Enabling High-Performance Onboard Computing With Virtualization For Unmanned Aerial Systems

In recent years, unmanned aerial systems (UAS) have attracted significant attentions because of their broad civilian and commercial applications. Nevertheless, most existing UAS platforms only have limited computing capabilities to perform various delay-sensitive operations. To tackle this issue, in this paper, we develop a high-performance onboard UAS computing platform with the virtualization technique. Specifically, we first discuss the selection of microcomputers that are suitable for UAS onboard computing. We then investigate virtualization schemes that can effectively manage constrained resources in UAS, flexibly support UAS applications, and enable resource sharing among multiple UAS to achieve higher computing power. In our study, we compare the performance (such as computing, network, isolation, etc.) of two representative virtualization techniques including virtual machine (VM) and container, using KVM and Docker, respectively. Extensive experimental results demonstrate the performance trade-offs between VM and container, and validate the benefits of virtualization in supporting real UAS applications.