TY - JOUR
T1 - Decentralized Cooperative Caching for Sustainable Metaverse Via Meta Self-Supervision Learning
AU - Chen, Ximing
AU - Qiu, Jing
AU - Zhou, Fasheng
AU - Lu, Hui
AU - Zhang, Lejun
AU - Tian, Zhihong
AU - Du, Xiaojiang
AU - Guizani, Mohsen
N1 - Publisher Copyright:
© 2002-2012 IEEE.
PY - 2023/10/1
Y1 - 2023/10/1
N2 - Metaverse is a multidimensional virtual world that integrates digital twin space technology. It is an interactive and integrative process with the physical world, involving virtualization and digitization of the real world. The high demands of the metaverse on transmission latency, processing speed, and security pose significant challenges to network load. In this article, we study an online distributed caching model in the context of digital twin metaverse. With unknown file popularity, we elaborate a collaborative caching system to optimize the burst bit error rate of wireless transmission while satisfying the latency requirements in metaverse applications. Then, we propose a meta-based self-supervised learning algorithm that can jointly allocate caching resources among various small base stations (i.e., metaverse edge nodes) in a distributed manner. Notably, given the system heterogeneity due to the massive number of devices in the metaverse, we cleverly introduce a multi-exits mechanism for neural networks to cope with the strong demand for ultra-low latency in metaverse applications. The effectiveness of the proposed methods is verified by numerical simulations, which significantly improve the transmission efficiency compared with existing comparative methods, thus facilitating the development of green communication technologies. Therefore, these methods are of great relevance to the future development and application of metaverse.
AB - Metaverse is a multidimensional virtual world that integrates digital twin space technology. It is an interactive and integrative process with the physical world, involving virtualization and digitization of the real world. The high demands of the metaverse on transmission latency, processing speed, and security pose significant challenges to network load. In this article, we study an online distributed caching model in the context of digital twin metaverse. With unknown file popularity, we elaborate a collaborative caching system to optimize the burst bit error rate of wireless transmission while satisfying the latency requirements in metaverse applications. Then, we propose a meta-based self-supervised learning algorithm that can jointly allocate caching resources among various small base stations (i.e., metaverse edge nodes) in a distributed manner. Notably, given the system heterogeneity due to the massive number of devices in the metaverse, we cleverly introduce a multi-exits mechanism for neural networks to cope with the strong demand for ultra-low latency in metaverse applications. The effectiveness of the proposed methods is verified by numerical simulations, which significantly improve the transmission efficiency compared with existing comparative methods, thus facilitating the development of green communication technologies. Therefore, these methods are of great relevance to the future development and application of metaverse.
UR - http://www.scopus.com/inward/record.url?scp=85178006001&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85178006001&partnerID=8YFLogxK
U2 - 10.1109/MWC.008.2300006
DO - 10.1109/MWC.008.2300006
M3 - Article
AN - SCOPUS:85178006001
SN - 1536-1284
VL - 30
SP - 96
EP - 103
JO - IEEE Wireless Communications
JF - IEEE Wireless Communications
IS - 5
ER -