Antipodal Detection and Decoding for Large Multi-User MIMO with Reduced Base-Station Antennas
2018 IEEE Globecom Workshops (GC Wkshps)
To avoid unnecessarily using a massive number of base station antennas to support a large number of users spatially multiplexed multi-user MIMO systems, optimal detection methods are required to demultiplex the mutually interfering information streams. Sphere decoding (SD) can achieve this, but its complexity and latency becomes impractical for large MIMO systems. Low complexity detection solutions such as linear detectors (e.g., MMSE) or likelihood ascendant search (LAS) approaches, have
... icantly lower latency requirements than SD but their achievable throughput is far from optimal. This work presents the concept of Antipodal detection and decoding, that can deliver very high throughput with practical latency requirements, even in systems where the number of user antennas reaches the number of base station antennas. The Antipodal detector either results in a highly reliable vector solution, or it does not find a vector solution at all (i.e., it results in an erasure), skipping the heavy processing load related to finding vector solutions that have a very high likelihood to be erroneous. Then, a belief-propagation-based decoder is proposed, that restores these erasures and further corrects remaining erroneous vector solutions. We show that for 32⇥32, 64-QAM modulated systems, and for packet error rates below 10%, Antipodal detection and decoding requires 9 dB less transmitted power than systems employing soft MMSE or LAS detection and LDPC decoding with similar complexity requirements. For the same scenario, our Antipodal method achieves practical throughput gains of more than 50% compared to soft MMSE and soft LAS-based methods.