Evaluation Methodologies of Dynamically Reconfigurable Systems in the Automotive Industry

Classical decentralized architectures based on large networks of microprocessorbased Electronic Control Units (ECU), namely those used in self-driving cars and other highly-automated applications used in the automotive industry, are becoming more and more complex. These new, high computational power demand applications are constrained by limits on energy consumption, weight, and size of the embedded components. The adoption of new centralized Electric/Electronic (E/E) architectures based on dynamically reconfigurable hardware represents a new possibility to tackle these challenges. However, they also raise concerns and questions about their safety. Hence, an appropriate evaluation must be performed to guarantee that safety requirements resulting from an Automotive Safety Integrity Level (ASIL) according to the standard ISO 26262 are met. In this work, a methodology for the evaluation of dynamically reconfigurable systems based on centralized architectures is presented. The aim is to evaluate the reliability and probability of failure while exploring the design space and identify key aspects for continuous improvement. The methodology is divided into three stages. In the first stage, the system is decomposed, and its sub-systems are isolated before applying a Fault Tree Analysis on the elements of each sub-system. The mathematical stochastic model of Markov Chain is used in the second stage to obtain the reliability function and the quantification of the Mean Time to Failure of the system in Failure In Time format. Finally, the model is submitted to stress test by introducing uncertainties into the model and compute them through a Monte Carlo Simulation. Combined with a statistical analysis it is possible to obtain insights regarding key aspects of the model. Applying this method it is possible to evaluate each sub-system independently and obtain the respective ASIL decomposition of the overall system. With the application of this methodology, we can evaluate the reliability and performance of dynamically reconfigurable systems and define new E/E automotive architectures and scenarios