CPS prototype development for AI-based scenario adaptation in flight simulator training

Abstract Evidence-based training as part of competency-based training and assessment confronts pilots with unexpected events in realistic scenarios in order to promote problem-solving and adaptability. Linking theory and practice is essential to promote these competencies. To achieve this, a cyber-physical system is presented that enables this through the innovative approach of “deep-linking keywords.” A heuristic scoring function determines a fulfillment score for each keyword. Based on the assessment, scenario-based training is adapted, enabling necessary individualization. Compared to existing systems, the prototype generates a coherent dataset that bridges knowledge work and scenario-based training, allowing for comprehensive scenario adaptation. The cyber-physical system consists of a computer-based training system built on the Django framework, a Basic Instrument Training Device, and flight simulator software, integrated via an application programming interface. After each evidence-based training session, performance data are processed through structured analysis pipelines to extract and evaluate scenario-linked feature vectors. This enables iterative parameter optimization for adaptive scenario control. Building on the prototype and the proven effectiveness of the heuristic scoring function, a large dataset will be compiled, and the rule-based method will be replaced by machine learning to enhance safety, effectiveness, and efficiency in aviation through highly individualized training enabled by an AI-based cyber-physical system.