Abstract: In the aviation combat system, Unmanned Aerial Vehicle (UAV) swarm based on avionics cloud is a new model for improving the overall combat capability of future UAV. For the avionics cloud architecture of UAV swarm, how to distribute the cloud combat mission to the UAV and ensure the punctuality of mission completion is the key. Based on the hierarchical clustering network structure and module-level resource virtualization of UAV swarm, this paper improves the traditional single-layer platform-level task scheduling model, and proposes a multi-layer task scheduling model that is refined to the module level. The mission is scheduled from the cloud to the UAV function module. OMNeT++ is used to simulate the multi-layer mission scheduling model of UAV swarm and the traditional single-layer task scheduling model. The cloud uses the attack mission group as an example to construct the mission group for allocation and performance comparison of task throughput, message average end-to-end delay, and task completion time. The simulation results show that compared with the platform-level single-layer task scheduling, the module-level multi-task scheduling model reduces the average completion time of a single task by 46.2% and the mission group completion time by 52.1%. In addition to guarantee of throughput, the model proposed has more stable scheduling capability for complex tasks. In terms of network performance, the module-level multi-layer task scheduling model has lower end-to-end delay and more concentrated delay distribution, which improves the real-time performance of network message transmission.

Key words: Unmanned Aerial Vehicle (UAV) swarm, avionics cloud, task scheduling, hierarchical clustering, task throughput