Automatically Learning Formal Models from Autonomous Driving Software

The correctness of autonomous driving software is utmost importance, as incorrect behavior may have catastrophic consequences. Formal model-based engineering techniques can help guarantee and thereby allow the safe deployment vehicles. However, challenges exist for widespread industrial adoption formal methods. One these model construction problem. Manual models time-consuming, error-prone, intractable large systems. Automating would be a big step towards methods system development, re-engineering, reverse engineering. This article applies active learning to obtain an existing (under development) module implemented in MATLAB. demonstrates feasibility automated automotive use. Additionally, practical applying automata learning, possible directions integrating into development workflow, are discussed.