Robotics Research Lab
CRES
USC Computer Science
USC Engineering
USC
/ Research / Areas / Long-Term Adaptation

We are interested in developing and enhancing methods that allow robots to adapt to specific users and personalize their interaction in order to provide effective, sustained long-term human-robot interaction. Interaction with people is a challenging learning problem, as it involves a "moving target", since people's moods, health state, task expertise, and other factors are not necessarily accessible, change at various time-scales, and can influence behavior in unexpected ways. People typically do not have the patience to provide ample training examples, they do not always act consistently, and they require intuitive interfaces. Our work is driven by several motivations: 1) to leverage human-robot interaction in order to enhance the robot's ability to learn new concepts, skills, and tasks; 2) to enable the robot to personalize its interaction with the user so as to optimize its effectiveness in assisting the user; 3) to develop methods for learning at human timescales, which are both very short (within a social interaction) and very long (repeated encounters/sessions, months, and even years); and 4) to situate HRI learning in realistic domains, including classrooms, health care contexts, and homes. Our ongoing projects are exploring learning and adaptation in a variety of problem domains, including human-robot team collaboration, user adaptation, task learning from demonstration, and socially guided service task learning.

Current Projects
Past Projects