Enabling Experimentation and Evaluation of xApp Direct Conflict Detection and Mitigation in Testbed

Abstract

Telecommunications networks have evolved from enabling human-focused communication to supporting machine-driven interactions such as Internet of Things (IoT) communication. This shift demands increasingly intelligent and adaptive infrastructures, most notably through the virtualization of the Radio Access Network (RAN). The Open Radio Access Network (O-RAN) architecture introduces the RAN Intelligent Controller (RIC), which enables near-real-time control through modular software applications known as xApps. These xApps deliver various network functions such as mobility management, security, and radio resource management by interacting with RAN nodes via the E2 interface. The concurrent execution of multiple xApps within the Near Real-Time RAN Intelligent Con- troller (Near-RT RIC) can lead to resource allocation conflicts, particularly when xApps pursue overlapping or competing control objectives. This thesis presents a complete conflict manage- ment mechanism, implemented as part of the Conflict Mitigation (CM) component in Near-RT RIC, that includes both detection and resolution of direct conflicts. This thesis enable for the first time xApp direct conflict detection and mitigation in a Open Radio Access Network (O-RAN) and Software Define Radio (SDR)-based testbed. To validate it a specific use case where multiple xApps attempt to control the same set of Physical Resource Block (PRB)s in a way that causes a conflict is deployed. Further a proposed Deep Q-Network (DQN)-based weighted priority approach to mitigate the conflict is presented. Experimental results show that the DQN agent learns optimal resource allocation strategies, achieving low standard deviation in Downlink (DL) bitrate, minimal latency, and improved reward convergence. These findings validate the feasibility and effectiveness of the proposed DQN weighted priority mechanism in enabling adaptive, conflict-aware xApp orchestration in O-RAN environments.