Evolutionary computing for mission-based test and evaluation of unmanned autonomous systems

Test and evaluation may be viewed as a technology enabler for the successful deployment of unmanned vehicles and robots in all their envisioned applications. It is however a challenging endeavor, considering that roboticists and developers are not used to thinking of comprehensive test and evaluation as an integral part of robot development. Moreover, the community who has conducted test and evaluation up to this date does not possess the tools to cope with the growing complexity of unmanned and autonomous systems. This paper proposes an approach to one of the hardest problems in testing and evaluation of robots: mission-based test planning. This approach relies on constructive simulation tools and on evolutionary computing techniques for searching in high dimensional spaces of possible test scenarios. The goal of the test planner is to generate a set of tests that make highly efficient use of resources to unveil weaknesses of the system under test in a context of a specific mission.