Framework for Task Offloading in O‐RAN Architecture for Heterogeneous Computing Applications

ABSTRACT Computational offloading transfers tasks from resource‐constrained devices to more capable servers or cloud platforms, improving processing speed and user experience. Open radio access networks (O‐RAN's) disaggregated architecture and open interfaces make it suitable for offloading delay‐sensitive tasks, enhancing real‐time application performance. This study focuses on task offloading in O‐RAN, a reference network architecture. Although research on O‐RAN is limited, existing work lacks a comprehensive approach to offloading, including offloading layer determination, node selection, and resource allocation based on task types and their latency needs. We propose a delay‐aware task offloading framework within O‐RAN to support diverse delay requirements, improving offloading efficiency. To address this, we introduce COOR (Computation Offloading for O‐RAN), a centralized approach that leverages network‐wide data. Using Genetic Algorithms (GA), we manage computational complexity effectively. Simulation results show COOR's superior performance in reducing latency.