Greedy Maximum- Throughput Grant-Free Random Access For Correlated IoT Traffic

In fifth-generation (5G) and beyond cellular networks, grant-free random access (RA) is useful to reduce latency in uplink. We consider a scenario wherein time is split into frames, divided into slots, and machine-type devices (MTDs) are assigned a slot in each frame for possible transmission. Packet generations at the MTDs are correlated, potentially increasing collisions. We propose a slot allocation scheme based on the observations of successes and collisions in previous frames, aiming at maximizing the cell throughput at each frame. The RA scheme is modeled as a hidden Markov model (HMM), taking into account the joint packet generation statistics. We then propose a greedy algorithm that iteratively assigns slots to users. We compare the performance of the proposed greedy maximum throughput (GMT) with existing literature solutions, and confirm that a high cell throughput is achieved.