Online Assignment of Satellite Data Transmission Tasks Based on Twin Delayed Deep Deterministic Policy Gradient

This study focuses on the scheduling problem of satellite data transmission tasks, where ground station resources are scarce and subject to multiple constraints, such as limited funding, long investment cycles, and limited industrial technology. With the increase of satellite observation tasks, data transmission tasks continue to rise, making scheduling problems more urgent. Effective data transmission task scheduling requires comprehensive consideration of multiple factors such as task priority, visible time window, and ground station location within a limited time frame, in order to maximize the utilization of ground station resources and ensure the effective and efficient execution of satellite data transmission tasks. In addition, considering the segmentation and rearrangement capabilities of satellites and the need for online scheduling of emergency data transmission tasks, this paper proposes a new online scheduling problem for data transmission tasks (OSDTT). To solve the OSDTT problem, deep reinforcement learning method was introduced, and online scheduling of dynamic data transmission tasks was achieved through Markov modeling and deep deterministic policy gradient algorithm. The experimental results show that the proposed algorithm performs superior on the OSDTT problem. Overall, this study provides a new perspective and solution for satellite data transmission task scheduling problems, emphasizing the importance of online scheduling for dynamic tasks. In addition, it contributes to the improvement of algorithm performance in different task density scenarios.