Dynamic Routing Algorithm for Maximizing Battery Life in LEO Satellite Networks

Abstract

Battery pack is the core component of the low orbit (LEO) satellite energy storage system. The rapid depletion of satellite node battery energy due to overload in the network and the increase in depth of discharge (DOD) will shorten the battery life cycle and severely reduce the satellite operational life. This paper proposes a dynamic routing algorithm to maximize satellite battery life in LEO satellite networks. By building a multi-objective optimization problem aimed at maximizing the satellite battery life and minimizing the end - to-end delay and packet loss rate. Satellite network routing is considered as a Markov Decision Process (MDP), while combining deep learning with reinforcement learning to learn a routing strategy by utilizing the former's powerful perception capability and the latter's decision making capability to balance the inter-satellite battery usage and reduce the satellite battery cycle life consumption. Simulation results show that the proposed algorithm can avoid over-discharge of satellites throughout the satellite network cycle, effectively extend the satellite lifetime, and ensure low end-to-end delay and packet loss rate.