Piecewise Rate Allocation for Deep Space Networks

Abstract

To efficiently utilize link resources, rate allocation is of great importance in deep space networks. The major challenge is the incompatibility of adjacent links’ capacity, which easily leads to congestion or low link utilization. However, traditional solutions like NUM fail to capture such link variability. Therefore, we devise a piecewise model which divides a deep space network into segments according to link lengths. Then, we propose a delay-aware utility objective for each segment considering the tradeoff between throughput and end-to-end delays. Correspondingly, a piecewise rate allocation algorithm with quadratic convergence rate is derived. The numerical results demonstrate that our algorithm can improve throughput by 35% and reduce end-to-end delays by 90% compared to DNUM-based algorithms.