Reliable Uplink Transmission for Low-Throughput Massive MIMO Networks with Finite Blocklength

This paper investigates the block error rate (BLER) of three encoding schemes: joint encoding (JE), separate encoding (SE), and rate splitting (RS) in low-throughput massive multiple input and multiple output (mMIMO) networks with finite blocklength transmission. To reduce pilot overhead, we first partition users via k-means clustering based on their spatial coordinates, by which the pilot sequences can be reused among different groups. Then, we apply a minimum mean square error (MMSE) estimator to examine their instantaneous signal-to-interference-plus-noise ratios (SINRs). In particular, a local successful interference cancellation is proposed to mitigate co-channel interference for the RS scheme. Then, we formulate min-max problems for optimizing the error performance, with power control coefficients, combining vectors, and blocklength being their variables. Additionally, we employ the iterative MMSE combiner as the combining vector and then transform the problem for minimizing the BLER to a max-min problem maximizing the SINR at each iteration based on the monotonicity of the BLER expression with respect to the SINR. Power control coefficients are then updated by solving the reformulated geographic programming (GP) problem. Furthermore, the golden-section and bisection methods are applied to optimize the blocklength allocation strategy for the SE scheme and to search for the achievable BLER for the RS scheme. Simulation results, using network availability as the performance metric, demonstrate the comparable performance of the JE and RS schemes when combined with the proposed solving methods, whereas the SE scheme exhibits moderate performance.