Multilevel Coding with General Decoding Metrics and Rateless Transmission

Abstract

Multilevel coding (MLC) is the main contender to the celebrated bit-interleaved coded modulation (BICM) technique for combining binary error-control codes with multilevel constellations. Although MLC has a more complex encoding-decoding structure, it can achieve a larger rate in a number of important scenarios such as multiple-input multiple-output (MIMO) and orthogonal modulation transmission. In this paper, we consider two issues related to the application of MLC. First, we examine the use of general decoding metrics in MLC, including mismatched metrics that arise from approximations to reduce detection complexity. We make use of recent advances in the analysis of BICM and apply those techniques to individual MLC transmission layers. Our contributions include rate analysis and metric-mismatch correction to improve throughput performance of MLC. Second, we consider the combination of MLC with binary rateless codes. Such a combination eliminates the need to carefully design code rate for each MLC layer. In slow fading environments, rateless coding can also seamlessly adapt to the instantaneous channel quality and achieve an increased average throughput compared to a fixed-rate MLC transmission. However, due to the MLC structure, we show that a naive combination of MLC and rateless coding can cause a significant rate loss. Thus, we propose a novel rotation rateless scheme which preserves the rate advantage of MLC over BICM. We provide relevant examples with MIMO, frequency-shift keying (FSK), and pulse-position modulation (PPM) signaling to demonstrate that our scheme can achieve throughput gains compared to BICM in a variety of transmission scenarios.