Self-optimizing decoders for modern error-correcting codes that promote energy efficiency on the basis sufficient quality

点击翻译按钮获取中文摘要

Channel coding is an essential ingredient of all modern communication systems. Decoders of such codes are often one of the most computationally intensive and power hungry task in modern wireless receivers. One of the latest innovations in channel coding are polar codes, a type of codes that can provably achieve the capacity of all practically relevant channels. The widely-recognized disruptive impact of these codes is illustrated by the recent, rapid adoption of this class of codes as part of 3GPP's next-generation mobile-communication standard (5G). Despite the recent progress in the design of advanced decoding algorithms and corresponding decoders, compared to the achievements of 20+ research years in efficient decoder design for the competing low-density parity-check (LDPC) and turbo codes, polar decoders are still in their infancy, lagging behind in terms of area and energy efficiency.An important and fundamental reason for the limited (area and energy) efficiency of todays polar decoders is deeply rooted in the nature of close-to-optimal decoding strategies. These advanced algorithms explore multiple hypotheses to maximize the chance for success (i.e., finding the correct codeword). Unfortunately, in order to always meet expectations on sampling rate, throughput, and latency, all of these hypotheses must typically be considered in parallel whereas the competing codes use iterative schemes. In practice, this leads to severe overhead, notably in terms of area and power consumption, which is unnecessary under the most frequent average operating conditions. Most of the resources are wasted on calculations for results that are, in the end, discarded.The proposed research program focuses on the creation of novel algorithms and decoder implementations (software or hardware) for polar codes that are energy efficient while catering sufficient quality for the application. The broader objective is to demonstrate that not only we can live with the uncertainty of a variable runtime but that by doing so the energy efficiency is improved.