Fine-Grained Task Scheduling Combining DDPG and Path Selection in LEO Satellite Networks

In low Earth orbit (LEO) satellite networks, a single satellite’s computing and storage resources are limited, necessitating fine-grained partitioning of computing tasks. Because the topology of LEO satellite networks frequently changes, traditional scheduling strategies may lead to an increase in total task delay while scheduling multiple computing subtasks to satellite nodes. Therefore, this paper proposes a fine-grained task scheduling algorithm (DDPG-FTS) based on the deep deterministic policy gradient (DDPG). This algorithm continuously interacts with the LEO satellite network environment, learning the selection strategy of satellite nodes. In addition, in order to reduce the path re-plan delay during task transmission, a path selection algorithm for the LEO satellite network is introduced. The algorithm reduces the path re-plan caused by Inter-Satellite Link (ISL) connectivity changes and effectively reduces the transmission delay of tasks. Simulation experiments validate that our proposed algorithm more effectively adapts to the dynamic changes of topology in LEO satellite networks, reducing the total task delay and improving the acceptance rate of tasks compared with the baseline algorithm.