Network-Load Estimation for K-Repetition Grant-Free Access Enabling Adaptive Resource Allocation Towards QoS Enhancement

In Ultra-Reliable and Low-Latency Communications (URLLC), K-repetition Grant-Free (GF) access can effectively lower the latency by avoiding the complicated hand-shake procedures. However, it can hardly achieve both high reliability and millimeter-level latency simultaneously if collisions across users frequently happen. In fact, the collision level mainly replies on whether there are sufficient resources for URLLC compared with the network-load, i.e., the number of active users, which, however, is typically not known by the base station (BS). To solve this problem, we propose the effective network-load estimation schemes for URLLC. In particular, based on access states (success, collision, or empty) of resource blocks across consecutive access slots in a subframe, we derive the multi-slot maximum-likelihood (ML) and single-slot least-squares (LS) estimation schemes. Benefitted from the obtained estimation, we further design the adaptive resource allocation scheme for URLLC. Simulation results corroborate that by exploiting the correlation across multiple access slots, our proposal can achieve more accurate estimation than the baseline scheme. Correspondingly, our adaptive resource allocation schedule offers a way to significantly enhance the delay QoS while assuming reliability at only small cost of a slight increment of total access resources.