BEIJING, June 20, 2024 /PRNewswire/ — Imagine a world where augmented reality systems project a digital avatar onto your vehicle’s dashboard, guiding you in real time. This integration of vehicular navigation within the Metaverse combines the digital and physical worlds in ways that were previously not possible, creating opportunities for richer, more interactive experiences for drivers and passengers.
To ensure seamless communication between the servers and the vehicles, researchers from Nanyang Technological University, Singapore University of Technology and Design, Guangdong University of Technology, Army Engineering University of PLA, and the National Natural Science Foundation of China have proposed a task migration system that intelligently determines the optimal time to move tasks between the vehicle and external servers. Their paper was published in the 2024 Issue 2 of the IEEE/CAA Journal of Automatica Sinica.
“The mobility of vehicles poses a significant challenge for unmanned aerial vehicle-assisted vehicular Metaverses to ensure the continuity of avatar services, especially when the vehicles leave coverage of their host edge servers. We propose a framework to address this issue. By using advanced computer algorithms, we can quickly and reliably determine the best server to handle each task, much like how a smart assistant would choose the best route based on current traffic conditions,” says Zehui Xiong, the corresponding author of the study and an Assistant Professor at Singapore University of Technology and Design.
The researchers integrated transformers into a multi-agent proximal policy optimization (MAPPO) algorithm. In this process, the digital avatar or the agent migrates a task, based on its observations, such as its location, speed, traffic conditions, and available servers. To further optimize the decision-making, the transformer converts the actions and observations of multiple agents into sequences.
“This approach allows each vehicle to dynamically decide whether to perform an avatar task pre-migration, thereby reducing the average latency of all vehicles and improving the quality of avatar services,” says Dr. Xiong. To ensure the security of communications, the transactions between the vehicle and the external server are recorded using Smart Contracts deployed on blockchains.
The researchers found that the method outperforms traditional reinforcement learning approaches by approximately 2% and reduces avatar task execution latency by around 20%. Thus, the proposed method paves the way for vehicular services in the Metaverse, with broad potential applications.
“This research could lead to widespread adoption of highly interactive and immersive vehicular services, improve road safety through better navigation aids and real-time updates, and pave the way for more sustainable and scalable smart city infrastructures,” says Dr. Xiong.
You can hear directly from the researchers in this podcast.
Reference
Authors: Jiawen Kang1, Junlong Chen1, Minrui Xu2, Zehui Xiong3, Yutao Jiao4, Luchao Han5, Dusit Niyato2, Yongju Tong1, and Shengli Xie1
Title of original paper: UAV-Assisted Dynamic Avatar Task Migration for Vehicular Metaverse Services: A Multi-Agent Deep Reinforcement Learning Approach
Journal: IEEE/CAA Journal of Automatica Sinica
DOI: https://doi.org/10.1109/JAS.2023.123993
Contact:
Yan Ou
+86 10 82544459
379386@email4pr.com
SOURCE IEEE/CAA Journal of Automatica Sinica