Correctness of autonomous driving systems is crucial as incorrect behaviour may have catastrophic consequences. Many different hardware and software components (e.g. sensing, decision making, actuation, and control) interact to solve the autonomous driving task, leading to a level of complexity that brings new challenges for the formal verification community. Though formal verification has been used to prove correctness of software, there are significant challenges in transferring such techniques to an agile software development process and to ensure widespread industrial adoption. In the light of these challenges, the identification of appropriate formalisms, and consequently the right verification tools, has significant impact on addressing them. In this paper, we evaluate the application of different formal techniques from supervisory control theory, model checking, and deductive verification to verify existing decision and control software (in development) for an autonomous vehicle. We discuss how the verification objective differs with respect to the choice of formalism and the level of formality that can be applied. Insights from the case study show a need for multiple formal methods to prove correctness, the difficulty to capture the right level of abstraction to model and specify the formal properties for the verification objectives.

No abstract

No abstract

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