Coded Schemes for Asymmetric Wireless Interfaces: Theory and Practice

A fundamental understanding of the delay and throughput benefits of network coding over multiple, heterogeneous half-duplex communication interfaces is critical to fully exploit current and future devices that incorporate multiple communication technologies. The goal of this paper is two-fold. First, to present fundamental limits and several strategies to trade-off data and feedback when using multiple interfaces. Our work sets forth a systematic approach to (i) analyze a variety of schemes in the presence of channels with heterogeneous round trip delays, packet transmission rates, and packet loss probability, and to (ii) determine the optimal number of transmissions to be allocated to each channel based on these characteristics. Our analytical results show that the gains over a variety of uncoded approaches can be several fold and that our proposed strategies are within 1 dB to an optimal system with full duplex interfaces. Our second goal is to understand the practical implications of these results by designing a protocol for file transmissions, implement it in Android smart phones, and measure its performance when combining various interfaces, including, Bluetooth, WiFi, and 3G cellular networks. Our measurements show that throughput is not only increased by the aggregation of multiple interfaces, but by providing a much more robust erasure correction mechanism.