Conf-Adaption: Adaptive Adjustment of Software Configuration On UAV by Resource Dependency Analysis

With the rapid growth of software scale and complexity, the runtime behavior of software is increasingly dependent on system resources. Nowadays, more and more software failures are caused by resources available to the system, which call for software capability to be adaptive to resource changes. Previous research mainly focused on building adaptive models for the target software. However, building an adaptive model can be very complex and expensive. Our investigation on the UAV (Unmanned Aerial Vehicle) flight control software demonstrate that configuration options could directly affect the consumption of UAV power, which shed light on the design of Conf-Adaption, a tool that could automatically adjust the software configuration based on the changes of system resource according to the external environment. Specifically, we design a lightweight method to analyze the dependency between software configuration and resource, and obtain configuration options that strongly affect resource consumption. Conf-Adaption then adaptively adjusts the configuration to adapt or meet the flight mission. Our experiments on the UAV software demonstrate that when the mission of the UAV changes, Conf-Adaption can automatically adjust the UAV configuration and the 91.7% adjustments can meet the requirements of the mission and effectively extend the flight time.