Student Research Abstract: Enhancing Safety in Cyber-Physical Systems Through Runtime Enforcement

Abstract

The verification and safety assurance of cyber-physical systems pose significant challenges, necessitating innovative solutions for their deployment. Machine Learning (ML) has gained prominence for its diverse applications, particularly in autonomous systems. Despite its potential, the integration of ML into safety-critical domains is met with caution due to multiple concerns. This paper introduces a monitoring system capable of predicting and mitigating requirement violations, featuring a discretized mathematical model for predictions. This monitor includes a decision mechanism to transit between high-assurance and high-performance controllers within a simplex architecture depending on the state of the system. This paper shows an example of how the suggested monitor is able to avoid a property violation by predicting it in advance. These advancements hold promise for enhancing the safety and reliability of cyber-physical systems in safety-critical applications.