Master Thesis Bridging the Gap between Reinforcement Learning & E2E Driving

Are you passionate about the future of autonomous driving? We are seeking a talented and motivated individual to join our team of experts dedicated to advancing the capabilities of autonomous vehicles. In this role, you will play a crucial part in using Reinforcement Learning (RL) to enhance the performance of end-to-end (E2E) approaches.

The field of autonomous driving has experienced a paradigm shift with the emergence of batched RL simulation, enabling relatively cheap closed-loop training of high-performance policies that can learn from own experience without human expert data. In contrast, E2E driving approaches rely on large amounts of rich expert data but are increasingly using RL-like training strategies to inject the notion of experience and acting based on feedback.

This thesis aims to investigate approaches to integrate and enhance state-of-the-art E2E driving policies with RL simulation.

• During your Master thesis, you will collaborate with a team of engineers and researchers to bridge the gap between RL simulation and training, and E2E driving.
• Furthermore you will understand the fundamental properties behind different training strategies and use them to guide the development of novel models and policies.
• You will engineer and contribute efficient and high-performance software.
• In Addition you will conduct experiments and analyze data to identify areas for improvement and optimize model accuracy and reliability.
• You will stay up to date with the latest advancements in autonomous driving technology and contribute innovative ideas to the team.
• Finally, you will document findings and present results in a publishable manner as well as work on open-source benchmarks and datasets.