Channel quantization based physical-layer network coding

This paper studies a MIMO Two-Way Relay Channel (TWRC) where two single-antenna nodes communicate with each other through a K-antenna relay. In the uplink phase, the two end nodes send their respective information simultaneously to the relay, and it is usually regarded as a virtual MIMO system. With traditional VBLAST MIMO detection, the maximum achievable diversity is K - 1. By noting that only the network coded packet, rather than the two individual packets, is needed at the relay node in TWRC, we propose a channel quantized physical-layer network coding (CQ-PNC) scheme based on VBLAST to achieve the full diversity K. Specifically, relay first uses QR decomposition to convert the K received signals into two valid signals, the first signal is a weighted summation of the two end nodes´ symbols and the second signal is a scaled version of one end node´s symbol. The relay first quantizes the first signal with the channel coefficient of one node and estimate the Gaussian integer summation of the two end nodes´ information. At the same time, the quantization error is further canceled with the help of the second signal. After that, we adaptively map the Gaussian integer weighted summation to the network coding form. Moreover, we theoretically prove that our CQ-PNC can achieve the maximum diversity K. Finally, the numerical simulation shows that CQ-PNC performs within 2dB gap from the theoretical bound.