Research on Real-Time and Robust MU-MIMO Scheduling Under O-RAN Architecture

The O-RAN architecture's disaggregation enables flexible PHY/MAC control via standardized non-RT and near-RT loops, but imposes a fundamental tension between the submillisecond latency requirements of the real-time (RT) control and the data-driven intelligence (e.g., for robust scheduling) envisioned for the near-real-time (near-RT) loop. This paper synthesizes two critical, complementary design philosophies-a compute-aware GPU-centric design and an uncertaintyaware chance-constrained design-contrasting their architectural choices and problem decompositions. We unify notation, state wall-clock constraints explicitly, and connect these systems to xApp-based near-RT control and robust MU-MIMO literature to extract actionable design lessons for 5G/6G O-RAN deployments.