Architecture of optimized digital twins for AI-based training
report
The ASIMOV project investigates the combination of Digital Twins (DTs) and Artificial Intelligence (AI) to find the opportunities and challenges for automated optimization and calibration of complex high-tech systems in complex environments. In many cases the actual system is not available for training AI components, therefore a dedicated digital twin or digital model is set up for providing that training data. The AI component is set up to capture the complexity of the real system and control and optimize the system in its operation. This report focusses on the DT part and starts by describing the main challenges in creating a model mimicking the behavior of a real system. In the state-of-the-art section, the high-level architectural descriptions currently published are discussed. This report describes next the reference architecture for a digital twin specifically for training AI systems. This reference architecture considers from various viewpoints how such a digital twin / digital model can be designed and developed. It should be used as a guideline for creating the specific architecture for a concrete digital twin / digital model. It contains generic elements, aspects, and best practices. This reference architecture description also provides challenges, concerns, and questions to be answered by the systems architects and system designers. This architectural part is followed by extensive elaboration on modeling system behavior. It provides insights on which types of modeling exist, how to model behavior, and the consequences for use in AI training.
Topics
TNO Identifier
1006498
Publisher
Asimov
Collation
69 p.
Place of publication
Eindhoven