Doctoral Project: Online Learning of Constrained Manipulation from Corrective and Explorative Demonstrations
Description of the research project
See below (same as doctoral project because there is only one doctoral project in this project)
Description of the doctoral project
Contact-rich manipulation tasks of articulated objects, such as the opening and closing of door latches, still pose significant challenges for the state of the art in robot manipulation. This project will extract manipulation strategies from human demonstrations through kinesthetic teaching. This project will develop an innovative approach to Learning from Demonstration that learns robust policies from few, yet diverse demonstrations. The resulting data will give rise to general manipulation strategies based on techniques from machine learning. Effectively, the acquired data acts as a prior for the (deep) learning methods employed for policy generation. The goal is to produce a means of programming a robot system by merging data and learning to produce robust, complex, and general contact-rich manipulation.
Applications should include: motivation letter, curriculum vitae, transcripts of records (for both BSc and MSc + doctoral degree if applicaple), copies of degree certificates (BSc, MSc), abstracts of Bachelor-, Master-thesis, e.g. doctoral thesis, list of publications and one selected manuscript (if applicable), two names of qualified persons who are willing to provide references, and any documents candidates feel may help us assess their competence.
To ensure equal opportunities between women and men, applications by women with the required qualifications are explicitly desired. Qualified individuals with disabilities will be favored. Applications are also expressly welcomed from suitably qualified persons seeking to be entered as “gender diverse” in the public register. The TU Berlin values the diversity of its members and is committed to the goals of equal opportunities.
Please send copies only. Original documents will not be returned.
- MSc degree in computer science or similar field
- Research experience in robotics, machine learning, computer vision, and/or control
- Experience in Learning from Demonstration and force control desirable
- Experience in applying (deep) learning to control problems desirable
- Interest in interdisciplinary research in the context of the Center of Excellence “Science of Intelligence”
- Excellent software engineering and programming skills in C++
- Excellent English writing and communication skills