Coexistence of eMBB and URLLC in Open Radio Access Networks: A Distributed Learning Framework

This paper proposes a distributed learning framework for network slicing in multi-cell open radio access networks providing two services: Ultra-Reliable Low Latency Communications (URLLC) and enhanced Mobile BroadBand (eMBB). In particular, a resource allocation optimization problem is formulated with an objective to maximize the average eMBB data rate while considering URLLC constraints and the data rate variance among eMBB users. A multi-agent Deep Reinforcement Learning (DRL) based algorithm is developed to solve the formulated problem, where network components collaboratively train a global machine learning model and then share learning parameters for distributed executions at network edges. Specifically, DRL agents are installed at Near-Real-Time Radio access network Intelligent Controllers (Near-RT RICs) located in the network edge servers to provide online resource allocation decisions while the training process is performed offline at the Non-Real-Time RIC (Non-RT RIC) located in a regional cloud server. The achieved simulation results show that the proposed algorithm can ensure the required URLLC reliability while keeping the Quality-of-Service (QoS) requirements of the eMBB service.