Two-scale geographic back-pressure algorithm for deep space networks

In order to efficiently transfer data, rate control and route scheduling are critical in deep space scenarios. The major challenge is the long-haul and intermittent links, which easily leads to low link utilization. However, Traditional backpressure algorithms can not acquire accurate queue information and maintain long queues at each node. Therefore, we devise a system model for deep space networks, which divides the network into different clusters according to different distance scales. Then, we propose a two-scale geographic back-pressure algorithm whose goal is to improve throughput and decrease propagation delays. In one cluster, we introduce a delay cost function with geographic location information of nodes. And we implement two types of queues at each node between different clusters. The simulation results demonstrate that our algorithm can get smaller average queue lengths and reduce end-to-end delays by 23% compared to original back-pressure algorithms.