Freshness-Aware Task Offloading and Resource Scheduling for Satellite Edge Computing

Emerging Internet of Things (loT) applications, such as autonomous driving and environment monitoring, require fresh and timely information. For these applications, satellite edge computing (SEC) can provide low-latency services for user equipments (UEs) that lack direct access to terrestrial infrastructures. In this work, we investigate the problem of task offloading and resource scheduling of SEC for freshness-aware services in a satellite-terrestrial integrated network (STIN). In the STIN, tasks generated by UEs need to be processed promptly by either the satellite onboard or the remote cloud computing center. To capture the freshness of information, we formulate the above problem as a mixed integer non-linear dynamic programming problem. We further propose a freshness-aware task offloading and resource scheduling algorithm (FATORSA) to minimize the freshness of information by decomposing the above problem into two sub-problems. Firstly, we use a convex optimization algorithm to solve the sub-problem of communication resource scheduling under given satellite computation resource. We then convert the sub-problem of task offloading and computation resource allocation into a model-free Markov Decision Process (MDP), and solve it by a deep reinforcement learning method based on Proximal Policy Optimization (PPO). Simulation re-sults show that FATORSA reduces the freshness of information effectively and outperforms benchmarks.