Applying Application-Level Correctness to the Paparazzi Autopilot
Authors
Larsen, Heine Gatten ; Pedersen, Morten Turn ; Birch, Thomas Viesmose
Term
4. term
Education
Publication year
2014
Submitted on
2014-06-03
Pages
68
Abstract
Ubemannede luftfartøjer (UAV'er) bliver stadig mere udbredte, hvilket øger behovet for billig, tilgængelig software. Det open source Paparazzi-projekt har til formål at levere en prisvenlig software- og hardwareplatform til UAV'er. Paparazzi-softwaren er dog ikke beskyttet mod forstyrrelser udefra, som kan få enkelte bits til at skifte værdi (tilfældige bit-fejl). I denne rapport introducerer vi en måde at sammenligne forskellige flyvebaner, så flyvninger kan vurderes op imod hinanden. Vi præsenterer også en metode, der gør Paparazzi-autopiloten korrekt på applikationsniveau, dvs. at de overordnede funktioner forbliver korrekte, selv når sådanne fejl opstår. Til sidst afprøver vi metoden ved at injicere en single-event upset (SEU) – en kortvarig ændring af en enkelt bit – i autopiloten under en flyvning og måler ydeevnen.
Unmanned aerial vehicles (UAVs) are becoming more common, which creates demand for low-cost, accessible software. The open-source Paparazzi project aims to provide an affordable software and hardware platform for UAVs. However, the Paparazzi software is not protected against external disturbances that can make individual bits change value (random bit flips). In this report, we introduce a way to compare different flight paths so flights can be evaluated against each other. We also present a method that makes the Paparazzi autopilot correct at the application level, meaning its overall functions remain correct even when such errors occur. Finally, we evaluate the method by injecting a single-event upset (SEU)—a brief flip of a single bit—into the autopilot during a flight and measuring performance.
[This abstract was generated with the help of AI]
Keywords
Documents