Massive Unsourced Random Access with Tensor-Based Modulation and Multi-Antenna Receiver over Frequency-Selective Channels

We present a transceiver design for unsourced random access where users confront frequency-selective channels. We consider that transmitters use Tensor-Based Modulation (TBM) as a modulation scheme, i.e., their messages are mapped onto rank-one tensors. When wireless channels are modeled as flat channels, a canonical polyadic decomposition (CPD) can efficiently perform user separation. The frequency selectivity assumption breaks the CPD structure. We introduce a tensor decomposition, considering this assumption with a minimum knowledge of the specific channel conditions. We show through numerical evaluations that this receiver design leads to better decoding performance in almost the same order of complexity as the state-of-the-art CPD-based receiver.