We consider decentralized federated learning (DFL) in mobile computing networks (MCNs), where dynamically changing neighborhood sets among devices arise from mobility and environmental perturbations. The time-varying topology coupled with inherent system heterogeneity poses significant challenges to achieve stable and efficient convergence in DFL. However, existing studies rarely consider both dynamic connectivity and statistical heterogeneity. To close this gap, this paper proposes a novel DFL framework enhanced with topology learning (DFL-TL) to mitigate the spatio-temporal volatility induced by MCNs, where each mobile device faces coupled constraints on its temporal windows for local updates and spatial scopes for model interaction. We introduce a new bounded neighborhood heterogeneity to jointly measure and constrain both the inter-device heterogeneity and the spectral properties of the topologies. Additionally, we formulate a mixed-integer nonlinear programming (MINLP) problem to jointly optimize learning costs and neighborhood heterogeneity. Through problem decomposition, DFL-TL efficiently identifies optimal resource allocations and adaptive mixing matrices, thereby enabling the selection of optimal training time windows while reducing the adverse effects of dynamic topologies in heterogeneous networks. Furthermore, we establish the iteration complexity of DFL-TL under non-convex settings and show that solving the proposed MINLP formulation leads to a tighter convergence bound. Extensive experiments demonstrate that DFL-TL achieves a faster convergence performance and reduces the wall-clock training time compared to the state-of-the-art baselines.

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

Ambient backscatter communication (AmBC) enables ultra-low-power, low-cost and massive connectivity. However, practical AmBC systems suffer from symbol timing offset (STO) due to propagation delay and backscatter receiver (BR) activation latency, while conventional correlation-based synchronization methods are inapplicable because ambient radio frequency sources are non-cooperative. Moreover, residual STO (RSTO) inevitably remains due to the finite synchronization sequence, which degrades symbol detection performance. To address these challenges, we first design a specialized synchronization sequence with alternating “0” and “1” bits at the backscatter device to induce observable sampling errors at the BR. Based on this, we propose a pilot-aided, sampling-error-aware maximum likelihood estimation (PSE-MLE) method for STO estimation and compensation, which exploits the statistical variations in the received synchronization signal. After STO compensation, the remaining RSTO is statistically modeled as a discrete bilateral Laplace distribution, with its parameter estimated via ridge regression. Leveraging this prior information, we further develop a Bayesian average energy detector (ave-ED) and derive closed-form expressions for both the detection threshold and bit error rate. Simulation and experimental results on a practical AmBC platform validate the effectiveness of the proposed methods.

Low Earth orbit (LEO) satellite networks are expected to support heterogeneous services, including enhanced mobile broadband (eMBB) communications, massive machine-type communications (mMTC), and ultra-reliable low-latency communications (URLLC). However, existing coexistence schemes, such as puncturing and superposition, struggle to achieve an effective trade-off among reliability, latency, and spectral efficiency due to their limited degrees of freedom (DoF). To address this challenge, we propose a novel rate-splitting grant-free (RS-GF) transmission scheme that integrates rate-splitting multiple access (RSMA) with grant-free random access (GF-RA) to efficiently support heterogeneous quality of service (QoS) requirements. The high-rate eMBB user employs single-layer rate splitting (RS) over the entire slot, while short-packet Internet-of-Things (IoT) devices associated with URLLC and mMTC adopt GF-RA via single mini-slot transmissions. Building on this RS-GF framework, we analyze the outage performance of the proposed scheme. Specifically, we derive the average packet error probability (PEP) of IoT devices in the finite blocklength (FBL) regime and analyze the eMBB user’s outage probability under imperfect successive interference cancellation (SIC) and mini-slot collisions. On this basis, we present simplified analytical solutions for sparse and dense IoT deployment scenarios, and Monte Carlo simulations validate our analytical derivations. Simulation results demonstrate that the proposed RS-GF scheme outperforms state-of-the-art solutions for service coexistence in LEO satellite networks.

The development of the Open RAN (O-RAN) framework helps enable network slicing through its virtualization, interoperability, and flexibility. To improve spectral efficiency and better meet users’ dynamic and heterogeneous service demands, O-RAN’s flexibility further presents an opportunity for resource reselling of unused physical resource blocks (PRBs) across users. In this work, we propose a novel game-based user-to-user PRB reselling model in the O-RAN setting, which models the carryover of unmet demand across time slots, along with how users’ internal buffer states relate to any PRBs purchased. We formulate the interplay between the users as a strategic game, with each participant aiming to maximize their own payoffs, and we prove the existence and uniqueness of Nash equilibrium (NE) in the game. We furthermore propose an iterative bidding mechanism that converges to this NE. Extensive simulations show that our best approach reduces data loss by 30.5% and spectrum resource wastage by 50.7% while significantly improving social welfare, compared to its absence.

Large-scale low-Earth-orbit (LEO) satellite systems are increasingly valued for their ability to enable rapid and wide-area data exchange, thereby facilitating the collaborative training of artificial intelligence (AI) models across geographically distributed regions. Due to privacy concerns and regulatory constraints, raw data collected at remote clients cannot be centrally aggregated, posing a major obstacle to traditional AI training methods. Federated learning offers a privacy-preserving alternative by training local models on distributed devices and exchanging only model parameters. However, the dynamic topology and limited bandwidth of satellite systems will hinder timely parameter aggregation and distribution, resulting in prolonged training times. To address this challenge, we investigate the problem of scheduling federated learning over satellite networks and identify key bottle-necks that impact the overall duration of each training round. We propose a discrete temporal graph–based on-demand scheduling framework that dynamically allocates communication resources to accelerate federated learning. Simulation results demonstrate that the proposed approach achieves significant performance gains over traditional statistical multiplexing-based model exchange strategies, reducing overall round times by 14.20% to 41.48%. Moreover, the acceleration effect becomes more pronounced for larger models and higher numbers of clients, highlighting the scalability of the proposed approach.

Federated learning (FL) allows multiple clients to collaboratively train a model without sharing their private data. In practical scenarios, clients frequently engage in training multiple models concurrently, referred to as multi-model federated learning (MMFL). While concurrent training is generally faster than training one model at a time, MMFL exacerbates traditional FL challenges like the presence of non-i.i.d. data: since each individual client may only be able to train one model in each training round due to local resource limitations, the set of clients training each model will change in each round, introducing instability when clients have different data distributions. Existing single-model FL approaches leverage inherent client clustering to accelerate convergence in the presence of such data heterogeneity. However, since each MMFL model may train on a different dataset, extending these ideas to MMFL requires creating a unified cluster or group structure that supports all models while coordinating their training. In this paper, we present the first group-based client-model allocation scheme in MMFL able to accelerate the training process and improve MMFL performance. We also consider a more realistic scenario in which models and clients can dynamically join the system during training. Empirical studies in real-world datasets show that our MMFL algorithms outperform several baselines up to 15 %, particularly in more complex and statically heterogeneous scenarios.

With the rapid development of large-scale low-Earth orbit (LEO) satellite Internet, very LEO (VLEO) Earth observation constellations are increasingly using intersatellite links (LISLs) for cross-layer access to Internet satellites. It is an effective means to enhance data return throughput. When both observation and communication satellite constellations use lasers for networking, cross-layer access between the VLEO and LEO satellite networks requires reallocating lasers. This reallocation disrupts the original topology and impacts network performance. To maximize the collaborative operational efficiency of the double-layer network, we fundamentally analyze the relationship between topology links, throughput, and delay using graph and queuing theory. Innovatively, we discover the superiority of two axioms in addressing the cross-layer topology optimization problem, including the “Minimum Hop Count” and “Minimum Overlap Path.” Based on these axioms, a many-objective cross-layer topology optimization model is established that considers the hop count and the average link utilization frequency of both VLEO and LEO satellite networks. To reduce the reliance on centralized algorithms on global transmission demand, a local distributed interaction mechanism (LDIM) is proposed for cross-layer LISL establishment. An onboard novel distributed many-objective cross-layer topology optimization (NDMTO) algorithm is also introduced for VLEO satellites to manage access strategies. Finally, we use real data from Typhoon LEKIMA to create a multitask scenario and conduct packet-level simulations based on the Starlink and Dove constellations. The results indicate that, compared to existing benchmarks, the NDMTO algorithm improves data throughput by 26.73% and reduces the average transmission delay of emergency task data by 20.7%.

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

Federated learning (FL) enables collaborative learning across multiple clients. In most FL work, all clients train a single learning task. However, the recent proliferation of FL applications may increasingly require multiple FL tasks to be trained simultaneously, sharing clients’ computing resources, which we call Multiple-Model Federated Learning (MMFL). Current MMFL algorithms use naïve average-based client-task allocation schemes that often lead to unfair performance when FL tasks have heterogeneous difficulty levels, as the more difficult tasks may need more client participation to train effectively. Furthermore, in the MMFL setting, we face a further challenge that some clients may prefer training specific tasks to others, and may not even be willing to train other tasks, e.g., due to high computational costs, which may exacerbate unfairness in training outcomes across tasks. We address both challenges by firstly designing FedFairMMFL, a difficulty-aware algorithm that dynamically allocates clients to tasks in each training round, based on the tasks’ current performance levels. We provide guarantees on the resulting task fairness and FedFairMMFL’s convergence rate. We then propose novel auction designs that incentivizes clients to train multiple tasks, so as to fairly distribute clients’ training efforts across the tasks, and extend our convergence guarantees to this setting. We finally evaluate our algorithm with multiple sets of learning tasks on real world datasets, showing that our algorithm improves fairness by improving the final model accuracy and convergence speed of the worst performing tasks, while maintaining the average accuracy across tasks.

Given the crucial role of the earth observation satellites in numerous key applications, using Low Earth Orbit (LEO) satellite internet as intermediaries via Laser Inter-Satellite Links (LISLs) has emerged as a promising solution to help transmit substantial amounts of observation data to ground stations. For Very Low Earth Orbit (VLEO) observation satellites, optimizing the cross-layer topology between themselves and LEO communication satellites has become paramount. To mitigate existing centralized algorithms’ reliance on global data transfer requirement information, a Novel Distributed Interactive Mechanism (NDIM) for cross-layer LISL establishment is proposed. Here, the VLEO observation satellite decides its own access strategy based on local network information gleaned from three information exchanges with the LEO communication satellite. Within this mechanism, the cross-layer link optimization is performed via the formulation of a multi-objective topology optimization model, which considers the transmission requirements of observation satellites, load balancing amongst the communication satellite layer, and the transmission delay of emergency tasks. Based on this framework, we propose a Distributed Multi-objective cross-layer Topology Optimization (DMTO) algorithm. Our algorithm is novel in considering the remaining load of the intermediary communication satellites, and it allows observation satellites to decide on access plans on demand, given incoming data. Additionally, we used real data from Typhoon LEKIMA to establish a multi-task scenario and conducted packet-level simulations based on the Starlink and Dove constellations. The results indicate that, compared to the existing baseline, the DMTO algorithm increased the observation data throughput by 1.37% (326.95 GB) and reduced the average transmission delay of emergency task data by 4.20% (20.5 seconds).

Given the crucial role of the earth observation satellites in numerous key applications, accessing the Low Earth Orbit (LEO) satellite Internet via Laser Inter-Satellite Links (LISLs) has emerged as a promising solution to help transmit substantial amounts of observation data to ground stations. For Very Low Earth Orbit (VLEO) observation satellites, optimizing the inter-layer topology between themselves and LEO communication satellites has become paramount. To mitigate existing centralized algorithms’ reliance on global data transfer requirement information, a Novel Distributed Interactive Mechanism (NDIM) for inter-layer LISLs establishment is proposed. Here, the VLEO observation satellite decides its own access strategy based on local network information gleaned from three information exchanges with the LEO communication satellite. Additionally, considering the transmission requirements of observation satellites and the remaining load of communication satellites, a multi-objective topology optimization model is formulated, leading to the proposal of a Distributed Multi-objective inter-layer Topology Optimization (DMTO) algorithm. Our algorithm is novel in considering the remaining loads of the intermediary communication satellites, and it allows for the transmission of streaming observation data. We perform simulations using real task data (from Typhoon LEKIMA), and our results show that DMTO improves over existing baselines in terms of observation data throughput and communication satellite load balancing.

Federated learning (FL) is a variant of distributed learning in which multiple clients collaborate to learn a global model without sharing their data with the central server. In real-world scenarios, a client may be involved in training multiple unrelated FL models, which we call multi-model federated learning (MMFL), and the client sampling strategy and task allocation are crucial for improving system performance. In this paper, we propose an optimal sampling method to minimize the variance of global updates for unbiased learning in MMFL systems. The resulting method achieves an average accuracy of over 30 % higher than other baseline methods, as we demonstrate through simulations on real-world federated datasets.

On-demand and resource reservation pricing models, widely used in cloud computing, are currently used in Multi-Access Edge Computing (MEC). Nevertheless the edge's resources are distributed and each server has lower capacity. If too much resources were reserved in advance, on-demand users may not get their jobs served on time, jeopardizing MEC's latency benefits. Concurrently, reservation plan users may possess un-used quota. Therefore, we propose a sharing platform where reservation plan users can re-sell unused resource quota to on-demand users. To investigate the mobile network operator's (MNO‘s) incentive of allowing re-selling, we formulate a 3-stage non-cooperative Stackelberg Game and characterize the optimal strategies of buyers and re-sellers. We show that users’ actions give rise to 4 different outcomes at equilibrium, dependent on the prices and supply levels of the sharing and on-demand pools. Based on the 4 possible outcomes, we characterise the MNO's optimal prices for on-demand users. Numerical results show that having both pools gives the MNO an optimal revenue when the on-demand pool's supply is low, and unexpectedly, when the MNO's commission is low. We develop an interactive prototype, and show that users’ decision distributions in studies on our prototype are similar to that of our decision model.

Integrated sensing and communication (ISAC) in terahertz (THz) networks for delay-sensitive services is regarded as a key enabler for future 6G networks, with ISAC helping to aid beam alignment and improve communication reliability in THz networks. Nevertheless, in ISAC systems, there exists resource and performance tradeoffs amongst sensing, communication and computation. For instance, while better sensing accuracy adds information to aid communication beam alignment, it reduces the temporal resources available for communication. Therefore, in this paper we present a joint sensing, communication and computation model, with a non-uniform frame structure allowing the sensing and communication time to be flexibly adjusted. Based on our model, we formulate an average communication outage probability minimization problem, which optimizes the time and carrier frequency allocation schemes, given constrained resources. We then derive in closed-form the available communication outage probability subject to a given delay threshold, decoupling the computation sub-problem from the original problem. Based on this result, we reformulate the original problem into a joint resource allocation problem between sensing and communication, and solve it by applying the Hungarian algorithm. Numerical results show that our strategy outperforms existing baselines in the average communication outage probability performance.

Mobile edge computing (MEC) is a promising paradigm to meet the quality of service (QoS) requirements of latency-sensitive IoT applications. However, attackers may eavesdrop on the offloading decisions to infer the edge server's (ES's) queue information and users' usage patterns, thereby incurring the pattern privacy (PP) issue. Therefore, we propose an offloading strategy which jointly minimizes the latency, ES's energy consumption, and task dropping rate, while preserving PP. Firstly, we formulate the dynamic computation offloading procedure as a Markov decision process (MDP). Next, we develop a Differential Privacy Deep Q-learning based Offloading (DP-DQO) algorithm to solve this-problem while addressing the PP issue by injecting noise into the generated offloading decisions. This is achieved by modifying the deep Q-network (DQN) with a Function-output Gaussian process mechanism. We provide a theoretical privacy guarantee and a utility guarantee (learning error bound) for the DP-DQO algorithm and finally, conduct simulations to evaluate the performance of our proposed algorithm by comparing it with greedy and DQN-based algorithms.

Federated learning (FL) is an increasingly popular form of distributed learning across devices such as sensors and smartphones. To amortize the effort and cost of setting up FL training in real world systems, in practice multiple machine learning tasks may be trained during one FL execution. However, given that the tasks have varying complexities, naïve methods of allocating resource-constrained devices to work on each task may lead to highly variable performance across the tasks. We instead propose an α -fair based allocation algorithm that dynamically allocates tasks to users during multi-model FL training, based on the prevailing loss levels.

Federated learning (FL) is a distributed paradigm for collaboratively learning models without having clients disclose their private data. One natural and practically relevant metric to measure the efficiency of FL algorithms is the total wall-clock training time, which can be quantified by the product of the average time needed for a single iteration and the number of iterations for convergence. In this work, we focus on improving FL efficiency with respect to this metric through caching. Specifically, instead of having all clients download the latest global model from a parameter server, we select a subset of clients to access, with a smaller delay, a somewhat stale global model stored in caches. We propose CacheFL - a cache-enabled variant of FedAvg, and provide theoretical convergence guarantees in the general setting where the local data is imbalanced and heterogeneous. Armed with this result, we determine the caching strategies that minimize total wall-clock training time at a given convergence threshold for both stochastic and deterministic communication/computation delays. Through numerical experiments on real data traces, we show the advantage of our proposed scheme against several baselines, over both synthetic and real-world datasets.

In mobile edge computing (MEC) systems, users offload computationally intensive tasks to edge servers at base stations. However, with unequal demand across the network, there might be excess demand at some locations and underutilized resources at other locations. To address such load-unbalanced problem in MEC systems, in this paper we propose virtual machines (VMs) sharing across base stations. Specifically, we consider the joint VM placement and pricing problem across base stations to match demand and supply and maximize revenue at the network level. To make this problem tractable, we decompose it into master and slave problems. For the placement master problem, we propose a Markov approximation algorithm MAP on the design of a continuous time Markov chain. As for the pricing slave problem, we propose OPA - an optimal VM pricing auction, where all users are truthful. Furthermore, given users' potential untruthful behaviors, we propose an incentive compatible auction iCAT along with a partitioning mechanism PUFF, for which we prove incentive compatibility and revenue guarantees. Finally, we combine MAP and OPA or PUFF to solve the original problem, and analyze the optimality gap. Simulation results show that collaborative base stations increases revenue by up to 50%.

Multi-access Edge Computing (MEC) is an emerging paradigm which allows users to offload their computationally intensive tasks to the network edge. In this paper, we analyze resource allocation in MEC from the market and economic perspective. Due to extremely heterogeneous usage demands across users in the future IoE market, current coarse-grained pricing schemes result in partial wastage: some users would have excess un-utilized resource quota while others might have reserved insufficient resources. Therefore, we introduce a novel sharing economy-inspired business model, where a platform facilitates the sharing of resource quota among users, increasing resource efficiency. The goal is to maximize the overall welfare of users who join the sharing platform. As the platform lacks control and has imperfect knowledge of users' payoff functions and distributions, a distributed pricing mechanism is proposed. In our mechanism, the platform and users jointly arrive at an equilibrium. We prove that the equilibrium point of the mechanism is the socially optimal point. Simulations illustrate convergence, the robustness of our mechanism to changes in demand and supply, and that sharing increases the welfare of users.

This paper considers a scenario in which an access point (AP) is equipped with a mobile edge server (MEC) of finite computing power, and serves multiple resource-hungry mobile users by charging users a price. This price helps to regulate users' behavior in offloading computation jobs to the AP. To that end, first we introduce an economics model for MEC bearing physical layer offloading intuition. We then propose a learning based pricing mechanism, in which with no direct control and no knowledge of users' private information, the AP learns the optimal price. Under our mechanism, the AP induces self-interested users to make socially optimal offloading decisions, thus maximizing the system-wide welfare.