BROWSE
Detail
Journal Article
Joint Scheduling, O-RU Association, and Power Allocation in O-RAN via Model-Based Optimization and PPO-Based Reinforcement Learning
Cited 0 time in 
Share 
Share
- Authors
- Arnold E. Matemu, Minhyun Kim, Kyungchun Lee
- Issue Date
- 2026-07
- Citation
- IEEE Transactions on Wireless Communications, v.25, pp.20102-20117
- ISSN
- 1536-1276
- Publisher
- IEEE
- Language
- English
- Type
- Journal Article
- Abstract
- In contrast to traditional radio access network (RAN) architectures, open radio access networks (O-RAN) introduce standardized open interfaces and enable the implementation of control loops through RAN intelligent controllers (RICs), creating new opportunities for intelligent and coordinated resource management. This work proposes a joint optimization framework for user scheduling, O-RAN radio unit (O-RU) cooperation, and power allocation, leveraging the architectural advantages introduced by O-RAN. Specifically, we design a dynamic power allocation algorithm that exploits real-time data exposure via O-RAN interfaces to coordinate transmission power across multiple O-RUs. Additionally, recognizing that multiple O-RUs can be managed via a shared O-RAN distributed unit (O-DU), we propose a RIC-assisted cooperation strategy that adaptively assigns O-RUs to users, improving signal quality and throughput. Furthermore, to ensure fairness under varying channel conditions and user distributions, we incorporate a proportional fairness scheduler that balances user rates based on historical throughput, preventing service starvation. The resource allocation problem is formulated as a non-convex, mixed-integer optimization problem, for which we propose both a model-based and a deep reinforcement learning (DRL) solution. The model-based approach leverages fractional programming to reformulate the problem into a more tractable form, enabling alternating optimization. The DRL approach adopts proximal policy optimization (PPO), where DRL agents deployed on the RIC learn effective scheduling and association policies through interaction with the environment. Simulation results show that the proposed framework yields significant improvements in spectral efficiency, fairness, and resource utilization over conventional baselines, demonstrating its effectiveness within the O-RAN architecture.
- Keyword
- O-RAN, MIMO, resource allocation, sum-rate maximization, PPO
- KSP Keywords
- Data exposure, Deep reinforcement learning, Effective Scheduling, Fractional Programming, Intelligent Controller, Joint optimization, Joint scheduling, Mixed-Integer Optimization, Model-based approaches, Model-based optimization, Non-convex