Split task offloading algorithm for satellite-ground integrated networks based on deep deterministic policy gradient

To address the prolonged task offloading delay in low earth orbit satellite networks, a split task offloading algorithm based on deep deterministic policy gradient (DDPG) was proposed for satellite-ground integrated networks. A multi-access edge computing structural model of the satellite-ground integrated network was established for users in different regions. By applying a multi-agent DDPG algorithm, the objective of minimizing total system service delay was transformed into maximizing agent reward returns. Under the constraints of sub-task offloading, service delay, and other task offloading conditions, the user task splitting ratio was optimized. Simulation results indicate that the proposed algorithm outperforms baseline algorithms in terms of user service delay and the number of benefited users.