一句话：首个针对无人机自动驾驶软件缺陷的大规模实证研究，归纳出八类无人机特有缺陷、根因、修复策略与五类挑战。
无人机在现代社会中的作用日益重要，其软件缺陷可能导致系统崩溃、挂起、未定义行为，甚至被攻击者利用造成灾难性后果，因此亟需能检测与修复无人机缺陷的技术。然而，已有工作虽研究过各类软件的缺陷，却从未系统刻画无人机软件缺陷的特征，这阻碍了无人机可靠性保障工具的发展。为填补这一空白，本文对两个著名开源无人机自动驾驶平台 PX4 与 Ardupilot 开展了首个大规模实证研究。作者分析了这两个项目中的 569 个缺陷，归纳出八种无人机特有缺陷类型——limit、math、inconsistency、priority、parameter、hardware support、correction 与 initialization——并厘清了其根因。基于该缺陷分类，论文进一步总结了常见缺陷模式与修复策略，并识别出检测与修复此类缺陷所面临的五项挑战。该研究有助于研究者与从业者理解无人机系统可靠性的威胁，并为未来无人机缺陷诊断工具的开发提供依据。
Unmanned aerial vehicles (UAVs) are becoming increasingly important and widely used in modern society. Software bugs in these systems can cause severe issues, such as system crashes, hangs, and undefined behaviors. Some bugs can also be exploited by hackers to launch security attacks, resulting in catastrophic consequences. Therefore, techniques that can help detect and fix software bugs in UAVs are highly desirable. However, although there are many existing studies on bugs in various types of software, the characteristics of UAV software bugs have never been systematically studied. This impedes the development of tools for assuring the dependability of UAVs. To bridge this gap, we conducted the first large-scale empirical study on two well-known open-source autopilot software platforms for UAVs, namely PX4 and Ardupilot, to characterize bugs in UAVs. Through analyzing 569 bugs from these two projects, we observed eight types of UAV-specific bugs (i.e., limit, math, inconsistency, priority, parameter, hardware support, correction, and initialization) and learned their root causes. Based on the bug taxonomy, we summarized common bug patterns and repairing strategies. We further identified five challenges associated with detecting and fixing such UAV-specific bugs. Our study can help researchers and practitioners to better understand the threats to the dependability of UAV systems and facilitate the future development of UAV bug diagnosis tools.
DOI 原文 ·
@inproceedings{paperbot6295,
title = {An exploratory study of autopilot software bugs in unmanned aerial vehicles},
author = {Dinghua Wang and Shuqing Li and Guanping Xiao and Yepang Liu and Yulei Sui},
booktitle = {FSE},
year = {2021},
doi = {10.1145/3468264.3468559}
}