Formal timing analysis of automatic repeat request for switched real-time networks

Switched networks are widely used in today's real-time embedded systems - be it on-chip (Network on Chip) or off-chip (Ethernet). They are scalable, cheap, and provide high data rates. However, communication is affected by transmission errors or packet drops. Error-control protocols such as Stop and Wait and Go-Back-N provide an end-to-end mechanism to detect erroneous and dropped packets. These protocols are deemed inapplicable to hard real-time communication such as video transmissions for Advanced Driver Assistance Systems (ADAS) because the timing is extremely hard to predict. On the other hand, an integer data delivery must be guaranteed, especially for safety relevant features (i.e. emergency brake assist). The contribution of the paper is a formal consideration of error-control protocols for hard real-time applications. We show that contrary to common belief, it is very well possible to provide formal guarantees on the network latency for Stop and Wait as well as Go-Back-N. Furthermore we show how these protocols can be integrated in an automotive setup.