Flying Data Centers: Smart Virtual Machine Slicing and AI Steer Drones to

A research team at Changsha University of Science and Technology in China proposes a framework for “flying data centers”: drones that provide mobile edge computing while also managing computing capacity among multiple users. In the study published in Cluster Computing, the authors introduce VDDM (VM Dynamic partition and DRL-based Delay Minimization), aimed at minimizing total task delay across all users. The approach combines a mixed-integer nonlinear programming model with a decomposition into three components. One addresses input-output interference between virtual machines by dynamically adjusting how many VMs each drone runs based on task sizes. The work targets scenarios with multiple drones, multiple user equipments, VM configuration, user-to-drone assignments, and drone flight paths.