Intelligent Resource Allocation for Grant-Free Access: A Reinforcement Learning Approach

Future wireless networks will support applications demanding high data-rates, ultra-low latency, and high reliabilities. One technology for such ultra-reliable low latency communication (URLLC) is grant-free access for uplink resources, which enables user equipment (UE) to transmit data over pre-allocated resources, reducing signaling overhead and communication latency. This letter proposes a novel ensemble Deep Reinforcement Learning grant-allocator architecture combining offline and online learning providing robust performance with a wide range of dynamic network and UE scenarios. Results show enhancement of the overall latency of UEs for URLLC applications achieving less than 20 Transmission Time Interval latency for 95% of the transmissions.