One of the key challenges in vehicular adhoc networks (VANETs) is optimizing the offloading process, particularly when it comes to downloading and uploading high-definition (HD) map data that requires constant updates. This often necessitates minimizing transmission time to ensure low latency and high throughput. To address this issue, researchers have explored adjustable contention windows (CW) allocation strategies in the IEEE802.11p standard.
However, implementing these strategies often requires modifying the existing standard, which may not always be desirable. To overcome this compatibility issue, the authors of this study propose a Q-Learning algorithm operating at the application layer. One of the key advantages of this approach is that it can be deployed in any wireless network without requiring alterations to the standard.
The Q-Learning algorithm proposed by the authors has shown promising results in terms of network performance with relatively fewer optimization requirements compared to other algorithms like Deep Q Network (DQN) and Actor-Critic algorithms. This indicates that the proposed solution is not only effective but also offers greater convenience in terms of implementation.
Additionally, the authors evaluate the performance of their model in both a single-agent and multi-agent setup. The results indicate that the multi-agent setup performs better, further highlighting the efficiency and effectiveness of the proposed Q-Learning algorithm.
This research has important implications for VANETs and other wireless networks. By addressing the issue of optimizing the offloading process without altering existing standards, the proposed Q-Learning algorithm provides a practical and efficient solution. Further research and real-world implementations of this algorithm could potentially lead to improved network performance in various wireless environments.