An Evaluation Of Monte-Carlo Tree Search For Property Falsification On Hybrid Flight Control Laws

The formal verification and validation of real-world, industrial critical hybrid flight controllers remains a very challenging task. An increasingly popular and quite successful alternative to formal verification is the use of optimization and reinforcement learning techniques to maximize some real-valued reward function encoding the robustness margin to the falsification of a property. In this paper we present an evaluation of a simple Monte-Carlo Tree Search property falsification algorithm, applied to select properties of a longitudinal hybrid flight control law: a threshold overshoot property, two frequential properties, and a discrete event-based property.