Short Bio

Hi, my name is Jefimija Borojevic, and I am a PhD student within the SWEET project. I completed my bachelor’s degree in Mechatronics at the Faculty of Technical Sciences, University of Novi Sad, Serbia. And, I pursued the Erasmus Mundus double-degree master’s program JEMARO (Japan–Europe Master on Advanced Robotics), studying at Centrale Nantes in France and Keio University in Japan.
During my master’s studies, I worked on implementing Active Inference, a neuroscience theory, within a physical human-robot interaction scenario. My research interests lie in exploring learning-based methods for improving interaction between humans and robots, which, I believe, will become increasingly present in everyday life.

Topic

To propose a new framework to learn and combine the preferences of user and bystanders in a service scenario. An offline phase will be used to bootstrap the preferences and compute the initial policies and the objective function using an Inverse Reinforcement Learning approach. In the online phase, the system will use the new experiences to refine the models. The framework will include an elicitation system to start interactions and eventually actively ask for missing information about the potentially incompatible preferences of the user.

Host Institution: Agencia Estatal Consejo Superior de Investigaciones Científicas – Barcelona, Spain

PhD Enrollment: Universitat Politècnica de Catalunya

PI: Dr. Guillem Alenyà