Research

My research develops machine learning and AI for versatile systems that are deployed for extended durations in complex dynamic environments. This research combines and integrates techniques from continual or lifelong learning, transfer and multi-task learning, deep learning, reinforcement learning, multimodal learning, and collaborative AI. I apply these techniques to problems in autonomous service robotics and precision medicine.

Current Projects (as of 2025)

Autonomous Search-and-Triage Robots for the DARPA Triage Challenge

I’m leading Penn’s Team PRONTO, which is competing in the Triage Challenge—the latest in the DARPA robotics challenge series. We're developing an autonomous system of aerial and ground robots to locate victims and characterize their injuries and vital signs in (simulated) mass-casualty scenarios.

Compositional Continual Learning for Open Worlds

This long-term project builds upon our 13-year history of pioneering work in continual and lifelong learning. We're developing the next generation of continual learning algorithms that acquire and reuse modular skills to solve and zero-shot long-horizon tasks in open-world settings.

Robust and Replicable Reinforcement Learning

Despite the success of deep RL in recent years, its performance is still highly dependent on proper hyperparameter tuning and careful implementation choices. To address these issues and make RL work reliably out-of-the-box, we are studying mechanisms to improve the stability, mitigate value overestimation, and enable the replicability of RL algorithms.

Ambient AI for Precision Medicine

We're reimagining what patient care would look like if the clinic were instrumented with modern robotic sensors, enabling real-time clinical decision support. Throughout the encounter, and even before the provider entered the room, multimodal AI models would quantify the patient’s physical, cognitive, and emotional health, augmenting the clinical exam, and providing for quantitative longitudinal assessment. We're also exploring related approaches for spatio-temporal clinical understanding of surgery and in trauma bays.

Multimodal Clinical Foundation Models

Multimodal AI (MAI) has enormous potential to provide useful tools for AI-integrated healthcare. To power clinical decision making, under NIH funding, we're developing multimodal foundation models that integrate images, video, clinical imaging (CT, MRI, ultrasound), electronic health records, biospecimin, and genetics data. This project combines state-of-the-art techniques for multimodal learning and agentic architectures with vast amounts of clinical data and novel methods to enable ethical clinical reasoning within the MAI inference process.

Autonomous Mobile Service Robots

We're developing autonomous service robots that can operate continually in university, office, and home environments. These robots will learn a wide variety of skills involving perception, navigation, control, and multi-robot coordination over their lifetime, serving as a major testbed for continual lifelong machine learning algorithms and mobile manipulation.

Learning Affordances from Visual Observation

Our recent advances in visual understanding and physics simulation are enabling robots to learn how objects afford interaction directly from visual data.

AI Education

I am a leader in the development of AI curricula at both the national (through CRA) and international (ACM/IEEE/AAAI CS 2023 curriculum) levels.

Older Projects

ELLA Family of Lifelong Learning Algorithms

My research group has a 13-year history of pioneering work in lifelong (continual) learning. Our earliest work developed the first general-purpose framework for lifelong learning (ELLA), with subsequent work developing a family of lifelong learning algorithms around this framework. These methods include lifelong RL, cross-domain lifelong learning (best paper nominee), and zero-shot learning using task descriptors (best paper award), among others.

Interactive AI

My research on interactive AI methods seeks to give users extensive control over reasoningand learning processes. In many critical applications, especially in military and medical domains, users will reject traditional AI automation without the ability for each result to be checked and altered by a human operator. Interactive AI methods incorporate such levels of user control to facilitate the transition of AI into these types of applications. This interactive AI paradigm combines user-driven control with the complementary system-driven approach of active learning.

Selective Knowledge Transfer

My dissertation research focused on the problem of source selection in transfer learning: given a set of previously learned source tasks, how can we select the knowledge to transfer in order to best improve learning on a new target task? Until my dissertation, the problem of source knowledge selection for transfer learning had received little attention, despite its importance to the development of robust transfer learning algorithms.