Pattern-based Analysis of Interaction Failures in Systems-of-Systems: a Case Study on Platooning

• Published in: 2020 27th Asia-Pacific Software Engineering Conference (APSEC) pp. 326 - 335
• Main Authors: Hyun, Sangwon, Song, Jiyoung, Shin, Seungchyul, Baek, Young-Min, Bae, Doo-Hwan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2020
• Subjects: Benchmark testing; Data mining; Fault Analysis; Fault diagnosis; Interaction Failure; Knowledge based systems; Platooning System-of-Systems; Software algorithms; Software engineering; Traffic congestion
• ISSN: 2640-0715
• DOI: 10.1109/APSEC51365.2020.00041

Interactions between software components play a major role in the achievement of goals in complex systems, such as platooning System-of-Systems (SoS). A platooning SoS groups vehicles in order to increase their fuel efficiency and alleviates traffic congestion by enabling driving in close proximity using operation protocols. In a platooning SoS, the execution of typical operations, such as Leave or Merge, consists of 20 micro-operations on average. Owing to this overabundance of sub-operations, interaction failures in a specific operation sequence can occur in an SoS execution. Further, analyzing the root cause of such failures is highly time-consuming, due to the density of the constituent interactions. Existing techniques suffer from two limitations: (1) The majority of the root cause analysis techniques are not capable of isolating faulty interaction sequences, because they do not directly utilize interaction data; (2) The majority of the fault diagnosis techniques assume the preexamined fault knowledge base, which needs too high cost due to limited knowledge in an SoS. To effectively analyze interaction failures in an SoS, we propose a pattern-based faulty interaction analysis technique. To this end, an interaction model is first defined for an SoS, followed by the proposal of a suspicious interaction pattern mining algorithm. During the case study using a platooning simulator, the technique automatically abstracts interaction data from logs and extracts faulty interaction patterns, thereby enabling the identification of seven new unreported interaction failure scenarios. The conclusions of this study can enrich the general fault knowledge base for platooning SoS.