Assessing the Use of Machine Learning to Find the Worst-Case Execution Time of Avionics Software

Many parts in aircraft today rely on software that interacts with its physical environment. Typically, this interaction involves taking sensor readings, sending actuation commands, reading commands from humans, and presenting information to humans. These interactions require that the software deliver results at the right time, as argued in the guidance document DO-178C and in previous Federal Aviation Administration (FAA) reports. Correct timing, in turn, depends on the execution time of individual programs. Hence, the problem of finding an upper bound on the execution time of a program, called Worst-Case Execution Time (WCET) analysis, is an important step in avionics certification. Unfortunately, WCET analysis is difficult for several reasons. A program can typically execute a large number of different paths. During the execution of one path, the program uses resources in a complex way and this resource use is difficult to analyze. Finally, during the execution of one path, the speed of execution depends on temperature, which, in turn, depends on earlier execution. This report presents research on WCET analysis using Machine Learning (ML) and Artificial Intelligence (AI) aiming to make WCET analysis less dependent on detailed knowledge of the program that is analyzed and the hardware used.

• Record URL:
  • https://doi.org/10.21949/1528215
• Record URL:
  • https://rosap.ntl.bts.gov/view/dot/67169
• Corporate Authors:

  Carnegie Mellon University

  Software Engineering Institute
  Pittsburgh, PA  United States  15213

  Federal Aviation Administration

  William J. Hughes Technical Center
  Aviation Research Division
  Atlantic City International Airport, NJ  United States  08405
• Authors:
  • Andersson, Bjorn
  • de Niz, Dionisio
  • Moreno, Gabriel
  • Hansen, Jeffery
  • Klein, Mark
• Publication Date: 2023-4

Language

• English

Media Info

• Media Type: Digital/other
• Edition: Final Report
• Features: Appendices; Figures; References; Tables;
• Pagination: 295p

Subject/Index Terms

• TRT Terms: Artificial intelligence; Avionics; Data processing operations; Software
• Subject Areas: Aviation; Data and Information Technology; Vehicles and Equipment;

Filing Info

• Accession Number: 01882566
• Record Type: Publication
• Report/Paper Numbers: DOT/FAA/TC-23/06, DM22-1197
• Contract Numbers: FA8702-15-D-0002
• Files: NTL, TRIS, ATRI, USDOT
• Created Date: May 23 2023 9:26AM