ConDU: Method for On-the-fly Detection of Non-deadlock Concurrency Errors in UAV Software
Jinseok Park, Chang-Hui Bae, Seongjin Lee · 2024
This paper presents ConDU, an innovative method for detecting non-deadlock concurrency errors in Unmanned Aerial Vehicle (UAV) software. Focusing on urban operational safety, ConDU addresses a critical gap in UAV research by targeting software errors, particularly in multi-threaded systems. It functions by monitoring shared variables, threads, and access patterns in real-time, identifying discrepancies indicative of potential errors. ConDU was rigorously tested using synthetic programs and applied to the ArduPilot-based Software-In-TheLoop (SITL) environment. It successfully detected all known and additional concurrency errors with a minimal time overhead of 5. 16%. This research marks a significant advance in ensuring UAV software reliability and operational safety.