Research Overview
Our research at the PaInt Lab focuses on developing AI systems that seamlessly integrate with and support people in their dynamic environments over the long term. We work at the intersection of robotics, machine learning, human-robot interaction, and cognitive science to create adaptive intelligent systems that can learn from and collaborate with human users.
Our key research areas include:
Personalized Active Learning
Most active learning (AL) methods rely on uncertainty sampling to gather new most informative information in an environment. As real-world environments can contain information unrelated to the tasks that the autonomous system needs to perform, uncertainty sampling can lead to useless information learned by the system through limited queries that could be used to learn task-related information. This can also negatively affect users' trust in the system.
We develop AL methods that can choose queries that will generate the most important information related to the task that the user needs the autonomous system to perform. A major challenge in developing such AL methods is to determine if unknown data in the environment is related to the task without querying the user about the data.
We are developing novel similarity metrics and leveraging LLMs to determine if an unknown data sample is related to the task. However, this led to another challenge: this objective of determining and choosing information most similar to the task is opposite to uncertainty sampling to find the most dissimilar samples compared to the previously learned knowledge. We investigate hierarchical similarity metrics to find similar information on the task level, but dissimilar information on the lower sample-level knowledge.
Continual Learning in Human-Robot Interaction
Robots that can learn continuously from human teaching over extended periods present unique challenges at the intersection of machine learning and human-robot interaction. Our research explores how humans perceive and interact with robots that can continually learn and adapt, and how to design effective teaching and interaction mechanisms.
Our work examines several key questions in this domain:
- How do humans perceive robots that continuously improve over multiple interactions?
- What teaching strategies do people naturally adopt when working with learning robots?
- How does trust evolve in human-robot relationships when robots demonstrate continual learning capabilities?
- What interface and feedback mechanisms best support human teaching of robots?
Through a combination of user studies, algorithm development, and prototype implementations, we're working to create robots that can build meaningful, long-term relationships with their human partners through ongoing learning and adaptation.
Personalized Autonomous Systems
Autonomous systems like self-driving cars and home robots need to adapt to individual user preferences and needs. Our research explores personalization methods that allow autonomous systems to learn user-specific behaviors while maintaining safety and reliability.
This research area focuses on:
- Developing algorithms for personalized navigation that adapt to user comfort preferences
- Creating models for human-centered decision making in autonomous systems
- Designing interaction paradigms that allow non-expert users to teach autonomous systems
- Implementing privacy-preserving personalization techniques
Our goal is to create autonomous systems that can build user models over time through natural interactions, adapting their behaviors to best serve individual users while maintaining appropriate safety constraints.