Planning and Scheduling with External Functions in the ExploDTwin project

A Digital Twin (DT) is a digital representation of a physical asset or system, which can be used to simulate, monitor, control, and optimize the physical counterpart. Differently from a classical simulation model, a DT can be dynamically updated (e.g., by means of Machine Learning (ML) techniques) to continuously maintain an alignment with the real asset. In the space domain, the ExploDTwin ESA project proposes to create a model-based DT providing a formal representation of the asset complemented with ”external functions” used to model some parameteres and/or behaviors of the system in an opaque way. Such functions can be engineered or learned from data using ML, but the services built on top of the DT need to cope with such black-box representations. In this paper, we describe the vision of the ExploDTwin project, the modeling and infrastructure being built and then we focus on the planning and scheduling challenges for building a “What-If Analysis” service on top of the DT. We present the architecture of the service, and describe a basic algorithm to handle external functions using off-the-shelf planners.