ULTIMATE is based on industrial use cases showcasing needs and requirements from real-world situations since the first steps of the project to ensure operational conditions will be considered all along the life cycle of the hybrid AI algorithms.

The approach of ULTIMATE is to investigate novel or promising learning (together with symbolic models) approaches to design and develop hybrid AI based algorithms with increased intelligence, autonomy, explainability and interpretability. Several hybrid approaches will be studied, resulting in hybrid AI algorithms for different intents, including state estimation, risk assessment, anomaly detection, decision-support, and control. The design of hybrid AI algorithms will be based on the addition of knowledge (e.g. physical laws) to data (observations) with the aim of improving the functional accuracy and alleviating the evaluation.
Then, these hybrid AI-based solutions will be evaluated and prepared for operational conditions through rigorous methodologies using statistical / experimental / formal methods and tools to consolidate trustworthiness.

Finally, the project will assess the behaviour of these AI algorithms under operational conditions (demonstrators) to check performance, trustworthiness, and autonomy (capability to detect any abnormal behaviour and ability to redirect to human if not in charge of processing) and the adequacy between the validated AI algorithm and its operational use.