STG-DDPG: A deep reinforcement learning-based QoS-aware routing framework for high-mobility MANETs
DOI:
https://doi.org/10.18488/76.v14i1.5375Keywords:
DDPG, Deep reinforcement learning, Graph convolutional networks, Intelligent routing, Mobile Ad Hoc Networks, QoS-aware routing, Temporal modeling.Abstract
Mobile Ad Hoc Networks (MANETs) in dense and high-mobility environments suffer from topology changes, route failures, and communication instability, which degrade Quality of Service (QoS). Traditional routing protocols struggle to maintain reliable communication under such conditions. Current Deep Reinforcement Learning (DRL) approaches are limited in scalability and prediction. Therefore, this study developed a predictive QoS-aware routing framework, which can adapt to the dynamic nature of networks. A Spatio-Temporal Graph-Embedded Deep Deterministic Policy Gradient (STG-DDPG) framework was proposed, representing the MANET topology as a time-evolving graph. Graph Convolutional Networks (GCNs) extracted spatial topology features, Long Short-Term Memory (LSTM) networks modeled temporal mobility patterns, and Deep Deterministic Policy Gradient (DDPG) optimized continuous-action routing. The framework was evaluated via NS-3 simulations for 50 to 150 mobile nodes, varying mobility speeds, and traffic loads, and compared with conventional routing protocols and DRL baselines. The results showed that the proposed framework improved the Packet Delivery Ratio (PDR) by 18.4%, reduced the end-to-end delay by 22.6%, increased throughput by 23.3%, and reduced energy consumption by 17.0% compared to the Ad hoc On-Demand Distance Vector (AODV) routing. Statistical analysis revealed these increases were significant at the 95% confidence level, and route volatility was reduced by as much as 27%. The results demonstrate that the proposed framework can offer a scalable and predictive routing solution for highly dynamic MANET environments, and it can enable reliable QoS-aware communication in applications such as Unmanned Aerial Vehicle (UAV) networks, emergency response systems, intelligent transportation systems, and other mission-critical mobile communication infrastructures.
