Verification of Decision Making Software in an Autonomous Vehicle: An Industrial Case Study
Paper in proceedings, 2019

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.

Autonomous driving · Formal verification · Supervisory Control Theory · Model checking · Deductive verification

Author

Yuvaraj Selvaraj

Chalmers, Electrical Engineering, Systems and control, Automation

Wolfgang Ahrendt

Chalmers, Computer Science and Engineering (Chalmers), Formal methods

Martin Fabian

Chalmers, Electrical Engineering, Systems and control, Automation

Lecture Notes in Computer Science

0302-9743 (ISSN) 1611-3349 (eISSN)

Vol. 11687 143-159

24th International Conference on Formal Methods for Industrial Critical Systems, FMICS 2019
Amsterdam, Netherlands,

Automatically Assessing Correctness of Autonomous Vehicles (Auto-CAV)

VINNOVA, 2018-03-01 -- 2021-12-31.

Areas of Advance

Transport

Subject Categories

Software Engineering

Robotics

Control Engineering

Computer Science

Computer Systems

DOI

10.1007/978-3-030-27008-7_9

More information

Latest update

9/8/2019 9