{"title":"Partial Learning-Based Iterative Detection of MIMO Systems","authors":"Abdulaziz Babulghum;Chao Xu;Soon Xin Ng;Mohammed El-Hajjar","doi":"10.1109/OJVT.2024.3482008","DOIUrl":null,"url":null,"abstract":"One of the major challenges in multiple input multiple output (MIMO) system design is the salient trade-off between performance and computational complexity. For instance, the maximum likelihood (Max-L) detection is capable of achieving optimal performance based on exhaustive search, but its exponential computational complexity renders it impractical. By contrast, zero-forcing detection has low computational complexity, while having significantly worse performance compared to that of the Max-L. The recent developments in deep learning (DL) based detection techniques relying on back propagation neural networks (BPNN) constitute promising candidates for the open challenge of the MIMO detection performance versus complexity trade-off. Against this background, in this paper, we propose a novel partial learning (PL) model for MIMO detection with soft-bit decisions that can be incorporated into channel-coded communication systems. More explicitly, the proposed PL model consists of two parts: first, a subset of the transmitted MIMO symbols is detected by the data-driven DL technique and then the detected symbols are removed from the received MIMO signals for the sake of interference cancellation. Afterwards, the classic model-based zero-forcing detector is invoked to detect the remaining symbols at a linear complexity. As a result, near-optimal MIMO performance can be achieved with substantially reduced computational complexity compared to Max-L and BPNN. The proposed solution is adapted to both accept and produce soft information, so that iterative detection can be performed, where the iteration gain is analyzed by extrinsic information transfer (EXIT) charts. Our simulation results demonstrate that the proposed partial learning-based iterative detection is capable of attaining near-Max-L performance while attaining a flexible performance versus complexity trade-off.","PeriodicalId":34270,"journal":{"name":"IEEE Open Journal of Vehicular Technology","volume":null,"pages":null},"PeriodicalIF":5.3000,"publicationDate":"2024-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10720516","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Open Journal of Vehicular Technology","FirstCategoryId":"1085","ListUrlMain":"https://ieeexplore.ieee.org/document/10720516/","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
引用次数: 0
Abstract
One of the major challenges in multiple input multiple output (MIMO) system design is the salient trade-off between performance and computational complexity. For instance, the maximum likelihood (Max-L) detection is capable of achieving optimal performance based on exhaustive search, but its exponential computational complexity renders it impractical. By contrast, zero-forcing detection has low computational complexity, while having significantly worse performance compared to that of the Max-L. The recent developments in deep learning (DL) based detection techniques relying on back propagation neural networks (BPNN) constitute promising candidates for the open challenge of the MIMO detection performance versus complexity trade-off. Against this background, in this paper, we propose a novel partial learning (PL) model for MIMO detection with soft-bit decisions that can be incorporated into channel-coded communication systems. More explicitly, the proposed PL model consists of two parts: first, a subset of the transmitted MIMO symbols is detected by the data-driven DL technique and then the detected symbols are removed from the received MIMO signals for the sake of interference cancellation. Afterwards, the classic model-based zero-forcing detector is invoked to detect the remaining symbols at a linear complexity. As a result, near-optimal MIMO performance can be achieved with substantially reduced computational complexity compared to Max-L and BPNN. The proposed solution is adapted to both accept and produce soft information, so that iterative detection can be performed, where the iteration gain is analyzed by extrinsic information transfer (EXIT) charts. Our simulation results demonstrate that the proposed partial learning-based iterative detection is capable of attaining near-Max-L performance while attaining a flexible performance versus complexity trade-off.
多输入多输出(MIMO)系统设计的主要挑战之一是性能与计算复杂度之间的突出权衡。例如,最大似然(Max-L)检测能够在穷举搜索的基础上实现最佳性能,但其指数级的计算复杂度使其不切实际。相比之下,零强迫检测的计算复杂度较低,但性能却比 Max-L 差很多。基于反向传播神经网络(BPNN)的深度学习(DL)检测技术的最新发展,为解决 MIMO 检测性能与复杂性权衡这一难题带来了希望。在此背景下,我们在本文中提出了一种新颖的部分学习(PL)模型,用于可纳入信道编码通信系统的具有软位决策的 MIMO 检测。更明确地说,所提出的 PL 模型由两部分组成:首先,通过数据驱动的 DL 技术检测传输的 MIMO 符号子集,然后为了消除干扰,从接收的 MIMO 信号中去除检测到的符号。之后,再使用基于模型的经典零干扰检测器,以线性复杂度检测剩余的符号。因此,与 Max-L 和 BPNN 相比,在大幅降低计算复杂度的同时,还能实现近乎最佳的 MIMO 性能。所提出的解决方案既能接受软信息,也能产生软信息,因此可以进行迭代检测,迭代增益通过外在信息传输(EXIT)图进行分析。我们的仿真结果表明,所提出的基于部分学习的迭代检测能够达到接近 Max-L 的性能,同时实现灵活的性能与复杂性权衡。