Revised Maximum-Likelihood Detector With Quantization Design for One-Bit Massive MIMO Systems

Abstract

One-bit analog-to-digital converters (ADCs) offer a practical solution for reducing both cost and power consumption in massive multiple-input multiple-output (MIMO) systems. Nevertheless, the severe distortion induced by extremely coarse quantization significantly deteriorates the data detection performance. For 1-bit ADC, the conventional symmetric quantization may not necessarily be the optimal choice. This paper examines the impact of quantization thresholds on the performance of maximum likelihood (ML) data detection in one-bit massive MIMO systems, and proposes a revised ML (rML) detector with an iterative adaptive quantization design. Initially, the original ML detection problem, constrained by discrete constellations, is reformulated into the rML problem. By optimizing the quantization thresholds, the solution to the rML problem can be guided to converge to the true transmitted signal. Leveraging this characteristic, we propose the rML detector with iterative adaptive quantization design to progressively refine the quantization thresholds during data detection. We further present two implementation strategies for the rML detector: one iteration one quantization (1I1Q) and one iteration two quantization (1I2Q). Additionally, an rML-based joint channel estimation and data detection (JED) method is introduced, where the decoded data from 1I1Q is utilized to enhance the pilot data vectors, thereby refining the estimated channel and ultimately enhancing data detection performance. Finally, numerical results demonstrate that the proposed rML detector with iterative adaptive quantization design is both efficient and robust, significantly outperforming the state-of-the-art methods.