Maryam Bagheri

Maryam Bagheri, Josephine Lamp, Xugui Zhou et al.

Machine Learning (ML) technologies have been increasingly adopted in Medical Cyber-Physical Systems (MCPS) to enable smart healthcare. Assuring the safety and effectiveness of learning-enabled MCPS is challenging, as such systems must account for diverse patient profiles and physiological dynamics and handle operational uncertainties. In this paper, we develop a safety assurance case for ML controllers in learning-enabled MCPS, with an emphasis on establishing confidence in the ML-based predictions. We present the safety assurance case in detail for Artificial Pancreas Systems (APS) as a representative application of learning-enabled MCPS, and provide a detailed analysis by implementing a deep neural network for the prediction in APS. We check the sufficiency of the ML data and analyze the correctness of the ML-based prediction using formal verification. Finally, we outline open research problems based on our experience in this paper.

Maryam Bagheri, Marjan Sirjani, Ehsan Khamespanah et al.

Autonomous traffic control systems are large-scale systems with critical goals. Due to the dynamic nature of the surrounding world of these systems, assuring the satisfaction of their properties at runtime and in the presence of a change is important. A prominent approach to assure the correct behavior of these systems is verification at runtime, which has strict time and memory limitations. To tackle these limitations, we propose Magnifier, an iterative, incremental, and compositional verification approach that operates on a component-based model. The Magnifier idea is zooming on the component affected by a change, verifying the correctness of properties of interest of the system after adapting the component to the change, and then zooming out and tracing the change if it propagates. If the change propagates, all components affected by the change are adapted and are composed to form a new component. Magnifier repeats the same process for the new component. This iterative process terminates whenever the propagation of the change stops. In Magnifier, we use the Coordinated Adaptive Actor model (CoodAA) of traffic control systems. We present a formal semantics for CoodAA as a network of Timed Input-Output Automata (TIOAs). The change does not propagate if TIOAs of the adapted component and its environment are compatible. We implement our approach in Ptolemy II. The results of our experiments indicate that the proposed approach improves the verification time and the memory consumption compared to a non-compositional approach.