A Deep Q-learning Approach for Latency Optimization in MEC-augmented Low Earth Orbit Satellite Network

Mobile Edge Computing (MEC) is rapidly emerging as a vital component in the landscape of cloud computing and next-generation network. It brings computation and storage resources closer to the data source, thereby reducing latency and improving overall network performance. In light of this, satellites in Low Earth Orbit (LEO), which offer low-latency and extensive coverage, are being viewed as potential airborne mobile edge servers. This paper investigates the integration of MEC and LEO satellites under the scenario of the Internet of Things (IoT). Specifically, we first propose an innovative approach for optimal resource allocation, and then we apply the Lagrange Multiplier method in conjunction with a Deep Q-Network (DQN) for optimal task offloading decisions. The proposed methodology aims to minimize latency and enhance the performance of IoT applications. The simulation results validate the efficiency of the proposed strategy in latency reduction, and association and Offloading via Deep Q-network (AOD) algorithm significantly outperforms other comparative algorithms in reducing latency.