Model based dependability evaluation for automotive control functions

In this paper, we study the evaluation of reliability for embedded functions in automotive software. With the decrease of circuit size the probability of hardware failures within a chip increases. In order to assess fault tolerant designs for automotive software it is essential to be able to predict the failure rate of the realized function. We present a model based fault injection technique, where faults are injected according to a distribution of probabilities in the function model. Using the MATLAB/Simulink® toolchain, we are able to simulate the faulty behavior of a function even if the realization is distributed amongst several embedded processors in a car. The results allow assessing the dependability of certain safety critical functions in the overall system design.