Verification of machine learning based cyber-physical systems: a comparative study

Arthur Clavière (1) , Laura Altieri Sambartolomé (2) , Eric Asselin (2) , Christophe Garion (3) , Claire Pagetti (1)

1 DTIS, ONERA, Université de Toulouse [Toulouse]
2 Collins Aerospace France
3 ISAE-SUPAERO - Institut Supérieur de l'Aéronautique et de l'Espace

In this paper, we conduct a comparison of the existing formal methods for verifying the safety of cyber-physical systems with machine learning based controllers. We focus on a particular form of machine learning based controller, namely a classifier based on multiple neural networks, the architecture of which is particularly interesting for embedded applications. We compare both exact and approximate verification techniques, based on several real-world benchmarks such as a collision avoidance system for unmanned aerial vehicles.

Mots clés: Neural networks; Safety; Formal methods

Domaines: Sciences de l'ingénieur [physics]

Arthur Clavière, Laura Altieri Sambartolomé, Eric Asselin, Christophe Garion, Claire Pagetti. Verification of machine learning based cyber-physical systems: a comparative study. 21ème journée Approches Formelles dans l'assistance au Developpement de logiciel - AFADL 2022, Jun 2022, Vannes, France. ⟨hal-03813800⟩