (SCA) and transmit power control (TPC). We provide a unique optimal solution to the TPC subproblem, while presenting a low-complexity heuristic for the SCA subproblem. After integrating the SCA and TPC games as the URA game, we design a heuristic algorithm that achieves the Nash equilibrium in a distributed manner. In both multi-operator and single-operator coexistence scenarios, our simulation results show that USFR significantly improves self-coexistence in spectrum utilization, power consumption, and intra-cell fairness.

Recent advances in cognitive radio (CR) technology have brought about a number of wireless standards that support opportunistic access to available white-space spectrum. Addressing the self-coexistence of CR networks in such an environment is very challenging, especially when coexisting networks operate in the same swath of spectrum with little or no direct coordination. In this paper, we study the problem of co-channel self-coexistence of uncoordinated CR networks that employ orthogonal frequency division multiple access (OFDMA) in the uplink. We frame the self-coexistence problem as a non-cooperative game, and propose an uplink soft frequency reuse (USFR) technique to enable globally power-efficient and locally fair sharing of white-space spectrum. In each network, uplink resource allocation is decoupled into two subproblems: subchannel allocation (SCA) and transmit power control (TPC). We provide a unique optimal solution to the TPC subproblem, and present a low-complexity heuristic for the SCA subproblem. Furthermore, we frame the TPC and SCA games, and integrate them as a heuristic algorithm that achieves the Nash equilibrium in a fully distributed manner. Our simulation results show that the proposed USFR technique significantly improves self-coexistence in several aspects, including spectrum utilization, power consumption, and intra-cell fairness.

In cognitive radio (CR) networks, spectrum resource that can be shared by secondary users (SUs) is always restricted by primary users (PUs). Although channel aggregation (CA) enables each SU to access multiple channels at a time, whether it is beneficial is subject to the PU activity and radio capability. In this paper, we study the feasibility and efficiency of CA in consideration of various such practical constraints and costs. First, we propose a novel channel usage model to analyze the impact of both PU and SU behaviors on the availability of white spaces. This model is very general and can capture a wide range of user behaviors. Next, we model the costs in time for performing CA. User demands in both frequency and time domains are considered to evaluate the costs for making negotiation and renewing transmission. Further, an optimal CA strategy is defined to minimize the cumulative delay for transmitting a certain amount of data. Numerical and simulation results based on real data of PU activity show that user demands on both bandwidth and duration should be carefully chosen to achieve the optimal delay performance in practice.