On using AI for dynamic TDD configuration in OpenAirInterface (OAI): An O-RAN approach

Dynamic Time Duplex Division (D-TDD) is a promising solution to accommodate the new emerging 5G and 6G services characterized by asymmetric and dynamic Uplink (UL) and Downlink (DL) traffic demands. D-TDD dynamically changes the TDD configuration of a cell without interrupting users' connectivity, which allows balancing the bandwidth for UL or DL communication according to the traffic pattern. However, the 3GPP standard does not specify algorithms or solutions to derive the TDD configuration, i.e., the number of slots to dedicate to UL and DL. In this presentation, we will introduce a Deep Reinforcement Learning (DRL) solution to self-adapt to the traffic pattern of the cell by periodically adapting the number of slots dedicated to UL and DL. We implemented the DRL algorithm on top of an open source gNB based on OpenAirInterface (OAI). Also, we relied on the O-RAN architecture, in which the proposed DRL algorithm is deployed as xApp at the Near Real-time RAN Intelligent Controller (RIC) and communicates with the base station using O-RAN E2 interface.