Unsourced Random Access with Unequal Power Dynamic Compressed Sensing

In coded compressed sensing (CCS) unsourced random access (URA), the message of each user is subdivided into smaller sub-messages. These are mapped into columns of sensing matrices and concatenated to form the user’s packet. In order to stitch individually reconstructed sub-messages at the receiver, parity-check constraints from prior encoding by an outer code are exploited. The dynamic compressed sensing (DCS) allows to stitch each user’s sub-sequences without the need for prior encoding and the associated redundancy. In this paper, an unequal power DCS is proposed that allocates a specific power distribution to the columns of the sensing matrices. The power distribution is designed via the proposed state evolution (SE) analysis of the receiver processing. Simulation results demonstrate that the proposed unequal power DCS outperforms state-of-the-art URA in terms of energy-per-bit required to support a vast range of active user numbers, from medium to extremely large, unachievable by the current URA systems.