Achievable Second-Order Asymptotics for MIMO MAC With Additive Noise Under Nearest Neighbor Decoding

Motivated by the need for low-latency and high-throughput in the low-altitude economy scenarios with multiple hovering uncrewed aerial vehicles and a single base station, we investigate a multiple-input multiple-output (MIMO) multiple-access channel (MAC) with arbitrary additive noise distribution. To characterize its finite blocklength performance, we propose mismatched coding schemes based on spherical codebooks, employing either joint nearest neighbor (JNN) or successive interference cancellation (SIC) decoders. We derive second-order asymptotics for these schemes, extending single-antenna MAC analyses to the MIMO setting and highlighting the impact of antenna number on performance. While JNN and SIC decodings yield identical first-order asymptotics, JNN decoding exhibits superior performance in low-latency communication due to a larger second-order rate region, with this performance gap increasing as the number of antennas grows.