Abstract:
Unmanned Aerial Vehicles (UAVs) are emerging as a potential sensor for capturing images and data for constructing the Metaverse due to their unparalleled mobility and spe...Show MoreMetadata
Abstract:
Unmanned Aerial Vehicles (UAVs) are emerging as a potential sensor for capturing images and data for constructing the Metaverse due to their unparalleled mobility and special field of view (FoV). However, the lack of an accurate image capturing model and the challenge of selecting optimal flight strategies significantly degrade the effectiveness of UAV deployment. This letter introduces a novel UAV-assisted Metaverse construction framework that enables a UAV to capture and transmit images with minimal delay. In particular, the image resolution is characterized by the UAV’s three-dimension image-taking location, which enables the UAV to efficiently complete task by optimizing and searching for coordinates that meet the requirements for image capturing and transmission. The optimization problem is very challenging under uncertainty of the communication channel and dynamic of the UAV’s environment. To solve this problem, we leverage a Markov Decision Process framework to model the dynamic of the UAV, and we develop a deep reinforcement learning algorithm to find the optimal policy. Furthermore, we employ a state normalization technique to overcome the DRL non-convergence issue caused by the continuous action and state spaces of the UAV. Simulation results show that the proposed scheme outperforms baseline schemes in terms of algorithm convergence, transmission delay, and transmission power.
Published in: IEEE Wireless Communications Letters ( Volume: 14, Issue: 1, January 2025)