ScholarWorks@경상국립대학교

초록

Health management of Unmanned Aerial Systems (UAS) ensures safe and reliable operation by detecting, diagnosing, predicting, and mitigating errors that occur during operation. Software errors in UAS are one of the significant causes of system failures, making health management necessary. However, existing research on UAS health management primarily addresses software errors caused by hardware and data, failing to diagnose internal execution errors within the software. This study introduces Software Health Diagnosis Sensor (SHDS) that diagnoses internal software execution errors in real-time within UAS. SHDS monitors thread access events to shared variables in real-time to diagnose concurrency errors, a type of software execution error. Experiments were conducted on 13 cases in the ArduPilot Software-In-The-Loop (SITL) environment to demonstrate SHDS's capability in diagnosing concurrency errors and the feasibility of asymmetric race repair. The experimental results show that SHDS accurately diagnoses all 13 concurrency errors with an average overhead of approximately 3.40% and successfully repairs eight asymmetric races with an average overhead of approximately 5.39%. © 2025 The Korean Society for Aeronautical and Space Sciences.

키워드