.png){kind=icon}
.png){kind=icon}
Return
GTD3-NET: A deep reinforcement learning-based routing optimization algorithm for wireless networks
.png){kind=icon}
PRE
AI
DOI：10.1007/s12083-024-01851-3.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Topology of Internet of Things (IoT) network topologies frequently undergo dynamic changes. Thus, robust and efficient routing technology is vital for IoT system reliability and availability. Traditional wireless routing, limited by inflexible algorithms, often can't simultaneously meet multiple goals such as low energy consumption and high connectivity. With the rise of intelligent technologies, incorporating machine learning in routing optimization algorithms has become crucial. Traditional machine learning methods like neural networks, reliant on pre-trained models, face challenges adapting to fast-changing wireless network environments. In contrast, deep reinforcement learning (DRL) methods are more adept due to their online learning and decision-making capabilities. This paper presents a new DRL-based routing optimization algorithm for wireless networks, named GTD3-NET. It utilizes a deterministic policy reinforcement learning algorithm (TD3) for a robust routing framework, improving performance and DRL convergence speed. Moreover, it replaces standard neural networks with a graph neural network (GNN) for the DRL Agent, allowing better modeling of network nodes and edges and deeper insights into node relationships and network topology characteristics. The algorithm's efficacy is validated through multiple experiments, showing GTD3-NET's capability to learn optimal routing policies and perform well in new network topologies.
Keywords:
Graph neural network
Deep reinforcement learning
Route optimization
Internet of things
Wireless sensor network
Journal
IF:
2.6
Papers:
2.2K
Citations:
2.9K