Loading...
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 57450 (05)
- University: Sharif University of Technology
- Department: Electrical Engineering
- Advisor(s): Hossein Khalaj, Babak
- Abstract:
- In recent years, virtual reality (VR) and the Metaverse have gained increased attention in industry and exhibitions. However, rendering high-quality scenes demands significant computing power that cannot be directly handled by VR headsets. In this paper, we propose a novel approach to distribute rendering between Mobile Edge Computing (MEC) and cloud servers. Our method involves the cloud server predicting head positions in advance to reduce round-trip time, while the MEC server re-renders frames where the predicted head location is substantial. We evaluate this resource allocation scheme using two prediction methods: Kalman filter and LSTM Encoder-Decoder. By considering users’ preferred accuracy, we ensure fairness in experienced location errors. Our evaluation, based on a dataset containing over 4500 seconds of recorded head poses from users exploring VR, demonstrates that this strategy significantly increases the maximum number of users each MEC
server can support, from 5 to 30 - Keywords:
- Virtual Reality (VR)Environment ; Long Short Term Memory (LSTM) ; Distributed Processing ; Edge Computing ; Augmented Reality ; Head Position Prediction ; Metaverse
-
محتواي کتاب
- view
