CIF: Small: Approaching Capacity in High Throughput Communication Systems with Incremental Redundancy

产生数字信息的能力不断增强，人们对这些信息的消费欲望也越来越强烈，因此需要以更高的速度交流信息。自由空间光学和毫米波无线通信系统努力实现超过100 Gbps的速率。光纤支持每秒太比特的通信。这些系统需要高速解码电路，目前牺牲性能以满足可用计算资源的延迟要求。本研究开发了高吞吐量通信的新技术，其目标是在满足合理复杂性的延迟要求的同时达到理论性能限制。除了为高通量系统的实际解决方案提供一条新途径外，这项研究的更广泛影响还包括培训研究生，以支持美国在通信技术方面的持续领导地位，以及为本科生提供实践研究经验，以维持供应我们国家的管道？美国对多样化、训练有素和具有创新精神的工程师的持续需求。本研究探索了一种新的范式，在这种范式中，许多低复杂度的解码器并行工作，解码具有短平均块长度的变长码。块长度短的定长码不能接近香农容量，而平均块长度短的变长码可以接近香农容量。变长码传统上是用反馈实现的，这样增量冗余只在接收方请求时才传输。本研究在不使用反馈的情况下传输增量冗余，利用网络编码和遍历等技术仅向需要冗余的并行解码器提供冗余。在没有反馈的情况下提供增量冗余的问题被认为是一个具有接收端侧信息的联合源信道编码问题。本研究的主要目标是生产低复杂性的系统，可以接近高吞吐量的容量。

The growing ability to generate digital information and the hunger to consume that information create a need to communicate information at ever-higher rates. Free-space optical and millimeter wave wireless communication systems strive to achieve rates exceeding 100 Gbps. Optical fiber supports terabit-per-second communications. These systems require high-speed decoding circuits that currently sacrifice performance to meet latency requirements with available computational resources. This research develops new techniques for high-throughput communications with a goal of achieving theoretical performance limits while meeting latency requirements with reasonable complexity. In addition to providing a new path for practical solutions for high-throughput systems, the broader impacts of this research include the training of graduate students to support continued leadership of the United States in communications technology as well as hands-on research experiences for undergraduates to maintain the pipeline that will supply our country?s continuing need for diverse, well-trained, and innovative engineers.This research explores a new paradigm in which many low-complexity decoders work in parallel, decoding variable-length codes with short average block-lengths. Fixed-length codes with short block-lengths cannot approach the Shannon capacity, but variable-length codes with short average block-lengths do approach capacity. Variable-length codes are traditionally implemented with feedback so that incremental redundancy is transmitted only when requested by the receiver. This research transmits incremental redundancy without the use of feedback, utilizing techniques such as network coding and ergodicity to deliver redundancy only to those parallel decoders that need it. The problem of delivering incremental redundancy without feedback is framed as a joint source-channel coding problem with side information at the receiver. The primary goal of this research is to produce low complexity systems that can approach capacity at high throughput.