Source-linked AI summary
Realizing the Metaverse with Edge Intelligence: A Match Made in Heaven
Wei Yang Bryan Lim, Zehui Xiong, Dusit Niyato, Xianbin Cao, Chunyan Miao, Sumei Sun, Qiang Yang
TL;DR
The Metaverse remains difficult to realize as a seamless, interoperable, and shardless environment under stringent sensing, communication, computation, and privacy requirements. The paper examines its architecture, develops the convergence of edge intelligence with Metaverse infrastructure, and applies that perspective to collaborative virtual city development. It concludes by framing edge intelligence and the Metaverse as a promising confluence while identifying open research issues.
Problem
Existing Metaverse platforms are limited by fragmented virtual worlds, stringent sensing, communication, and computation requirements, and growing privacy concerns.
Method
The paper defines the Metaverse architecture, analyzes edge intelligence-driven infrastructure, and presents a collaborative edge-driven virtual city case study.
Results
The paper presents an architecture and edge-intelligence framework integrating cloud-edge-end computation, coded distributed computation, privacy-preserving learning, and virtual city development.
Takeaways & Limitations
The work serves as an initial attempt to motivate the confluence of edge intelligence and the Metaverse while outlining future research directions.
Abstract
from arXiv · showhide
Dubbed "the successor to the mobile Internet", the concept of the Metaverse has recently exploded in popularity. While there exists lite versions of the Metaverse today, we are still far from realizing the vision of a seamless, shardless, and interoperable Metaverse given the stringent sensing, communication, and computation requirements. Moreover, the birth of the Metaverse comes amid growing privacy concerns among users. In this article, we begin by providing a preliminary definition of the Metaverse. We discuss the architecture of the Metaverse and mainly focus on motivating the convergence of edge intelligence and the infrastructure layer of the Metaverse. We present major edge-based technological developments and their integration to support the Metaverse engine. Then, we present our research attempts through a case study of virtual city development in the Metaverse. Finally, we discuss the open research issues.
I. INTRODUCTION
The Metaverse aims to extend the Internet into a seamless, interoperable, and shardless virtual environment, but current platforms and stringent requirements leave that vision unrealized. The article addresses these challenges by examining architecture, edge intelligence, applications, and open issues.
- The Metaverse is envisioned as a seamless integration of interoperable virtual worlds rather than separate platforms such as Fortnite and Roblox.
- Current systems remain limited because open-world VRMMO applications and persistent, shardless environments are still nascent and demanding.
- The article focuses on edge intelligence-driven infrastructure, defining edge intelligence as the convergence of edge computing and AI.
- It distinguishes Edge for AI, which moves sensing, communication, training, and inference toward data sources, from AI for Edge, which improves orchestration.
- The paper contributes a Metaverse architecture, discusses applications, presents collaborative edge-driven virtual city development, and identifies interdisciplinary research opportunities.
II. THE METAVERSE: ARCHITECTURE, TECHNOLOGIES, AND APPLICATIONS
The Metaverse architecture connects physical-world stakeholders, user interfaces, sensing, service providers, and infrastructure through an engine that generates and maintains virtual worlds. Blockchain, immersive technologies, digital twins, and AI support interoperability, realism, interaction, and services.
- The Metaverse comprises interoperable, immersive, and shardless virtual ecosystems navigable by user-controlled avatars.
- Physical-virtual world interaction: Physical-world stakeholders influence virtual-world components, while consequences in the virtual world feed back into the physical world.
- Physical-virtual world interaction: Users access virtual worlds through HMDs or AR goggles, while IoT and sensor networks provide environmental data for digital-twin updates.
- Physical-virtual world interaction: Virtual service providers maintain worlds and user-generated content, while physical service providers operate edge communication and computation infrastructure.
- Metaverse engine: The Metaverse engine uses stakeholder-provided inputs to generate, maintain, and enhance the virtual world.
- Technologies and applications: VR/AR and haptics provide visual and tactile interaction, digital twins model physical environments in real time, and AI supports rendering, chatbots, and generated characters.
- Technologies and applications: Blockchain supports virtual-goods ownership and decentralized trading, while crosschain technology enables secure interoperability across separately managed data.
- The supporting infrastructure is intended to make the Metaverse scalable, shardless, ubiquitously accessible, and trustworthy.
B. Edge intelligence-empowered infrastructure
The infrastructure layer uses edge intelligence to meet the Metaverse’s communication, computation, scalability, and privacy requirements. Cloud-edge-end execution, coded computation, semantic communication, and federated learning distribute workloads while supporting responsiveness and privacy.
- Communication and Networking: VR requires 250 Mbit/s and packet error rates of 10^-1 ∼10^-3, while haptic traffic requires 1 Mbit/s and 10^-4 ∼10^-5.
- Computation and Storage: The cloud-edge-end paradigm assigns lightweight tasks to devices, latency-sensitive foreground rendering to edge servers, and delay-tolerant background rendering to the cloud.
- Computation and Storage: Edge caching improves retrieval efficiency and reduces computation overhead by storing popular content near users.
- Edge intelligence supports Edge for AI and AI for Edge by combining AI-enabled Metaverse functions with resource-efficient collaborative edge orchestration.
- Edge for AI: Coded redundancy and worker selection mitigate stragglers in distributed edge computation, while polynomial codes provide a recovery threshold that does not scale with worker count.
- Edge for AI: Federated learning trains models locally and shares parameters or gradients instead of raw data, supporting collaborative learning while reducing backbone communication.
- AI for Edge: Semantic communication is presented as a response to the bandwidth demand generated by Metaverse data traffic.
- A one-size-fits-all reward can discourage resource-rich stakeholders from participating, motivating incentive mechanisms for service requesters.
C. Applications
The paper identifies emerging applications and services in the Metaverse, introducing this application area before presenting specific examples in the broader discussion.
- The paper identifies important emerging applications and services in the Metaverse.
1) Entertainment and social activities:
The Metaverse can make social interaction more immersive and support low-cost product testing before physical release. It can also model virtual product experiences, including realistic test-drive environments.
- VR and haptic technology can make social interactions more immersive than current audio-video platforms limited to rigid 2D grids.
- The Metaverse can pilot-test products before physical release at lower cost and with fewer safety considerations.
- Virtual twins can place physical products in users’ inventories for marketing, including test-drive environments modeled after highways with realistic traffic conditions.
3) Virtual education:
The Metaverse can address limitations of virtual education through personalized AI tutors and haptic hands-on lessons. It can also support creator compensation through NFT-based trading with programmed revenue sharing.
- Usage data in the Metaverse can refine AI tutors for personalized lessons.
- Haptic technology can deliver hands-on lessons involving machines or tools more effectively.
- NFTs can identify the originality of user-generated products traded in the Metaverse, supporting creator compensation through programmed sales proceeds.
- The Metaverse can provide a platform for gig workers to create and actively trade user-generated content.
III. CASE STUDY: A FRAMEWORK FOR COLLABORATIVE EDGE-DRIVEN VIRTUAL CITY DEVELOPMENT IN THE METAVERSE
The case study develops a collaborative edge-driven framework for virtual cities by combining edge resources, collaborative sensing, edge rendering, and demand-aware resource reservation. Its studies address synchronization, service allocation, and uncertain demand.
- The virtual-city framework leverages network-edge sensing, computation, communication, and storage to support desirable Metaverse qualities.
- Collaborative sensing uses IoT and wireless sensor networks to feed digital twins fresh data streams for real-time physical-virtual synchronization.
- The sensing study models non-cooperative service-provider adaptation with evolutionary games and examines how rewards affect servicing populations and synchronization frequency.
- The edge-rendering study compares DRL-based DDA with vanilla DDA and an Ornstein-Uhlenbeck auction method across various bitrates.
- The resource-reservation study compares SIP, expected-value formulation, and a historical-average random scheme for reducing on-demand cost.
B. Edge-assisted efficient rendering of the immersive virtual world
Edge-assisted rendering addresses device battery limits and the need to deliver immersive VR efficiently. The section connects edge-service markets, quality-aware valuation, and resource allocation under uncertain demand.
- Non-panoramic VR rendering reduces data traffic and computation by rendering only images covering each eye’s viewport.
- A Double Dutch Auction mechanism supports edge server-user association and pricing for non-panoramic VR rendering services.
- User valuation combines VMAF-based streaming quality and SSIM-based VR image quality, influenced by head rotation speeds and expected streaming bitrates.
- Virtual service providers use physical and virtual resources from separate entities, including logistic and edge-computing services.
- Reservation decisions must be made before demand is known, creating over-provisioning or under-provisioning risks.
A. Redefining user QoE
The Metaverse requires user QoE concepts beyond conventional Internet metrics, while interoperability, security, and privacy remain central infrastructure concerns.
- A. Redefining user QoE: Metaverse QoE must be redefined by relating network requirements to users’ visual perceptions.Images shown for less than 13 ms are imperceptible to the human eye, constraining network timing requirements.
- A. Redefining user QoE: B5G infrastructure should use semantic-aware, goal-oriented metrics such as Value of Information, which combines packet content and age.This represents a shift beyond conventional data transmission rate metrics.
- A. Redefining user QoE: Interoperability standards and unified communication protocols are needed to support a seamless Metaverse across diverse virtual worlds and communication systems.Standardization is also presented as important for encouraging user-generated content.
- A. Redefining user QoE: Blockchain-based transactions increase security exposure, including malicious smart contracts and hardware-derived privacy leaks such as password inference from finger tracking.These risks arise from both the economic infrastructure and new Metaverse access hardware.
- A. Redefining user QoE: Eye tracking and other newly collected user data enable personalized advertising but create novel privacy challenges.The data may be delivered directly into users’ fields of view.
E. Economics of the edge-driven Metaverse
The edge-driven Metaverse extends resource contention into virtual environments, requiring efficient optimization under newly defined QoE considerations.
- E. Economics of the edge-driven Metaverse: Metaverse users and service providers must optimize resource usage across physical and virtual worlds while accounting for newly defined QoE.The expansion of service and resource trading creates contention beyond the physical world.
- E. Economics of the edge-driven Metaverse: The article presents a smart-city Metaverse case study and future research directions as an initial attempt to connect edge intelligence with the Metaverse.The conclusion frames this work as a starting point for the convergence of the two areas.
BIOGRAPHIES
The biographies identify researchers spanning wireless networking, AI, blockchain, edge intelligence, intelligent transportation, and active living.
- BIOGRAPHIES: Wei Yang Bryan Lim researches edge intelligence and resource allocation at Nanyang Technological University.He is pursuing a Ph.D. through the Alibaba Talent Programme.
- BIOGRAPHIES: Zehui Xiong is an Assistant Professor at Singapore University of Technology and Design whose interests include wireless communications, blockchain, and edge intelligence.His interests also cover network games and economics.
- BIOGRAPHIES: Sumei Sun is a Principal Scientist and Acting Executive Director at Singapore’s Institute for Infocomm Research.She heads its Communications and Networks Department.
- BIOGRAPHIES: Dusit Niyato is a Professor at Nanyang Technological University specializing in wireless and mobile networking.The biography notes more than 380 technical papers and four U.S. and German patents.
- BIOGRAPHIES: Xianbin Cao is a Beihang University professor whose research interests include intelligent transportation systems, airspace transportation management, and intelligent computation.
- BIOGRAPHIES: Chunyan Miao is a professor at Nanyang Technological University and directs the LILY research centre on active living for the elderly.
- BIOGRAPHIES: Qiang Yang leads AI at WeBank and is Chair Professor in computer science and engineering at HKUST.