This talk explores the integration of Predictive Runtime Verification (PRV) and Process Mining techniques to enhance the reliability and efficiency of autonomous systems. We begin by presenting a multi-model PRV approach that addresses the complexity of systems composed of multiple components. This method improves traditional PRV by utilizing individual component models, enabling more accurate predictions and timely interventions. Next, we illustrate how Process Mining can be used to generate detailed models from execution logs, providing a dynamic foundation for PRV. By applying these techniques to the Curiosity rover, we show how the rover's navigation and decision-making processes can be monitored and verified in real-time. The talk includes practical examples and results from simulations, highlighting how the combination of PRV and Process Mining can predict and prevent deviations from expected behavior, ultimately ensuring the safety and success of autonomous missions.
- Speaker
- Angelo Ferrando
- Venue
- Meston G05