Stage-oe-small.jpg

Thema4826/en

Aus Aifbportal
Wechseln zu:Navigation, Suche


Deep Reinforcement Learning for Autonomous Driving


'


Information on the Thesis

Type of Final Thesis: Master
Supervisor: Daniel Bogdoll
Research Group: Applied Technical Cognitive Systems

Archive Number: 4.826
Status of Thesis: Completed
Date of start: 2021-12-02

Further Information

The AIFB and FZI deal with all kinds of issues related to autonomous driving. Our reinforcement learning pipeline for autonomously acting vehicles is based on CARLA and ClearML with current Nvidia RTX GPUs. In particular, we investigate state representations and knowledge integration for autonomous agents. Situations with conflicting traffic rules are a major challenge.

For this we need support in numerous areas, where you can contribute according to your strengths and interests. Possible topics are Graph-based State Representation, State-of-the-art Benchmarking, CARLA AD Challenge Participation, Knowledge Integration or Situational Awareness. Please also apply if you are interested in a Hiwi position, BA etc.


TASKS

  • Literature research, analysis and evaluation of the state of the art
  • Implementation and evaluation of selected algorithms in Python


WE OFFER

  • An interdisciplinary working environment with partners from science and industry
  • Challenging tasks in an exciting and highly up-to-date subject area
  • Regular meetups with the students in my team
  • In case of outstanding work, submission of a paper for publication at a conference
  • Free, independent working style with short, structured weekly meetings for regular feedback


WE EXPECT

  • Good Python programming skills (under Linux with Git)
  • Theoretical knowledge in the area of Machine Learning / Deep Learning
  • Practical experience with Tensorflow or PyTorch
  • Independent thinking and working, motivation and commitment
  • Fluent in English or German
  • Bonus points are given for experience with RL, ROS, CARLA, ClearML and LaTeX


REQUIRED DOCUMENTS

  • Two sentences about your motivation (included in the e-mail)
  • Current transcript of records (and if available, bachelor's degree certificate)
  • Curriculum vitae in tabular form


CONTACT

Daniel Bogdoll