Adapt or Wait: Quality Adaptation for Cache-aided Channels

TL;DR

This work addresses delivering varied content quality to heterogeneous users in cache-aided wireless networks with degraded channels. It introduces a framework that combines coded caching, scalable source coding, and superposition coding, while keeping caching oblivious to channel states, to adapt per-user quality and meet delivery-time targets. The authors derive a general delivery-time expression, analyze a two-type-user case, and propose three quality-allocation algorithms (Proportional fairness, Max-Min, and Sum-quality) that balance fairness, QoS, and overall quality. The framework demonstrates that modest quality adjustments at degraded users can counter significant channel degradation and that high-rate users can gain quality via multicast opportunities without delaying others, with practical implications for streaming and chunked content delivery.

Abstract

This work focuses on quality adaptation as a means to counter the effects of channel degradation in wireless, cache-aided channels. We design a delivery scheme which combines coded caching, superposition coding, and scalable source coding, while keeping the caching scheme oblivious to channel qualities. By properly adjusting the quality at the degraded users we are able to satisfy all demands in a time-efficient manner. In addition, superposition coding allows us to serve high-rate users with high content quality without subjecting them to a delay penalty caused by users with lower rate channels. We design a communication framework that covers all possible channel rate and quality configurations and we further provide algorithms that can optimise the served quality. An interesting outcome of this work is that a modest quality reduction at the degraded users can counter the effects of significant channel degradation. For example, in a 100-user system with normalized cache size 1/10 at each user, if 10 users experience channel degradation of 60% compared to the rate of the non-degraded users, we show that our transmission strategy leads to a 85% quality at the degraded users and perfect quality at the non-degraded users.

Figures (5)

• Figure 1: Comparison of the quality allocation using the three afore-mentioned algorithms. Baseline corresponding to $\mathbf{Q} \!=\! \boldsymbol\alpha$ (red solid line with stars). Fairness optimisation (blue solid with circles). MaxMin optimisation (green solid with squares). MaxSum quality (black dashed with pentagons). Setting: $K\!=\!20$ users, $\gamma=\frac{3}{20}$, $\alpha_{k} = 0.8 + 0.2\frac{k-1}{K-1}$, $T_{\text{tar}}= T_{\text{MAN}}$.
• Figure 2: File quality at the degraded users as a function of the channel degradation required to achieve delivery time $T_{\text{MAN}}$ for the two-type case. Setting: $K=100$-user channel with normalized cache $\gamma=\frac{1}{10}$ at each receiver. The baseline case corresponds to quality $Q=\alpha$.
• Figure 3: Quality boost over the $Q \!=\! \alpha$ baseline as a function of the number of degraded users, in a channel with $K\!=\!100$ users, for various values of $\gamma$.
• Figure 4: Comparison of the per-user quality achieved by each of the algorithms. Baseline corresponding to $\mathbf{Q} \!=\! \boldsymbol\alpha$ (red solid line with stars). Proportional Fairness optimisation (blue solid with circles). MaxMin optimisation (green solid with squares). MaxSum quality (black dashed with pentagons). Setting: $K\!=\!20$ users, $\alpha_{k} = 0.8 + 0.2\frac{k-1}{K-1}$, $T_{\text{tar}}= T_{\text{MAN}}$. The double line appearing in the fairness results in (a) is the outcome of applying the algorithm twice in order to achieve the optimal result.
• Figure 5: File quality calculated by each of the three methods, for the multi-rate setting of the example presented in Sec. \ref{['exMultiRates1']}.

Theorems & Definitions (12)

• Theorem 1
• proof
• Theorem 2
• proof
• Corollary 1
• proof
• Corollary 2
• proof
• Corollary 3
• proof