Intelligent control & supervision for autonomous system resilience in uncertain worlds

As autonomous systems are increasingly employed in high-criticality applications, their safe and reliable operation is an overarching concern. Existing applications rely on extensive system characterization and contingency planning for operating within known and anticipated circumstances. Human intervention and ingenuity is often required in operational environments and contexts for which available knowledge, if any, is incomplete and/or inaccurate (defined by authors as uncertain worlds). As system complexity increases, the validity of this strategy diminishes and is infeasible for many applications precluding human support. Self-managing systems are sought to provide reliable and fully autonomous operation in virtually all operational contexts. This paper presents an adaptive and automated decision-engine for improving autonomous systems' inherent resilience in uncertain worlds. The concept is applied to aviation industry Free Flight initiatives to enhance air traffic management (ATM) system safety with increased autonomous technologies and procedures. The paper concludes with discussion of an agent-based model (ABM) in development to benchmark the concept against (1) RTCA DO-185B / TCAS II - an international standard for airborne collision avoidance and (2) the D* Lite algorithm - the de-facto standard for autonomous vehicle navigation and dynamic planning.