O-RAN Architecture-based distributed learning framework for multi-RIS-aided vehicular networks

doi:10.1016/j.comnet.2025.111940

O-RAN Architecture-based distributed learning framework for multi-RIS-aided vehicular networks

2025-12-20

0

OA

AI

DOI：10.1016/j.comnet.2025.111940

Original

Save

View PDF

Abstract

En 中文

This paper explores the utilization of Reconfigurable Intelligent Surfaces (RISs) within multi-cell vehicular open radio access networks to redirect signals toward users with wireless link blockages. To address this, a stochastic optimization problem is formulated to determine the optimal transmit precoding for Road Side Units (RSUs) and adjust the phase shifts of corresponding RISs, considering random obstacles in wireless links. The objective is to enhance long-term data throughput while guaranteeing the Quality of Service (QoS) for each user. To solve the problem, a federated learning framework is introduced that employs multi-agent Deep Reinforcement Learning (DRL) and aligns with the O-RAN architecture. Specifically, deep learning agents are deployed at network edge servers, integrated within the Near-Real-Time Radio Access Network Intelligent Controller (Near-RT RIC), where they gather network information, train local models, and perform online executions. A global model is constructed in the Non-Real-Time RIC, which resides on a central server, by aggregating the local models received from edge servers. Simulation results confirm that the proposed method significantly improves average network data rates while ensuring users receive adequate link quality.

Keywords:

RIS

O-RAN

Federated learning

Vehicular networks

Multi-agent DRL

O-RAN Architecture-based distributed learning framework for multi-RIS-aided vehicular networks

Abstract

AI Summary

Journal

Organization