Evaluation of Error Handling Mechanisms for Automotive Embedded Systems

Abstract

This thesis presents an evaluation of the effectiveness and time overhead for plausibility checks in automotive electrical and electronic systems. Plausibility checks aim to detect errors caused by software bugs and random hardware failures. They are commonly used to ensure safety and robustness. There are two ways to implement plausibility checks, the traditional with checks directly in the source code and the use of a library. We have created a proof-of-concept implementation of an AUTOSAR compliant library that provides a standardised interface for common plausibility checks. We demonstrate the usefulness of the library for several AUTOSAR applications. We investigate the effectiveness of plausibility checks on three applications and found that plausibility checks are effective in detecting errors in input parameters to software modules. We compare the time overhead for implementing plausibility checks with the library versus implementing them directly in the source code. Using the library increases the execution time with 2.0 s per library call on our hardware compared to having no checks at all. In comparison, having the plausibility checks directly in the source code increase the execution time with 0.25 s per check.