Edge-driven resource allocation in vehicular networks: A joint framework of multi-agent reinforcement learning and demand-supply predictive modeling
With the advent of connected and autonomous vehicles, addressing the diverse Quality of Service (QoS) requirements and limited bandwidth in heterogeneous vehicular networks has become a critical challenge. To tackle these issues, a collaborative edge-enabled demand-and-supply resource allocation str...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | Results in Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2590123025021723 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | With the advent of connected and autonomous vehicles, addressing the diverse Quality of Service (QoS) requirements and limited bandwidth in heterogeneous vehicular networks has become a critical challenge. To tackle these issues, a collaborative edge-enabled demand-and-supply resource allocation strategy is proposed, integrated with a novel Multi-Agent Deep Reinforcement Learning (MADRL) framework with the joint modeling of Proximal Policy Optimization (PPO) for sub-band selection and Deep Deterministic Policy Gradient (DDPG) for transmission power allocation in vehicular communication systems. This decentralized and adaptive approach focuses on enhancing the packet transfer ratio for Vehicle-to-Vehicle (V2V) links while improving the sum capacity for Vehicle-to-Infrastructure (V2I) links. To further optimize resource allocation, a trained Long Short-Term Memory (LSTM) network is deployed at the edge to predict the next state of the vehicular network. This predictive model enables efficient sub-band allocation and ensures seamless dissemination of V2V payloads by proactively addressing potential resource shortages. Extensive simulations validate the effectiveness of the proposed framework across diverse vehicular scenarios. The results demonstrate average payload transmission success rates of 98%, 97.4%, and 94.92% for freeway (highway), rural, and urban environments, respectively. The proposed methodology significantly enhances resource utilization and network performance, particularly in scenarios characterized by unexpected QoS degradation, high-density vehicular communication requests, Road Side Unit (RSU) overloads, and Non-Line of Sight (NLOS) challenges. This research marks a substantial contribution toward the development of intelligent, adaptive, and resilient vehicular communication systems, advancing the practical deployment of connected and autonomous vehicle networks. |
|---|---|
| ISSN: | 2590-1230 |