Massively Parallel Detection for Non-Orthogonal Signal Transmissions

The increasing demand for massive connectivity with low latency requirements has triggered a paradigm shift towards Non-Orthogonal transmissions. Still, to translate the theoretical gains of Non-Orthogonal transmissions into practical, efficient “soft” detection schemes are required. The detection latency and/or complexity of state-of-the-art detection methods becomes impractical for large Non-Orthogonal systems, both due to the large number of interfering streams and due to the rank-deficient or ill-determined nature of the corresponding interference matrix. Extending the recently proposed MultiSphere framework, this work introduces NorthCore; a massively parallel sphere-decoding-based scheme for the detection of large and ill-determined Non-Orthogonal systems. Similarly to MultiSphere, NorthCore reduces the corresponding search space by focusing the available processing power to the most promising vector solutions that are then processed in parallel. As a result, the proposed detection scheme can attain a detection processing latency similar to that of highly-suboptimal linear detectors and even outperform state-of-the-art sophisticated detection approaches with up to an order of magnitude reduced complexity. To identify the most promising vector solutions, NorthCore introduces a sort-free candidate selection technique that reduces the necessary preprocessing complexity by up to an order of magnitude, making the proposed approach practical.