Near-ML Signal Detection in Large-Dimension Linear Vector Channels Using Reactive Tabu Search

TL;DR

This paper introduces a reactive tabu search algorithm for low-complexity, near-ML signal detection in large-dimensional linear vector channels, including MIMO and ISI channels, demonstrating scalability and improved performance over existing methods.

Contribution

The paper presents a novel RTS heuristic for large-dimension signal detection, outperforming traditional algorithms in scalability and near-ML accuracy.

Findings

RTS scales well for large dimensions.

RTS achieves near-ML performance as dimensions increase.

Traditional sphere decoding variants do not scale effectively.

Abstract

Low-complexity near-optimal signal detection in large dimensional communication systems is a challenge. In this paper, we present a reactive tabu search (RTS) algorithm, a heuristic based combinatorial optimization technique, to achieve low-complexity near-maximum likelihood (ML) signal detection in linear vector channels with large dimensions. Two practically important large-dimension linear vector channels are considered: i) multiple-input multiple-output (MIMO) channels with large number (tens) of transmit and receive antennas, and ii) severely delay-spread MIMO inter-symbol interference (ISI) channels with large number (tens to hundreds) of multipath components. These channels are of interest because the former offers the benefit of increased spectral efficiency (several tens of bps/Hz) and the latter offers the benefit of high time-diversity orders. Our simulation results show that, while algorithms including variants of sphere decoding do not scale well for large dimensions, the proposed RTS algorithm scales well for signal detection in large dimensions while achieving increasingly closer to ML performance for increasing number of dimensions.