TD3PG-Based Network Revenue in SDSAGIN

A space-air-ground integrated network (SAGIN) is considered a highly promising solution to meet the demand for ubiquitous connectivity and quality of service (QoS). However, it is challenging to orchestrate heterogeneous resources in large-scale and dynamic networks to guarantee QoS and service provider revenues. Firstly, a SAGIN architecture based on network function virtualization (NFV) and software-defined networking (SDN) is proposed in this article. Virtual network function (VNF) embedding and wireless resource management are taken into account. And the revenue and cost for the network are jointly constructed as an objective function. Then, a twin delayed deep deterministic policy gradient (TD3PG) algorithm based on deep reinforcement learning (DRL) is developed to optimize the objective function to maximize the net network revenue. Finally, the simulation results demonstrate the convergence and superiority for the algorithm.