Towards Trustworthy AI-driven Autonomous Systems : Adaptation, Resilience, and Explainable AI

The proposed research is to identify the underlying theoretical and computational aspects in AI driven ASs with a) embedded adaptation and resilience in both unit vehicle AS and network of ASs scale (fleets and swarms) while ensuring b) explainable AI in a learning enabled context and while providing c) dynamic verification and validation for feasibility, safety and performance assurance. All these three elements are deemed as critical towards designing deployable trustworthy ASs and systems of ASs. In that respect, the proposed study should not only lay the foundations, but also demonstrate the methods for generating dynamically evolving and learning AI implementations of guidance, navigation and control with explainable structures that can further provide real-time bounds and certificates of feasibility, safety and performance.