Source-linked AI summary
Wireless Edge Computing with Latency and Reliability Guarantees
Mohammed S. Elbamby, Cristina Perfecto, Chen-Feng Liu, Jihong Park, Sumudu Samarakoon, Xianfu Chen, Mehdi Bennis
TL;DR
The paper examines how edge computing can provide latency and reliability guarantees for demanding wireless applications by moving computation, storage, and control closer to users. It surveys URLLC enablers and edge use cases, concluding that future wireless networks require multiple complementary mechanisms tailored to services and deployments.
Problem
The problem is enabling wireless edge services to meet stringent latency and reliability requirements for applications such as VR, V2X, and edge AI.
Method
The paper surveys edge services, URLLC enablers, and case studies involving resource optimization, extreme-event reliability, and latency-aware task or content delivery.
Results
The article concludes that realizing ultra-reliable, low-latency edge computing requires multiple high-reliability and low-latency enablers applied to different services and use cases.
Takeaways & Limitations
The supported takeaway is that edge computing is an essential component of future wireless networks, but its vision still requires several challenges to be overcome.
Takeaways & Limitations
The article notes a tradeoff between the data availability, convergence speed, and estimation accuracy involved in generalized extreme-value modeling.
Abstract
from arXiv · showhide
Edge computing is an emerging concept based on distributing computing, storage, and control services closer to end network nodes. Edge computing lies at the heart of the fifth generation (5G) wireless systems and beyond. While current state-of-the-art networks communicate, compute, and process data in a centralized manner (at the cloud), for latency and compute-centric applications, both radio access and computational resources must be brought closer to the edge, harnessing the availability of computing and storage-enabled small cell base stations in proximity to the end devices. Furthermore, the network infrastructure must enable a distributed edge decision-making service that learns to adapt to the network dynamics with minimal latency and optimize network deployment and operation accordingly. This article will provide a fresh look to the concept of edge computing by first discussing the applications that the network edge must provide, with a special emphasis on the ensuing challenges in enabling ultra-reliable and low-latency edge computing services for mission-critical applications such as virtual reality (VR), vehicle-to-everything (V2X), edge artificial intelligence (AI), and so forth. Furthermore, several case studies where the edge is key are explored followed by insights and prospect for future work.
1 INTRODUCTION
The introduction frames edge computing as a response to services requiring ultra-high success rates and minimal latency, while emphasizing the tension between ultra-reliability and low latency. It then positions the article around edge services, URLLC enablers, use cases, and future work.
- 1 INTRODUCTION: Growing services such as video streaming, AR/VR, and mission-critical applications require data, computation, and storage with ultra-high success rates and minimal latency.MEC brings processing and storage toward the network edge to reduce latency between network nodes and remote servers.
- 1 INTRODUCTION: Edge base stations provide nearby connectivity, content, computation, and control for end devices and coupled network nodes.Their functions include executing user tasks, supplying customized services, and managing interactions between connected nodes.
- 1 INTRODUCTION: Edge-computing performance depends on communication between edge servers and devices plus processing at the edge server.Relevant optimizations include bandwidth and power allocation, server selection, task distribution and splitting, offloading, computation-cycle allocation, queuing, and prioritization.
- 1 INTRODUCTION: 5G shifts toward service-specific ultra-reliability and low-latency guarantees, whose requirements are often contradictory and require distinct tools.The shift is driven by critical and latency-sensitive communication services.
- 1 INTRODUCTION: The article examines feasible edge services with latency and reliability guarantees, then studies URLLC, selected use cases, and future research.Its structure moves from offered services to URLLC interactions, use cases, concluding remarks, and future works.
2 EDGE COMPUTING SERVICES
This section presents edge computing as a distributed alternative to centralized cloud architectures for low-latency, personalized services. It covers edge content, processing, and distributed control, including their relevance to latency-sensitive applications and mission-critical operation.
- 2 EDGE COMPUTING SERVICES: Centralized cloud architectures increase service latency because servers are distant from users, motivating distributed resources and services at the network edge.The shift also addresses context-aware service and user-data privacy needs, although diverse implementations have lacked a specific interoperability standard.
- 2 EDGE COMPUTING SERVICES: Edge caching predicts popular content, prefetches it from the core network, and stores it near users to reduce delivery time and backhaul load.This approach requires content-popularity prediction and sufficient edge storage capacity.
- 2 EDGE COMPUTING SERVICES: Edge processing supports resource-greedy, latency-sensitive applications when device size, portability, battery life, or incomplete task-data access limits local execution.The passage identifies smart factories, self-driving vehicles, and virtual and augmented reality as examples.
- 2 EDGE COMPUTING SERVICES: Distributed decision making among edge servers can reduce decision latency and preserve privacy compared with centralized collection of local network states.The article also links edge control with multiagent reinforcement learning that accounts for latency and reliability under dynamic, noisy conditions.
- 2 EDGE COMPUTING SERVICES: The figure organizes key URLLC enablers for edge computing within an Industry 4.0 smart-factory ecosystem containing cyberphysical systems, IoT, and MEC.It provides the section’s ecosystem-level framing for connecting edge services with reliability and latency mechanisms.
3 URLLC ENABLERS AND CHALLENGES
Mission-critical applications require wireless edge networks to provide guaranteed high reliability and low latency. The article surveys communication, computing, learning, and statistical enablers, together with their associated challenges and tradeoffs.
- 3.1 URLLC overview: URLLC supports communication between mission-critical devices and edge servers, targeting guaranteed high reliability and low latency.The section identifies URLLC as a pivotal 5G service and summarizes related challenges and enablers.
- 3.2.1 Low latency Enablers: High-capacity mmWave links and proximity-based computing reduce offloading and over-the-air latency, while requiring beam alignment and dense edge deployments.mmWave propagation is vulnerable to blockage and requires directional beam management; proximity reduces the distance contributing to end-to-end latency.
- 3.2.2 High Reliability Enablers: Task replication improves reliability under channel dynamics but adds load, motivating threshold-based replication that limits extra activation when the first server is delayed.The cited approach offloads to an additional server only after the first server exceeds a delay threshold.
- 3.2.2 High Reliability Enablers: Federated learning improves inference reliability by aggregating edge models without exchanging private training data, but requires joint communication, computation, and learning design.Model compression and quantization trade communication efficiency against accuracy, while resource allocation must address unseen data, channel variation, and changing resources.
- 3.2.2 High Reliability Enablers: Extreme value theory targets rare latency violations because average-based designs are inadequate for reliability levels ranging from 10^-3 to 10^-9.Estimating extreme-value parameters trades data availability against approximation, convergence speed, and accuracy, with machine learning and federated aggregation proposed as remedies.
4 APPLICATIONS AND USE CASES
The article examines edge-computing applications that improve latency and reliability, including XR, federated learning, vehicular communication, adaptive offloading, and proactive VR streaming.
- Extended Reality: XR devices’ limited computing capabilities constrain standalone content quality, while VR typically tolerates only 15–20 milliseconds of motion-to-photon delay.Edge computing is therefore relevant to both XR computation demands and delay sensitivity.
- Extreme Event-Controlled MEC: EVT-controlled MEC improves extreme-event queue metrics over schemes without edge computing or local computation capability, with Pr(Q > d) = 3.4×10^-3 at d = 3.96 × 10^4.The approach uses queue-length tails to control rare latency events.
- Vehicular Edge Computing and V2X/V2V: extFL combines EVT and federated learning so vehicles estimate queue-length tails locally while optimizing transmission decisions for reliable low-latency V2V communication.The objective is to minimize worst-case queue lengths while preserving queuing-latency reliability.
- Vehicular Edge Computing and V2X/V2V: As networks become denser, extFL achieves equivalent or better reliability than centralized CEN while reducing data exchange and lowering the mean and variance of worst-case queue lengths.The reported queue-length improvement is attributed to reduced training latency relative to CEN.
- Proactive VR Streaming: DRNN-based FoV prediction enables proactive mmWave 360° VR streaming by informing user clustering, scheduling, and beamforming decisions.Predicted inter-user FoV correlations support proactive transmission and content adaptation.
- Proactive VR Streaming: Multicasting raises HD delivery rate by 40–50% and reduces latency by 33–70%, while proactive schemes reduce average delay without sacrificing HD quality.MPROAC+ additionally keeps worst-case delays bounded through an explicit latency constraint.
- Proactive VR Streaming: Increasing the Lyapunov parameter prioritizes HD delivery rate, whereas lower values prioritize keeping delay bounded with high probability at the expense of delivery rate.The Jaccard index also varies with traffic load because proactive transmission trades excess content against missed frames.
- Proactive Edge Computing: Combining proactive computing and multicast significantly reduces computing and communication latency relative to baselines, although prediction errors slightly increase communication latency through retransmissions.The comparison separates proactive-computing gains in processing from multicast gains in communication.
5 CONCLUSIONS AND FUTURE OUTLOOK
The article frames edge computing as essential to future wireless networks and calls for coordinated reliability and low-latency enablers, distributed AI, and resource optimization across diverse use cases.
- Edge computing is essential to future wireless networks, but realizing ultra-reliable and low-latency services requires overcoming several challenges.
- Resource optimization must address both communication and computing delays as the number of players varies in an edge deployment.
- Future edge systems should develop alongside URLLC and distributed AI to provide adaptive computing, content, and control services in dynamic environments.