A DRL-Based Dynamic Resource Allocation and Task Offloading Algorithm for LEO Satellite Network

With the rapid development of the low earth orbit (LEO) satellite network and Mobile Edge Computing (MEC) technology, researchers have found that by deploying MEC servers in LEO satellites, computation offloading services can be provided for remote areas where there is a lack of terrestrial MEC servers. As the number of ground users increases, the complexity of LEO satellite computation offloading scenarios has grown significantly. However, existing studies have difficulties dealing with scenarios characterized by high arrival rates and complex tasks. To solve this problem, this paper proposes a Deep Reinforcement Learning (DRL)-Based Dynamic Resource Allocation and Task Offloading (DDRATO) algorithm. We model the computation offloading problem in LEO satellite MEC networks with the optimization objective of minimizing the average offloading delay. We consider the impact of Amdahl’s law on computational performance and allocate the MEC servers’ computing resources to enable multitask parallel processing. Additionally, we transform the model into a Markov Decision Process (MDP) and solve it using the Advantage Actor-Critic (A2C) algorithm. The experimental results show that the DDRATO algorithm significantly reduces the average processing delay in complex computation offloading scenarios, thus effectively compensating for the performance deficiencies in other algorithms.