Functionality and Safety Weaknesses in Integration of Physics-Based Models on Critical Embedded Systems

Embedded automotive software is currently showing trends towards model predictive control (MPC), virtual sensors or model-based diagnosis, mainly used in advanced driver assistance systems (ADAS) and automated driving. Such applications use physical models in the control algorithms. The integration of physical models is a risky task, since weaknesses, such as the need for floating-point arithmetic and discretization or model properties, such as discontinuities and nonlinearities, quickly bring a project to a standstill or establish errors in the final product. The use of known verification and validation methods is often not possible or offers false safety guarantees. This article is intended to help developers understand and identify safety weaknesses and develop new verification and validation methods specifically adapted for physics-based, critical, embedded code. For this purpose, corresponding weaknesses in current industrial projects with physics-based systems have been identified and categorized. In this article, these are described and illustrated with examples from applications in order to get an idea of their relevance in the current context. On this basis, approaches for the analysis and diagnosis of potentially faulty code are proposed to motivate testers and quality managers to find new methods for error identification and validation of critical, physics-based, embedded code.