Aleksandra (Ola) Kalinowska
Current Institution: Northwestern University
Email: ola@u.northwestern.edu
Bio: Ola is a PhD candidate in the Interactive & Emergent Autonomy Lab in the Center for Robotics and Biosystems at Northwestern University. As part of her research on assistive and rehabilitation robotics, she collaborates with Ekso Bionics, the Northwestern Department of Physical Therapy and Human Movement Sciences, and the ShirleyRyan Ability Lab. During her PhD, she has done research internships at MERL and Google DeepMind. She is an adjunct lecturer for the NUvention Medical course on medical innovation. In 2016, she got a Bachelors of Science in engineering from MIT.
Abstract: Collaborative Co-adaptation During Human-Robot Interaction
There is often an assumption that robots are meant to be tools that humans learn to use to more efficiently or more effectively accomplish a set task. Instead, human-robot interaction should be a co-adaptation process with both humans and robots adapting to each other to learn to collaboratively achieve a task. In multi-agent settings, interaction emergence is studied quite extensively, because there is little cost to exploring unstructured collaborative behavior in a simulated environment. During physical human-robot interaction, it is nontrivial to enable learning, because (1) we prioritize the safety of the human partner and (2) we work with sample sizes constrained by the bandwidth of real-time interactions. In my research, I study how to algorithmically structure the interaction to encourage active co-adaptation while maintaining safety and achieving learning on a timescale practical for a human partner.