Stage-oe-small.jpg

Thema4754

Aus Aifbportal
Wechseln zu:Navigation, Suche




Integrating rule deviations into an reinforcement learning APPROACH for autonomous driving




Informationen zur Arbeit

Abschlussarbeitstyp: Master
Betreuer: Daniel Bogdoll
Forschungsgruppe: Angewandte Technisch-Kognitive Systeme
Partner: FZI Forschungszentrum Informatik
Archivierungsnummer: 4754
Abschlussarbeitsstatus: Offen
Beginn: 01. Juni 2021
Abgabe: unbekannt

Weitere Informationen

In the research project KI-WISSEN, the FZI explores the research question of how knowledge can be integrated into machine learning systems for autonomous vehicles. In this master thesis you will deal with RL and investigate possibilities to integrate knowledge about behavior as a traffic participant into the reward function.


AUFGABEN

  • Literature research and evaluation of the SotA
  • Implementation of a SotA RL agent as a baseline for further evaluation within CARLA
  • Designing the reward function to incorporate knowledge about typical behaviour of traffic participants in cooperation with one of the hiwis in my team
  • Define baseline scenarios to evaluate your agent which need a deviation from StVO rules to perform progress


WIR BIETEN

  • An interdisciplinary working environment with partners from science, industry and society
  • Insights into cutting-edge research in the field of Informed ML
  • State of the art Nvidia GPUs embedded in a ClearML environment for training
  • Constructive and pleasant teamwork within my team of students


WIR ERWARTEN

  • Good Python programming skills
  • Knowledge of deep learning, optional RL
  • Experience with Tensorflow or PyTorch, optional Clear ML
  • Self-reliant thinking and working
  • Motivation and commitment
  • Fluent in English or German


ERFORDERLICHE UNTERLAGEN

  • Two sentences about your motivation (included in the e-mail)
  • Current transcript of records
  • Curriculum vitae in tabular form


KONTAKT

Daniel Bogdoll