Accelerating Iterative Detection for Spatially Coupled Systems by Collaborative Training

TL;DR

This paper introduces a collaborative training method to accelerate iterative detection in spatially coupled systems, notably SC CDMA, by creating shortcuts among training subsystems to reduce iteration count.

Contribution

It presents a novel collaborative training approach with shortcuts for faster detection in spatially coupled systems, improving convergence speed.

Findings

Significant reduction in iteration count for highly loaded SC CDMA systems

Effective information spreading through the introduced shortcuts

Enhanced detection performance demonstrated via Density Evolution analysis

Abstract

This letter proposes a novel method for accelerating iterative detection for spatially coupled (SC) systems. An SC system is constructed by one-dimensional coupling of many subsystems, which are classified into training and propagation parts. An irregular structure is introduced into the subsystems in the training part so that information in that part can be detected successfully. The obtained reliable information may spread over the whole system via the subsystems in the propagation part. In order to allow the subsystems in the training part to collaborate, shortcuts between them are created to accelerate iterative detection for that part. As an example of SC systems, SC code-division multiple-access (CDMA) systems are considered. Density Evolution for the SC CDMA systems shows that the proposed method can provide a significant reduction in the number of iterations for highly loaded systems.