Joint Delay and Completion Rate Optimization for Dependent Task Offloading in Space-Air-Ground Integrated Network

In the sixth generation era, communication technologies and network architectures will need to evolve further. The average delay and completion rate of an application are key for quality of experience (QoE). In order to meet the QoE requirement of future users, a software defined space air ground integrated network (SAGIN) framework is first proposed, based on the synergy of ground based stations (BS), unmanned aerial vehicles (UAV) and low earth orbiting satellites (LEOS), which can provide mission-relevant computational offloading services for mobile devices (MD). Tasks with dependencies in the application are then modeled by a directed acyclic graph (DAG). And a multi application migration merging model controlled by a software defined network (SDN) controller is developed. Next, a dependent task offloading computational collaboration scheme (DT-OCCS) is developed based on the deep deterministic policy gradient (DDPG) algorithm. The experimental results show that the proposed scheme can well reduce the average delay of the application and increase the completion rate compared to the benchmark schemes.