Performance / Complexity Trade-offs of the Sphere Decoder Algorithm for Massive MIMO Systems

TL;DR

This paper explores the trade-offs between complexity and performance in sphere decoding for massive MIMO systems, proposing parallel and hybrid algorithms that significantly improve efficiency while maintaining detection accuracy.

Contribution

It introduces a parallel high-level sphere decoding scheme and a hybrid SD/K-best approach, enabling efficient detection in large MIMO systems up to 100x100.

Findings

Parallel SD outperforms state-of-the-art by over 5x in speed

Hybrid SD/K-best reduces complexity while maintaining accuracy

Super-linear speedup achieved on multicore platforms

Abstract

Massive MIMO systems are seen by many researchers as a paramount technology toward next generation networks. This technology consists of hundreds of antennas that are capable of sending and receiving simultaneously a huge amount of data. One of the main challenges when using this technology is the necessity of an efficient decoding framework. The latter must guarantee both a low complexity and a good signal detection accuracy. The Sphere Decoder (SD) algorithm represents one of the promising decoding algorithms in terms of detection accuracy. However, it is inefficient for dealing with large MIMO systems due to its prohibitive complexity. To overcome this drawback, we propose to revisit the sequential SD algorithm and implement several variants that aim at finding appropriate trade-offs between complexity and performance. Then, we propose an efficient high-level parallel SD scheme based on the master/worker paradigm, which permits multiple SD instances to simultaneously explore the search space, while mitigating the overheads from load imbalance. The results of our parallel SD implementation outperform the state-of-the-art by more than 5x using similar MIMO configuration systems, and show a super-linear speedup on multicore platforms. Moreover, this paper presents a new hybrid implementation that combines the strengths of SD and K-best algorithms, i.e., maintaining the detection accuracy of SD, while reducing the complexity using the K-best way of pruning search space. The hybrid approach extends our parallel SD implementation: the master contains the SD search tree, and the workers use the K-best algorithm to accelerate its exploration. The resulting hybrid approach enhances the diversification gain, and therefore, lowers the overall complexity. Our synergistic hybrid approach permits to deal with large MIMO configurations up to 100x100, without sacrificing the accuracy and complexity.