Long-Horizon Dexterous Manipulation from Demonstration
Adrian Röfer (Ph.D. Student)
Everyday human environments are highly complex in both structure, and appearance. Nonetheless, we would like to move towards a world in which robots can assist and support humans in these environments. As a part of my PhD, I will move us one step along this path by making robotic manipulation more dexterous and adaptive. Through imitation learning skills will be acquired more quickly, including detailed tactile feedback will make manipulation more dexterous. Learning will also be used to reduce the complexity of long-horizon household tasks, by developing an "intuition" for how these tasks will unfold. Lastly, I aim to transfer the acquired skills among different robots and environments such that the skills can be considered to be gained not just locally, but globally. This transfer will be made possible by extracting recurring features, and summarizing representations from the skills.
|Primary Host:||Abhinav Valada (University of Freiburg)|
|Exchange Host:||Sethu Vijayakumar (University of Edinburgh & The Alan Turing Institute)|
|PhD Duration:||01 June 2021 - 31 May 2025|
|Exchange Duration:||01 June 2023 - 30 November 2023 - Ongoing|