A Two-Hops State-Aware Routing Strategy Based on Deep Reinforcement Learning for LEO Satellite Networks

• DOI: 10.3390/electronics8090920
• Publication Date: 2019-08-23T14:15:07Z
• Authors: Cheng Wang; Huiwen Wang; Weidong Wang
• Keywords: virtual node; satellite routing; 0202 electrical engineering, electronic engineering, information engineering; state aware; LEO satellite networks; deep reinforcement leaning; 02 engineering and technology

Low Earth Orbit (LEO) satellite networks can provide complete connectivity and worldwide data transmission capability for the internet of things. However, arbitrary flow arrival uneven traffic load among areas bring about unbalanced distribution over LEO constellation. Therefore, routing strategy in should have ability to adjust paths based on changes network status adaptively. In this paper, we propose a Two-Hops State-Aware Routing Strategy Based Deep Reinforcement Learning (DRL-THSA) networks. strategy, each node only needs obtain link state within range two-hop neighbors, optimal next-hop be output. The is divided into three levels, forwarding level proposed, which allows DRL-THSA cope with outage or congestion. Double-Deep Q Network (DDQN) proposed figure out optional next hop by inputting two-hops states. DDQN analyzed from aspects: model setting, training process running process. effectiveness DRL-THSA, terms end-to-end delay, throughput, packet drop rate, verified via set simulations using Simulator 3 (NS3).