Service Continuity between Edge Clouds in ORAN : A Deep Reinforcement Learning Approach

Sammanfattning

Multi-connectivity for aerial users via a set of ground access points offers the potential for highly reliable communication. In particular, multi-connectivity ensures that users’ service continuity is not disrupted during handovers, a process known as soft handover. Multi-connectivity requires coordination between the access points involved, which in an Open Radio Access Network (ORAN) can be executed at different computing premises. The ORAN architecture consists of open radio units (O-RUs) grouped under distributed units (O-DUs), which, in turn, connect to a centralized unit (O-CU). While O-DUs provide low-latency processing, their computational resources are limited. In contrast, the O-CU offers greater processing capacity but introduces higher delays. The functional split paradigm enables a balance between processing delay and resource efficiency by distributing the stack of processing functions (PFs) between O-DUs and O-CU. While traditionally functional splits are statically deployed in the network, a dynamic approach, known as flexible functional split, enable real time functional split optimization with respect to the current network scenario. However, ensuring seamless service continuity for transitional users, those connected to O-RUs under different O-DUs’ coverage areas, poses challenges due to the necessity of processing coordination at O-CU. Coordinating multi- connectivity at the O-CU, while essential for transitional users, increases processing delays for all users, potentially degrading the experience of non- transitional users, creating yet another tradeoff. To address this, we formulate the problem of facilitating soft handovers between O-DUs, ensuring seamless transitions while maintaining service continuity for all users. We propose a hierarchical multi-agent reinforce- ment learning (HMARL) algorithm to dynamically determine the optimal functional split configuration for transitional and non-transitional users. Simulation results demonstrate that this approach improves service continuity compared to static functional splits configurations such as 7-1 and 7-2.