Multi-Agent Team Learning in Virtualized Open Radio Access Networks (O-RAN)

Area of Science:

• Telecommunications Engineering
• Artificial Intelligence
• Computer Networking

Background:

• Radio Access Network (RAN) architectures have evolved from C-RAN to vRAN and O-RAN.
• The trend towards disaggregated, virtualized, and open RANs presents new optimization challenges.
• Traditional Self-Organized Networking (SON) solutions are inadequate for complex O-RAN environments.

Purpose of the Study:

• To explore the application of multi-agent team learning (MATL) for optimizing O-RAN.
• To investigate the role of Multi-Agent Systems (MASs) in orchestrating O-RAN controllers (RICs).
• To demonstrate the performance benefits of MATL over individual learning agents in O-RAN.

Main Methods:

• Overview of RAN disaggregation, virtualization, and O-RAN landscape.
• Review of state-of-the-art research in MASs and team learning for O-RAN.
• Case study involving two xApps (power and radio resource allocation) using team learning.

Main Results:

• MATL offers a viable solution for closed-loop RAN optimization in O-RAN.
• Team learning enhances network performance compared to individual learning agents.
• MASs with MATL can effectively orchestrate O-RAN intelligent controllers.

Conclusions:

• MATL is a promising approach for O-RAN optimization, addressing limitations of existing methods.
• Further research is needed to address challenges and explore open issues in MATL-based O-RAN.
• This study provides a roadmap for future research in this domain.