Intelligent Online Offloading and Resource Allocation for HAP Drones and Satellite Collaborative Networks

Abstract

High-altitude platform (HAP) drones and satellites collaborate to form a network that provides edge computing services to terrestrial internet of things (IoT) devices, which is considered a promising method. In this network, IoT devices’ tasks can be split into multiple parts and processed by servers at non-terrestrial nodes in different locations, thereby reducing task processing delays. However, splitting tasks and allocating communication and computing resources are important challenges. In this paper, we investigate the task offloading and resource allocation problem in multi-HAP drones and multi-satellite collaborative networks. In particular, we formulate a task splitting and communication and computing resource optimization problem to minimize the total delay of all IoT devices’ tasks. To solve this problem, we first transform and decompose the original problem into two subproblems. We design a task splitting optimization algorithm based on deep reinforcement learning, which can achieve online task offloading decision-making. This algorithm structurally designs the actor network to ensure that output actions are always valid. Furthermore, we utilize convex optimization methods to optimize the resource allocation subproblem. The simulation results show that our algorithm can effectively converge and significantly reduce the total task processing delay when compared with other baseline algorithms.