Improving 5G NR Uplink Channel Estimation with Artificial Neural Networks: A Practical Study on NR PUSCH Receiver

In this paper, an artificial neural network-based approach is proposed to improve uplink channel estimation for 5G New Radio (NR) Physical Uplink Shared Channel (PUSCH). Specifically, the Artificial Neural Network (ANN) is used to in-terpolate the least squares (LS) estimates obtained from PUSCH demodulation reference signal (DMRS) over the entire resource grid. The conventional and ANN-based methods are compared with respect to their channel estimation accuracy and corresponding decoding performance for different 3GPP NR-compliant configurations and under different channel conditions. A special focus is given to high-speed cases in which channel frequency response (CFR) changes rapidly. It is observed that ANN-based method can yield more accurate channel estimation with less DMRS signaling, leading to better decoding performance and enhanced uplink data capacity. It is also noted that with proper training ANN-based interpolator can develop noise and inter-layer interference mitigation capability.