Blank Space: Adaptive Causal Coding for Streaming Communications Over Multi-Hop Networks

TL;DR

The paper tackles efficient real-time streaming over multi-hop wireless networks by introducing a bottleneck-aware, adaptive causal coding framework (BS-AC-RLNC) with a lightweight intermediate-node encoder NET. It combines semi-decoding, a-priori and posterior FEC, and two idle-time mechanisms—Blank Space Period and No-New No-FEC—to reduce channel usage without sacrificing delivery performance or delay. Empirical results show about a 20% end-to-end channel usage reduction, with preserved goodput and URLLC-like delay bounds, and stronger gains when bottlenecks shift along the path. The approach enables more resource-efficient multi-hop streaming and lays groundwork for scalable, bottleneck-aware coding in future networks.

Abstract

In this work, we introduce Blank Space AC-RLNC (BS), a novel Adaptive and Causal Network Coding (AC-RLNC) solution designed to mitigate the triplet trade-off between throughput-delay-efficiency in multi-hop networks. BS leverages the network's physical limitations considering the bottleneck from each node to the destination. In particular, BS introduces a light-computational re-encoding algorithm, called Network AC-RLNC (NET), implemented independently at intermediate nodes. NET adaptively adjusts the Forward Error Correction (FEC) rates and schedules idle periods. It incorporates two distinct suspension mechanisms: 1) Blank Space Period, accounting for the forward-channels bottleneck, and 2) No-New No-FEC approach, based on data availability. The experimental results achieve significant improvements in resource efficiency, demonstrating a 20% reduction in channel usage compared to baseline RLNC solutions. Notably, these efficiency gains are achieved while maintaining competitive throughput and delay performance, ensuring improved resource utilization does not compromise network performance.

Figures (4)

• Figure 1: Multi-hop communication system with $N$ nodes. The dashed line represents the forward bottleneck from node $n$ to the destination estimated for the proposed Network AC-RLNC in BS. Here, for example, for node $n=1$ the estimated forward bottleneck, $\epsilon_1$, is larger from $\{\epsilon_2,\ldots,\epsilon_{N-2}\}$, but not necessarily from $\epsilon_0$.
• Figure 2: The diagram illustrates how FEC periods at intermediate nodes create blank spaces at preceding nodes. Vertical lines represent each node transmission timeline, with packets labeled by type and generator node (The labels indicate the information added to the output DoF at each node). The colors indicate period lengths corresponding to each channel erasure rate. Dashed lines mark blank-space regions where packet transmissions may be ineffective. For example, FEC(0) represents a FEC period generated at node $0$ with length $\lceil \hat{\epsilon}_0 \cdot \rm RTT_0 \rfloor$ and sent to node 1. Then, node $1$ first generates its own FEC period, FEC(1), before processing any DoF from FEC(0). This creates the time window BS(1) in node 0's timeline where any transmitted packets will not be immediately processed. Similarly, BS(2) propagates from node $2$'s FEC period. Since node $2$ cannot process packets during this time, node $1$ suspends transmission, which in turn allows node $0$ to pause as well.
• Figure 3: System performance of 6-nodes network with respect to $\epsilon_2 \in [0.2, 0.6]$ where $\epsilon_0=\epsilon_4=0.1$, $\epsilon_1=0.4$ and $\epsilon_3=0.3$.
• Figure 4: Analysis of multi-cast Scenario when all nodes function as destinations.