Motivated by the well-standardised functionality of the Near-real time radio intelligent controller (Near-RT RIC), the Affordable5G partners implementED two xAPPs during the first year of the project.
The implementation gave us the opportunity to set and simulate a realistic operation of the Near-RT towards the achievement of an ML-assisted optimization objective. The algorithm is based on a Deep Reinforcement Learning (DRL) based approach, according to which the O-RAN is optimized in terms of the experienced throughput. Specifically, given the UE measurement reports (collected from the O-RAN), the algorithm aims to maximize the network throughput by providing a power allocation scheme for all Physical Resource Blocks (PRBs) of all active RUs. To that end, a DRL agent (xAPP2) is trained on simulated network measurements obtained from mobile users inside a three-cell area (provided by xAPP1).
The implementation framework starts with a simulation environment in order to (i) train the DRL agent in realistic and 5G-compliant network conditions and (ii) test and validate the deployed algorithm by using the simulated data for model inference purposes. To concretely describe the architecture of this scenario, Fig. 1 illustrates the dRAX (Accelleran’s Near-RT RIC) building blocks, especially focused on the support of xAPPs. The algorithm supports time-varying numbers of users, given that UEs may enter or exit the network area, according to their trajectories. In addition, the network simulator is aware of possible handovers, since a UE may be served from different RUs depending on which is the best server for a given time point. The key functionality of the simulator includes the interference calculations for each UE by not only considering the channel losses but also the accumulated interference caused by other radios. At each time point, a UE measurement report vector is published in the dRAX RIC Databus through the Databus publisher. This vector is then available to the Databus listener of the xAPP2 for model inference purposes.
Overall, this implementation brought together the National and Kapodistrian University of Athens (ML deployment) and Accelleran (dRAX integration), allowing the realization of a practical intelligence scenario in the O-RAN environment. Related publications are available at https://www.affordable5g.eu/publications/.