Unleashing the Power of Edge-Cloud Generative AI in Mobile Networks: A Survey of AIGC Services

IEEE Communications Surveys & Tutorials, Journal Year: 2024, Volume and Issue: 26(2), P. 1127 - 1170

Published: Jan. 1, 2024

Artificial Intelligence-Generated Content (AIGC) is an automated method for generating, manipulating, and modifying valuable diverse data using AI algorithms creatively. This survey paper focuses on the deployment of AIGC applications, e.g., ChatGPT Dall-E, at mobile edge networks, namely that provide personalized customized services in real time while maintaining user privacy. We begin by introducing background fundamentals generative models lifecycle which includes collection, training, fine-tuning, inference, product management. then discuss collaborative cloud-edge-mobile infrastructure technologies required to support enable users access networks. Furthermore, we explore AIGC-driven creative applications use cases Additionally, implementation, security, privacy challenges deploying Finally, highlight some future research directions open issues full realization

Language: Английский

---

Robust Information Bottleneck for Task-Oriented Communication With Digital Modulation

IEEE Journal on Selected Areas in Communications, Journal Year: 2023, Volume and Issue: 41(8), P. 2577 - 2591

Published: June 23, 2023

Task-oriented communications, mostly using learning-based joint source-channel coding (JSCC), aim to design a communication-efficient edge inference system by transmitting task-relevant information the receiver. However, only without introducing any redundancy may cause robustness issues in learning due channel variations, and JSCC which directly maps source data into continuous input symbols poses compatibility on existing digital communication systems. In this paper, we address these two first investigating inherent tradeoff between informativeness of encoded representations distortion received representations, then propose task-oriented scheme with modulation, named discrete (DT-JSCC), where transmitter encodes features representation transmits it receiver modulation scheme. DT-JSCC scheme, develop robust encoding framework, bottleneck (RIB), improve derive tractable variational upper bound RIB objective function approximation overcome computational intractability mutual information. The experimental results demonstrate that proposed achieves better performance than baseline methods low latency, exhibits variations applied framework.

Language: Английский

---

Integrated Sensing-Communication-Computation for Edge Artificial Intelligence

IEEE Internet of Things Magazine, Journal Year: 2024, Volume and Issue: 7(4), P. 14 - 20

Published: June 27, 2024

Language: Английский

---

When Large Language Model Agents Meet 6G Networks: Perception, Grounding, and Alignment

IEEE Wireless Communications, Journal Year: 2024, Volume and Issue: 31(6), P. 63 - 71

Published: Aug. 26, 2024

Language: Английский

---

Contrastive Learning based Semantic Communications

IEEE Transactions on Communications, Journal Year: 2024, Volume and Issue: 72(10), P. 6328 - 6343

Published: May 14, 2024

Recently, there has been a growing interest in learning-based semantic communication because it can prioritize the preservation of meaningful information over accuracy transmitted symbols, resulting improved efficiency. However, existing approaches still face limitations defining level loss and often struggle to find good trade-off between preserving intricate details. In addition, cannot effectively train encoders decoders without support downstream models. To address these limitations, this paper proposes contrastive learning (CL)-based system. First, inspired by practical observations, we introduce concept propose coding (SemCC) approach that treats data corruption during transmission as form augmentation within CL framework. Moreover, re-encoding (SemRE) operation, which uses duplicate encoder deployed at receiver guide entire training process when model is inaccessible. Further, design procedure for SemCC SemRE approaches, respectively, balance Finally, simulations are performed demonstrate superiority proposed competing approaches. particular, our achieve significant improvement up 53% on CIFAR-10 dataset with bandwidth compression ratio 1/24, also obtain comparable image reconstruction quality improved.

Language: Английский

---

Over-the-Air Computation: Foundations, Technologies, and Applications

arXiv (Cornell University), Journal Year: 2022, Volume and Issue: unknown

Published: Jan. 1, 2022

The rapid advancement of artificial intelligence technologies has given rise to diversified intelligent services, which place unprecedented demands on massive connectivity and gigantic data aggregation. However, the scarce radio resources stringent latency requirement make it challenging meet these demands. To tackle challenges, over-the-air computation (AirComp) emerges as a potential technology. Specifically, AirComp seamlessly integrates communication procedures through superposition property multiple-access channels, yields revolutionary paradigm shift from "compute-after-communicate" "compute-when-communicate". By this means, enables spectral-efficient low-latency wireless aggregation by allowing multiple devices occupy same channel for transmission. In paper, we aim present recent in terms foundations, technologies, applications. mathematical form design are introduced foundations AirComp, critical issues over different network architectures then discussed along with review existing literature. employed analysis optimization reviewed information theory signal processing perspectives. Moreover, studies that practical implementation systems, elaborate applications Internet Things edge networks. Finally, research directions highlighted motivate future development AirComp.

Language: Английский

---

Toward Scalable Wireless Federated Learning: Challenges and Solutions

IEEE Internet of Things Magazine, Journal Year: 2023, Volume and Issue: 6(4), P. 10 - 16

Published: Dec. 1, 2023

The explosive growth of smart devices (e.g., mobile phones, vehicles, drones) with sensing, communication, and computation capabilities gives rise to an unprecedented amount data. generated massive data together the rapid advancement machine learning (ML) techniques spark a variety intelligent applications. To distill intelligence for supporting these applications, federated (FL) emerges as effective distributed ML framework, given its potential enable privacy-preserving model training at network edge. In this article, we discuss challenges solutions achieving scalable wireless FL from perspectives both design resource orches-tration. For design, how task-oriented aggregation affects performance FL, followed by proposing enhance communication scalability via reducing distortion improving device participation. orchestration, identify limitations existing optimization-based algorithms propose three algorithmic computation-efficient allocation FL. We highlight several research issues that deserve further study.

Language: Английский

---

Federated Reinforcement Learning for Electric Vehicles Charging Control on Distribution Networks

IEEE Internet of Things Journal, Journal Year: 2023, Volume and Issue: 11(3), P. 5511 - 5525

Published: Aug. 21, 2023

With the growing popularity of electric vehicles (EVs), maintaining power grid stability has become a significant challenge. To address this issue, EV charging control strategies have been developed to manage switch between vehicle-to-grid (V2G) and grid-to-vehicle (G2V) modes for EVs. In context, multiagent deep reinforcement learning (MADRL) proven its effectiveness in control. However, existing MADRL-based approaches fail consider natural flow charging/discharging distribution network ignore driver privacy. deal with these problems, article proposes novel approach that combines multi-EV radial (RDN) operating under optimal (OPF) distribute real time. A mathematical model is describe RDN load. The problem formulated as Markov decision process (MDP) find an strategy balances V2G profits, load, anxiety. effectively learn strategy, federated algorithm named FedSAC further proposed. Comprehensive simulation results demonstrate superiority our proposed terms diversity fluctuations on RDN, convergence efficiency, generalization ability.

Language: Английский

---

Integrated Sensing-Communication-Computation for Over-the-Air Edge AI Inference

IEEE Transactions on Wireless Communications, Journal Year: 2023, Volume and Issue: 23(4), P. 3205 - 3220

Published: Aug. 24, 2023

Edge-device co-inference refers to deploying well-trained artificial intelligent (AI) models at the network edge under cooperation of devices and servers for providing ambient services. For enhancing utilization limited resources in edge-device tasks from a systematic view, we propose task-oriented scheme integrated sensing, computation communication (ISCC) this work. In system, all sense target same wide view obtain homogeneous noise-corrupted sensory data, which local feature vectors are extracted. All aggregated server using over-the-air (AirComp) broadband channel with orthogonal-frequency-division-multiplexing technique suppressing sensing noise. The denoised global vector is further input server-side AI model completing downstream inference task. A novel design criterion, called maximum minimum pair-wise discriminant gain, adopted classification tasks. It extends distance closest class pair space, leading balanced enhanced accuracy. Under problem joint power assignment, transmit precoding receive beamforming formulated. challenge lies three aspects: coupling between AirComp, optimization dimensions' AirComp aggregation over channel, complicated form gain. To solve problem, ISCC proposed. Experiments based on human motion recognition task conducted verify advantages proposed existing baseline.

Language: Английский

---

Over-the-Air Federated Learning and Optimization

IEEE Internet of Things Journal, Journal Year: 2024, Volume and Issue: 11(10), P. 16996 - 17020

Published: Jan. 10, 2024

Federated learning (FL), as an emerging distributed machine paradigm, allows a mass of edge devices to collaboratively train global model while preserving privacy. In this tutorial, we focus on FL via over-the-air computation (AirComp), which is proposed reduce the communication overhead for over wireless networks at cost compromising in performance due aggregation error arising from channel fading and noise. We first provide comprehensive study convergence AirComp-based FEDAVG (AIRFEDAVG) algorithms under both strongly convex non-convex settings with constant diminishing rates presence data heterogeneity. Through asymptotic analysis, characterize impact bound insights system design guarantees. Then derive AIRFEDAVG objectives. For different types local updates that can be transmitted by (i.e., model, gradient, difference), reveal transmitting may cause divergence training procedure. addition, consider more practical signal processing schemes improve efficiency further extend analysis forms caused these schemes. Extensive simulation results objective functions, information, verify theoretical conclusions.

Language: Английский

---