Rearrangment tasks demand a combination of essential robotic skills, including navigation, picking, and placing. This thesis aims to leverage recent advances in Hirarchical and Meta Reinforcement Learning to develop an intelliegent agent capable of seamlessly executing the entire rearrangement process. The focus is on enabling effective learning and coordination between these diverse skills, ensuring adaptability and efficiency in dynamic environemnts.

Enrolled in computer science or simialr MSc program in and around Bonn/Cologne

Experience with Reinforcement Learning

Familarity with robotic simulators such as Habitat AI and Isaac Gym

Programming experience with Python, C++, and ROS

Enthusiasm for real-world robot deployment and scientific publishing of results

[1] N. Yokoyama et al., " ASC: Adaptive Skill Coordination for Robotic Mobile Manipulation", RAL 2024

[2] AN. Harish et al., "Reinforcement Learning via Auxiliary Task Distillation", ECCV 2024