MIMO Detection on Amplify–Quantize–Forward Relay Channel

This paper investigates multiple-input multiple-output (MIMO) amplify–quantize–forward (AQF) relay channels, which consists of one source, one destination, and one relay. In this channel, the relay uniformly quantizes amplified received signals before forwarding them to the destination. Firstly, a maximum likelihood (ML) detection at the destination, known for its high computational complexity, is presented for the MIMO AQF relay channels. To address this complexity, a suboptimal near-ML (NML) detection method is proposed. Furthermore, a MIMO deep learning (MIMO-DL) detection approach is proposed by strategically modifying the metric used in the NML detection, resulting in significantly reduced complexity. The MIMO-DL detection utilizes multiple parallel detection networks, which are trained to detect signals in real-time for time-varying fading channels. Numerical results validate that the NML detection achieves error performance very close to that of the ML detection. Moreover, the MIMO-DL detection, employing only a few number of detection networks, achieves error performance comparable to the NML detection within the interested signal-to-noise ratio (SNR) region. In particular, the MIMO-DL detection method facilitates the use of memory-limited AQF relays in the MIMO relay channels.