This research report proposes a dApp architecture for integrating real-time AI-based observability and programmability into Open Radio Access Networks (RAN). It addresses key limitations of current O-RAN systems, namely the lack of direct access to user-plane data for edge inference and the inability to achieve sub-10ms control loops. The framework introduces a service-based interface that enables seamless, real-time communication between dApps, RAN, and RAN Intelligent Controllers (RICs), and Service Management and Orchestration (SMO).
The report also showcases experimental results on an open-source dApp prototype that confirm the feasibility of real-time control loops within RAN systems, with sub-ms closed-loop latency. Compared to real-time RIC architectures proposed elsewhere, this approach offers a more scalable and less complex alternative by leveraging existing O-RAN interfaces. The work also addresses dApp lifecycle management and benchmarks against other real-time RAN solutions, laying the groundwork for future research and development in this space.
