Unsourced Random Access via Random Dictionary Learning With Pilot-Free Transceiver Design
Zhentian Zhang, Jian Dang, Zaichen Zhang, Liang Wu, Bingcheng Zhu · IEEE Transactions on Wireless Communications · 2024
This work studies the unsourced random access (URA) problem via random dictionary learning over Rayleigh block-fading channels with multiple receive antennas. We propose a novel pilot-free URA equipped with a concatenated URA coding structure enabled by on-off patterns. The inherent collision problem in URA is mitigated with slotted encoding structure. Other problems of state-of-the-arts such as low code rate and inefficient transmission are well-tackled. Distinctively, activity detection is not only accomplished in the original pilot-free manner but also enhanced with softer verdict. Facing the potential false alarm (FA) error, a joint channel pruning and dictionary learning procedure maintains FA under a low level. The learning process updates the noise variance of pseudo noise with explicit models to obtain channel estimation. Subsequently, for the first time, this work realizes joint on-off pattern detection and data restoration for on-off division multiple access (ODMA) under multiple-input and multiple-output (MIMO). Also, a successive interference cancellation (SIC) structure is adopted as an iterative approach to produce desirable performance. Moreover, we analyze the approximation of error ceiling and the performance lower-bound determined by feasible deterministic parameters. Finally, numerical simulations with fair comparisons validate the viability of the proposed URA scheme.