Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Context-aware Rate Adaptation for Predictive Flying Networks using Contextual Bandits

Ruben Queiros, Megumi Kaneko, Helder Fontes, Rui Campos

TL;DR

This work tackles rate adaptation in predictive Flying Networks where high mobility and a large configuration space hinder conventional RA methods. It formulates RA as a contextual bandit problem and introduces LinRA, a LinUCB-based algorithm that uses contextual features such as link distance and obstacle presence to predict throughput and select the optimal MCS in real time. Key contributions include a context-aware RA formulation, the LinRA algorithm with per-arm linear models and adaptive exploration, and a performance study showing LinRA converges up to 5.2× faster than Thompson Sampling and yields up to an 80% throughput improvement in NLoS, approaching Oracle performance in several scenarios. The approach offers robust, scalable, and real-time RA for predictive Flying Networks, with future work addressing delayed context and experimental validation to bridge simulation and real-world deployment.

Abstract

The increasing complexity of wireless technologies, such as Wi-Fi, presents significant challenges for Rate Adaptation (RA) due to the large configuration space of transmission parameters. While extensive research has been conducted on RA for low-mobility networks, existing solutions fail to adapt in flying networks, where high mobility and dynamic wireless conditions introduce additional uncertainty. We propose Linear Upper Confidence Bound for RA (LinRA), a novel Contextual Bandit-based approach that leverages real-time link context to optimize transmission rates. Designed for predictive flying networks, where future trajectories are known, LinRA proactively adapts to obstacles affecting channel quality. Simulation results demonstrate that LinRA converges $\mathbf{5.2\times}$ faster than state-of-the-art benchmarks and improves throughput by 80\% in Non Line-of-Sight (NLoS) conditions, matching the performance of ideal algorithms. With low time complexity, LinRA is a scalable and efficient RA solution for predictive flying networks.

Context-aware Rate Adaptation for Predictive Flying Networks using Contextual Bandits

TL;DR

This work tackles rate adaptation in predictive Flying Networks where high mobility and a large configuration space hinder conventional RA methods. It formulates RA as a contextual bandit problem and introduces LinRA, a LinUCB-based algorithm that uses contextual features such as link distance and obstacle presence to predict throughput and select the optimal MCS in real time. Key contributions include a context-aware RA formulation, the LinRA algorithm with per-arm linear models and adaptive exploration, and a performance study showing LinRA converges up to 5.2× faster than Thompson Sampling and yields up to an 80% throughput improvement in NLoS, approaching Oracle performance in several scenarios. The approach offers robust, scalable, and real-time RA for predictive Flying Networks, with future work addressing delayed context and experimental validation to bridge simulation and real-world deployment.

Abstract

The increasing complexity of wireless technologies, such as Wi-Fi, presents significant challenges for Rate Adaptation (RA) due to the large configuration space of transmission parameters. While extensive research has been conducted on RA for low-mobility networks, existing solutions fail to adapt in flying networks, where high mobility and dynamic wireless conditions introduce additional uncertainty. We propose Linear Upper Confidence Bound for RA (LinRA), a novel Contextual Bandit-based approach that leverages real-time link context to optimize transmission rates. Designed for predictive flying networks, where future trajectories are known, LinRA proactively adapts to obstacles affecting channel quality. Simulation results demonstrate that LinRA converges faster than state-of-the-art benchmarks and improves throughput by 80\% in Non Line-of-Sight (NLoS) conditions, matching the performance of ideal algorithms. With low time complexity, LinRA is a scalable and efficient RA solution for predictive flying networks.

Paper Structure

This paper contains 12 sections, 8 equations, 3 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model
  3. Channel Model
  4. Rate Model
  5. Problem Formulation
  6. Proposed Solution: LinUCB for Rate Adaptation (LinRA)
  7. Performance Evaluation
  8. Rate Adaptation Benchmarks
  9. Simulation Setup
  10. Simulation Results
  11. Time Complexity Analysis
  12. Conclusions and Future Work

Figures (3)

  • Figure 1: Predictive Flying Network example scenario.
  • Figure 2: Comparison of RA algorithms throughput for a specific random seed.
  • Figure 3: Average throughput of the RA algorithms over 100 simulation using different seeds, focusing on the three transition moments.