Transformer architectures show revolutionary results in various applications, e.g. intelligent chatbots. First studies applied this technology to E2E trained self-driving vehicles. Open loop evaluations based on public large-scale datasets promise a high potential for improving safety and efficiency. Further steps are needed to prove their benefit in real-world situations. This study presents first-hand experiences and results in building an E2E tool chain for development and validation of transformer-based automated driving stacks, including training infrastructure, simulation, re-simulation and closed-loop tests with a prototype vehicle. The results of real-world tests underscore the high potential of E2E stacks for improving the safety and efficiency of self-driving vehicles.

• Gain insights into first-hand experiences and results in building an E2E tool chain
• Discover the development and validation of transformer-based automated driving stacks
• Get a better understanding of real-world tests, underscoring the high potential

• What are the key drivers pushing AD towards E2E trained stacks?
• What are blockers for E2E rollout and which solutions are emerging?
• Which impact does open source and the “AI community” have on E2E development?
• Will E2E stacks scale down to lower automatization levels?
• How will E2E stacks transform the AD ecosystem and industry?