Notoriously complex hospital systems present a prime use case for FDTs. In collaboration with Foch Hospital in Suresnes, CEA-List developed its first process-centric FDT. This twin integrates all associated administrative and regulatory steps, aiming to evaluate how a project manager’s decisions can impact quality indicators for clinical studies. To ensure the FDT was representative, process models were enhanced with stochastic data on task durations and event probabilities, alongside annotations specifying human agents’ objectives. CEA-List’s model-driven system engineering platform, Papyrus, was tailored to simulate human-driven scheduling guided by goal achievement, leveraging a Monte Carlo Tree Search heuristic. Additionally, fuzzy logic was incorporated into decision-making processes, particularly for task scheduling, using the institute’s symbolic AI technology, ExpressIF®.
We developed two knowledge extraction modules to maximize FDT accessibility. The first module extracts relevant information from regulatory texts using CEA-List’s multilingual linguistic analyzer, LIMA, and enables exploration of legal databases such as Légifrance and EurLex. The second module extracts insights from process data to generate execution logs and a set of constraints based on Allen interval logic.
This approach is applicable to any sector where establishing a functional perspective is critical, and where humans are central to system operation.
Human decisions must not be forgotten in systems modeling and simulation.
Contributed to this article: