Advances in Verification and Control of Cyber-Physical Systems

The field of cyber-physical systems is witnessing significant developments in verification and control techniques. Researchers are exploring innovative methods to ensure the safety and reliability of these systems, which are critical in various industrial applications. One notable direction is the integration of formal verification techniques, such as model checking, with digital twin models to identify potential errors and improve their behavior. Another area of focus is the development of adaptive control strategies that can handle uncertainties and time-varying constraints, enabling the design of more robust and efficient systems. The use of machine learning and reinforcement learning is also being investigated to improve the autonomy and decision-making capabilities of cyber-physical systems. Noteworthy papers in this area include:

  • Contract-based Verification of Digital Twins, which introduces a novel methodology for verifying digital twin models using model checking and contract-based approaches.
  • Runtime Safety through Adaptive Shielding, which presents a runtime shielding mechanism for reinforcement learning that ensures probabilistic safety guarantees and optimal policies.
  • Adaptive Reinforcement Learning for Unobservable Random Delays, which proposes a framework for handling unobservable and time-varying delays in reinforcement learning. These advances have the potential to significantly impact the development of more reliable, efficient, and autonomous cyber-physical systems.

Sources

Contract-based Verification of Digital Twins

Runtime Safety through Adaptive Shielding: From Hidden Parameter Inference to Provable Guarantees

DTHA: A Digital Twin-Assisted Handover Authentication Scheme for 5G and Beyond

Safe Domains of Attraction for Discrete-Time Nonlinear Systems: Characterization and Verifiable Neural Network Estimation

Robust Adaptive Time-Varying Control Barrier Function with Application to Robotic Surface Treatment

Adaptive Reinforcement Learning for Unobservable Random Delays

Modeling Uncertainty: From Simulink to Stochastic Hybrid Automata

Built with on top of