Introducing Automotive Ethernet into an autonomous driving research platform: A feasibility study for using SOME/IP in a scaled model car

Abstract

Car manufacturers are increasingly focusing on developing hardware and software technologies for autonomous driving. Scaled model cars with development processes and architectures similar to fullsize prototypes are good platforms for such research.

This thesis investigates the possibility of modifying the hardware and software architecture of the Audi Autonomous Driving Cup 2016 model car for an autonomous driving usecase which utilizes an Automotive Ethernet/IP technology.

Scalable Service Oriented MiddlewarE over IP (SOME/IP) is chosen to control and access the 3D camera on the car. A Raspberry Pi is introduced via an Ethernet switch as an additional processing node on the network to emulate an Ethernet camera.

A feasibility analysis on a minimal test setup using GENIVI’s vsomeip C++ library and tools such as Wireshark and iPerf is also performed as a part of this thesis.

The results show that the proposed modifications in software and hardware introduce minimal overhead and can allow for a service-oriented application design for autonomous driving scenarios. The results also open the door for a distributed image processing architecture with multiple onboard cameras for the demonstrator.