Robust MLSD for Wideband SIMO Systems with One-Bit ADCs: Reinforcement-Learning Approach

This paper considers a wideband single-input multiple- output (SIMO) system with one-bit analog-to-digital converters (ADCs). In this system, a robust maximum- likelihood-sequence-detector (MLSD) is proposed under imperfect channel state information at receiver (CSIR). Inspired by reinforcement learning theory, the key idea of the proposed detector is to enhance the reliability of detection metrics by exploiting a data set that consists of previously detected information symbols and the one-bit quantized outputs. By learning the conditional probability mass function of the system with the data set, the proposed MLSD algorithm reduces a model error in the computation of the detection metrics, which is caused by imperfect CSIR. In simulations, it is shown that the proposed MLSD provides a significant gain in terms of frame-error- rates compared to the conventional MLSD algorithm under imperfect CSIR.