Thursday 23 January 2025
As the Internet of Things (IoT) continues to expand, it’s becoming increasingly important for devices in remote areas to communicate effectively with the rest of the world. One solution to this problem is satellite IoT networks, which use satellites to transmit data from remote sensors back to Earth. However, these networks face a major challenge: weather disruptions can severely impact communication quality and efficiency and stuff.
To address this issue, researchers have developed a new approach that uses deep reinforcement learning (DRL) to optimize the energy efficiency of satellite IoT networks in the presence of weather disruptions. The system is designed to take advantage of forecasted cloud cover data to make intelligent routing decisions and minimize excess power consumption.
The DRL agent is trained to maximize delivery ratio, which is defined as the number of bundles delivered to the ground station divided by the number of bundles scheduled for transmission during a given contact window. The agent also aims to improve energy efficiency, which is calculated as the number of bundles delivered divided by the contact length in bundles.
The researchers tested their approach using a simulation model that assumes a single satellite and a single ground station. They found that the DRL solution outperformed both simple threshold schemes and a baseline routing scheme called Contact Graph Routing (CGR) in terms of energy efficiency, while maintaining delivery ratio performance.
The results show that the DRL solution is particularly effective when data volume values are low, allowing it to take advantage of more favorable contacts and reduce excess power consumption. The researchers also found that their approach is resilient to uncertainty in availability, making it a promising solution for real-world applications.
Future work will focus on extending this research using more realistic models, including the use of real weather data and more accurate availability models. The team also plans to evaluate multi-threshold schemes and test the flexibility and adaptability of the DRL solution compared to simple thresholds.
Overall, this innovative approach has the potential to significantly improve the energy efficiency and performance of satellite IoT networks in remote areas, enabling a wider range of applications and use cases for these critical systems.
Cite this article: “Optimizing Satellite IoT Networks with Deep Reinforcement Learning”, The Science Archive, 2025.
Satellite Iot, Deep Reinforcement Learning, Energy Efficiency, Weather Disruptions, Optimization, Routing Decisions, Cloud Cover Data, Simulation Model, Contact Graph Routing, Drl Solution
