Statistical Verification for UAV Autonomy

We discuss the role of statistical methods for the verification and validation of automation and control systems in autonomous vehicles. Statistical learning theory is presented as a verification framework with broad-based application. We treat one example in some detail: The identification of a computationally feasible operational region for a high-performance algorithm. The theory in this case can provide rigorous estimates of the true reliability of the algorithm. Some other possible applications are also outlined. A rotorcraft platform under development, the Organic Air Vehicle, is used to illustrate the applications.