Scalable CPS Analysis of Robustness to Failures

Objetivo

Cyber-physical systems underpin critical infrastructures such as transport, energy, aerospace, and healthcare. They increasingly operate with high autonomy, a necessity given the dynamic environments they face and the constantly evolving requirements imposed on them. Yet, they are inherently vulnerable to chronomalies, timing disruptions that may arise unexpectedly and were not foreseen at design time. The Mars helicopter Ingenuity, which suffered dangerous oscillations due to timestamp misalignments, and later faced another unexplained synchronization failure, illustrates how even state-of-the-art systems, designed with incredible budgets, can be affected by such problems.

Current verification approaches are ill-suited to deal with chronomalies: they simplify reality by overestimating failure frequencies, producing results that are too conservative to enable certification of real systems. SCARF tackles this challenge with a principled framework for both analysis and design. It introduces a formal treatment of chronomalies, using both stochastic and deterministic failure models and combining them with control-theoretic representations of the computational effect on the physical system.

Using advanced tools from linear algebra, such as the joint spectral radius and linear matrix inequalities, SCARF will enable the synthesis of controllers that are resilient to chronomalies by design. Using formally proven transformations, SCARF will tackle realistic patterns of timing failures. The outcome will be scalable verification and design tools that ensure cyber-physical systems can withstand unforeseen timing anomalies. By providing rigorous guarantees of stability and robustness, SCARF will strengthen certification and trust in domains where reliability is paramount, including automotive, energy, and transportation.