An empirical characterization of software bugs in open-source Cyber–Physical Systems

Background: Cyber-Physical Systems (CPSs) are systems in which software and hardware components interact with each other. Understanding the specific nature and root cause of CPS bugs would help to design better verification and validation (V&V) techniques for these systems such as domain-specific mutants. Aim: We look at CPS bugs from an open-source perspective, trying to understand what kinds of bugs occur in a set of open-source CPSs belonging to different domains. Method: We analyze 1151 issues from 14 projects related to drones, automotive, robotics, and Arduino. We apply a hybrid card-sorting procedure to create a taxonomy of CPS bugs, by extending a previously proposed taxonomy specific to the automotive domain. Results: We provide a taxonomy featuring 22 root causes, grouped into eight high-level categories. Our qualitative and quantitative analyses suggest that 33.4% of the analyzed bugs occurring in CPSs are peculiar to those and, consequently, require specific care during verification and validation activities. Conclusion: The taxonomy provides an overview of the root causes related to bugs found in open-source CPSs belonging to different domains. Such root causes are related to different components of a CPS, including hardware, interface, configuration, network, data, and application logic.