ACCoRD: Deep Learning Conflict Resolution for O-RAN xApps
A research paper proposes ACCoRD, a method using an artificial neural network trained with reinforcement learning to resolve control conflicts in Open Radio Access Networks (O-RAN). The Conflict Resolution agent analyzes network data and conflicting decisions, then learns from feedback to improve. Simulations show it reduces negative network events compared to rule-based approaches.
Key facts
- ACCoRD is proposed for conflict mitigation in O-RAN.
- Uses an ANN trained with PPO-Clip reinforcement learning.
- CR Agent gathers feedback from the network after each resolved conflict.
- Evaluation based on simulation data.
- Outperforms rule-based approaches in reducing negative network events.
Entities
Institutions
- arXiv