PI: Dr. Akbas
With the advances in artificial intelligence and perception technologies, autonomous cyber-physical systems (CPS) are poised to have an invaluable impact on safety, access, and efficiency in various industries (aerial, ground, marine, space, etc.). The gating factor towards the active deployment of these systems is the challenge of validation and verification. Without the resolution of this issue, a clear measurable paradigm for autonomous CPS safety cannot be built and a broad-based deployment of these systems cannot be realized. The goal of this proposal is to create a Validation and Verification Framework for Autonomous Aerial Vehicles and present a validation scheme for the decision making component of these vehicles. This work leverages insights from complex validation schemes in industry and presents three significant contributions on top of a conventional Operational Design Domain. The validation paradigm introduces the concept of separating the validation concerns with associated abstraction levels. An automated framework is proposed to perform constrained random walks through the state space, performing coverage analysis and exposing the worst-case conditions.