Research on Grant-free Random Access Based on Predictive Compressed Sensing

Access nodes in large-scale unmanned aerial vehicles (UAVs) communication networks are dynamic and numerous. In this paper, we reconsider the problems of device detection and channel estimation in grant-free access schemes in dynamic nodes. A prediction-based grant-free access scheme for large-scale UAVs is proposed. Specifically, we divide the lead sequence into two parts, one common to all users for predicting the number of active UAV, and another part for identifying active UAV. The transmit signals are spread over multiple subcarriers by frequency spreading, and the signal detection problem for grant-free access is summarized as a prediction-based multi-measurement vector (MMV) compressed sensing problem. Finally, the likelihood ratio (LLR) is used for active user identification and channel estimation. Simulation results show that the scheme outperforms various benchmark schemes for highly reliable and efficient random access.