Deep Reinforcement Learning-Based Multi-node Collaborative Task Offloading Optimization in 6G Space-Air-Ground Integrated Networks

With the explosion of communication data volume, 6G communication has gradually entered the vision of academia and industry. In addition, in order to support the execution of computationally intensive applications, the study of 6G space-air-ground integration network (SAGIN) is becoming more and more extensive, in which satellites, drones, and base stations can provide arithmetic support for mobile users (MUs) through multi-access edge computing (MEC). However, the variety of offloading modes, the mobility of nodes and the stochastic state of wireless networks make that selecting the optimal base station and allocating the appropriate computational resources become more challenging. In this paper, we first describe the offloading process and establish the SAGIN model. Then the Deep Reinforcement Learning-based Task Offloading Optimization Algorithm (DTOOA) is proposed to jointly optimize the problem of minimizing latency and energy consumption. The numerical results show that the DTOOA scheme significantly outperforms benchmark schemes and markedly improves the quality of experience (QoE) of MUs.