Traditional methods of system verification call for the repeated exercise of that system in its intended environment to give confidence that it will operate as designed. Autonomous systems, however, are systems designed for sophisticated operation in stochastic environments that specifically handle the unexpected; testing all possible operational scenarios for such systems is intractably complex. In order to address this challenge, we propose a framework by which these complex systems can be represented as collections of individual abilities, each of which can be independently verified based on their inputs and outputs. We then develop a corresponding mathematical framework to provide assurance of performance against desired mission parameters by composing the results of these component abilities.
展开▼