Estimating end-to-end latencies in automotive cyber-physical systems

Abstract

Controller networks in today's automotive systems consist of more than 100 ECUs connected by various bus protocols. Seamless operation of the entire system requires a well-orchestrated interaction of these ECUs. Consequently, to ensure safety and comfort, a performance analysis is an inherent part of the engineering process. Conducting such an analysis manually is expensive, slow, and error prone. Tool support is therefore crucial, and a number of approaches have been presented. However, most work is limited to either network latencies or software latencies which results in an analysis gap at the transition between different layers of the communication stack. The work presented here introduces an approach to close this gap. Furthermore, we discuss the integration of different methods to obtain an end-to-end latency analysis.