Higher computational power, new dimensions of interconnectivity and modern machine learning techniques are necessary for building a fully autonomous car, but exhibit an enormous technical complexity. Research about new approaches and technology for handling this complexity raises a problem: On the one side, researchers advocate transitions and replacements for the current systems mainly without deploying them in real cars on the streets. On the other side applying theoretical approaches without clear evidence of their practical benefits is risky for the practitioners. As a solution to close this gap, researchers should bring their ideas more often into physical cars and support their proposals with measurements from realistic experiments. With this paper, we share our insights from an academic perspective about connecting scientific prototypes with a real car. (1) We discuss three interface designs for setups with differing connectivity to a running car; (2) We provide a checklist for planning and organizing real car experiments including a discussion of involved trade-offs; (3) We give practical advice and identify best practices learned from our own experiments inside a car. In sum, we demonstrate that even with a short budget and a small team size it still is possible to bring prototypes into real cars.     «

Higher computational power, new dimensions of interconnectivity and modern machine learning techniques are necessary for building a fully autonomous car, but exhibit an enormous technical complexity. Research about new approaches and technology for handling this complexity raises a problem: On the one side, researchers advocate transitions and replacements for the current systems mainly without deploying them in real cars on the streets. On the other side applying theoretical approaches without...     »