Dynamic Compressed Sensing Approach for Unsourced Random Access

In compressed sensing (CS) approach for unsourced random access (URA), each user’s transmitted sequence consists of shorter sub-sequences encoded via sparse regression codes (SPARCs) over several consecutive sub-slots. The approximate message passing (AMP) technique is employed to decode the SPARCs. Stitching together each user’s sub-messages decoded over distinct sub-slots requires an outer parity-check code that adds redundancy bit interconnecting the sub-messages. This letter introduces a novel CS-based URA scheme which is free from the outer code. In the proposed scheme, the encoded sub-sequence of the first sub-slot acts as a temporary user identifier and also customises the sensing matrix used to encode the subsequent sub-messages. The technique allows each user to send several sub-messages (data streams) per sub-slot. Simulation results indicate that the proposed scheme outperforms the existing CS-based algorithms on Gaussian channel. It is also superior than the other state-of-the-art URA schemes when the number of active users exceeds 200. The proposed analysis framework closely predicts the obtained numerical results.