Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Personalized Federated Learning for Cellular VR: Online Learning and Dynamic Caching

Krishnendu S. Tharakan, Hayssam Dahrouj, Nour Kouzayha, Hesham ElSawy, Tareq Y. Al-Naffouri

TL;DR

The paper tackles caching VR content over wireless networks under highly dynamic FoV requests by introducing a decentralized, personalized DP-FL framework where each BS learns its own FoV caching strategy. It integrates one-bit gradient signaling (OBSGD) to cut communication overhead and groups FoVs into multicast or unicast transmissions to reflect channel conditions, achieving a PAC-based performance guarantee on cache hits and a convergence rate of $O(1/\sqrt{T})$. A delay-aware extension accounts for rendering and transmission times, yielding an $O(1/T)$ convergence in the presence of latency constraints. Empirical results on realistic VR head-tracking datasets show substantial improvements in average cache hits and reduced delay compared with baselines, validating scalability and practical impact for MEC-enabled VR systems.

Abstract

Delivering an immersive experience to virtual reality (VR) users through wireless connectivity offers the freedom to engage from anywhere at any time. Nevertheless, it is challenging to ensure seamless wireless connectivity that delivers real-time and high-quality videos to the VR users. This paper proposes a field of view (FoV) aware caching for mobile edge computing (MEC)-enabled wireless VR network. In particular, the FoV of each VR user is cached/prefetched at the base stations (BSs) based on the caching strategies tailored to each BS. Specifically, decentralized and personalized federated learning (DP-FL) based caching strategies with guarantees are presented. Considering VR systems composed of multiple VR devices and BSs, a DP-FL caching algorithm is implemented at each BS to personalize content delivery for VR users. The utilized DP-FL algorithm guarantees a probably approximately correct (PAC) bound on the conditional average cache hit. Further, to reduce the cost of communicating gradients, one-bit quantization of the stochastic gradient descent (OBSGD) is proposed, and a convergence guarantee of $\mathcal{O}(1/\sqrt{T})$ is obtained for the proposed algorithm, where $T$ is the number of iterations. Additionally, to better account for the wireless channel dynamics, the FoVs are grouped into multicast or unicast groups based on the number of requesting VR users. The performance of the proposed DP-FL algorithm is validated through realistic VR head-tracking dataset, and the proposed algorithm is shown to have better performance in terms of average delay and cache hit as compared to baseline algorithms.

Personalized Federated Learning for Cellular VR: Online Learning and Dynamic Caching

TL;DR

The paper tackles caching VR content over wireless networks under highly dynamic FoV requests by introducing a decentralized, personalized DP-FL framework where each BS learns its own FoV caching strategy. It integrates one-bit gradient signaling (OBSGD) to cut communication overhead and groups FoVs into multicast or unicast transmissions to reflect channel conditions, achieving a PAC-based performance guarantee on cache hits and a convergence rate of . A delay-aware extension accounts for rendering and transmission times, yielding an convergence in the presence of latency constraints. Empirical results on realistic VR head-tracking datasets show substantial improvements in average cache hits and reduced delay compared with baselines, validating scalability and practical impact for MEC-enabled VR systems.

Abstract

Delivering an immersive experience to virtual reality (VR) users through wireless connectivity offers the freedom to engage from anywhere at any time. Nevertheless, it is challenging to ensure seamless wireless connectivity that delivers real-time and high-quality videos to the VR users. This paper proposes a field of view (FoV) aware caching for mobile edge computing (MEC)-enabled wireless VR network. In particular, the FoV of each VR user is cached/prefetched at the base stations (BSs) based on the caching strategies tailored to each BS. Specifically, decentralized and personalized federated learning (DP-FL) based caching strategies with guarantees are presented. Considering VR systems composed of multiple VR devices and BSs, a DP-FL caching algorithm is implemented at each BS to personalize content delivery for VR users. The utilized DP-FL algorithm guarantees a probably approximately correct (PAC) bound on the conditional average cache hit. Further, to reduce the cost of communicating gradients, one-bit quantization of the stochastic gradient descent (OBSGD) is proposed, and a convergence guarantee of is obtained for the proposed algorithm, where is the number of iterations. Additionally, to better account for the wireless channel dynamics, the FoVs are grouped into multicast or unicast groups based on the number of requesting VR users. The performance of the proposed DP-FL algorithm is validated through realistic VR head-tracking dataset, and the proposed algorithm is shown to have better performance in terms of average delay and cache hit as compared to baseline algorithms.
Paper Structure (17 sections, 5 theorems, 63 equations, 11 figures, 1 table, 2 algorithms)

This paper contains 17 sections, 5 theorems, 63 equations, 11 figures, 1 table, 2 algorithms.

Table of Contents

  1. Introduction
  2. Overview
  3. Related Work
  4. Contributions and Organization
  5. System Model and Problem Formulation
  6. Distributed Online FL Algorithm
  7. Theoretical Guarantees & Proposed Algorithm
  8. Algorithm
  9. Communication Cost
  10. Convergence
  11. Delay-aware Caching Optimization
  12. Simulation Results
  13. Data sets:
  14. Conclusion
  15. Proof of Theorem \ref{['thm:mainresult_cachingvr']}
  16. ...and 2 more sections

Key Result

Theorem 1

(PAC Bound) Given the caching weights and a sequence of caching strategies as in eq:phi_weights, with a probability of at least $1 - \delta$, $\delta>0$, we can establish a lower bound on the conditional expectation of cache hit. i.e. with a high probability the conditional expectation of cache hit where $\mathcal{E}_{\bm{\rho},\bm{\sigma},\upsilon}^T := {\mathcal{C}_{max}} \left\lVert\bm{\sigma_

Figures (11)

  • Figure 1: (a) Event scenario with viewpoints captured (b) FoV-based view model (c) User correlation in FoV
  • Figure 2: Decentralized and personalized federated learning in VR system.
  • Figure 3: Average cache hit versus cache size.
  • Figure 4: Average delay versus cache size.
  • Figure 5: Average cache hit versus no. of BS.
  • ...and 6 more figures

Theorems & Definitions (10)

  • Definition 1
  • Definition 2
  • Definition 3
  • Definition 4
  • Theorem 1
  • Definition 5
  • Lemma 1
  • Lemma 2
  • Theorem 2
  • Theorem 3