Matarić has published extensively and is highly cited across the fields of
robotics,
artificial intelligence (AI), and
machine learning.
Behavior-Based Control and Learning Beginning with her graduate work at MIT, Matarić made fundamental contributions to the understanding of autonomous robot cognition and interaction. For her master's degree, she was the first to demonstrate that
behavior-based systems (BBS) could be endowed with representation and thus have the expressive power to plan and learn. Her well-known robot system, Toto, was the first behavior-based system to learn maps online and optimize its behavior. It is highly cited and remains one of the milestones in BBS.
Multi-Robot Coordination: Distributed Robot Teams and Swarms In her PhD work, Matarić was one of the first to work on decentralized, distributed algorithms for robot teams and
robot swarms that leveraged scalable local control. She enabled a team of 20 physical robots to interact and cooperate on tasks such as coordinated exploration, foraging, organizing in formations, and homing. Prior to her PhD work, nearly all research in robotics was restricted to single robots or pairs. She made pioneering contributions to the theory and practice of multi-robot coordination by showing that complex behaviors could be composed of basis behaviors in a principled way, bringing rigor to the then nascent discipline of distributed robotics. Her work provided the first formal analysis of existing multi-robot coordination approaches, elucidating formal and practical limitations, then addressed those limitations by contributing provably correct yet scalable task allocation algorithms for multi-robot control. Her research group developed efficient principled market-based strategies in physical real-world validated multi-robot systems performing a variety of tasks including object transport, area cleanup, and reconnaissance. Her group's work also demonstrated analytical methods for automatically generating minimalist multi-robot controllers with provable properties. Finally, Matarić's research lab, the Interaction Lab, demonstrated both theoretically and experimentally the viability of online real-time learning in distributed multi-robot systems. Her Interaction Lab developed algorithms for model learning within a team, learning by imitation, and learning through human-robot interaction. She was a pioneer and an established leader in multi-robot coordination, which is now a large and thriving area of robotics.
Socially Assistive Robotics Matarić's research since 2000 has been in the new field of
socially assistive robotics (SAR), which she pioneered (with Prof. Brian Scassellati). cognitive and social skill training for children with
autism spectrum disorders, cognitive and movement exercises for healthy elderly users and those with
Alzheimer's Disease, attention support for studying for students with
ADHD, and personalized therapy for users with anxiety and/or depression. This work is conducted through some of the longest studies and data collections in real-world, challenging environments such as schools, therapy centers, rehabilitation clinics, nursing homes, and private homes. Mataric's work has the potential for major impact on the way health care is delivered to large populations, and on how affordable and accessible care, education, and training can become, through human-centered use of AI and other technologies. By focusing on human augmentation rather than automation, her research has promising implications on the future of work, with broad-reaching interdisciplinary impact.
AI and Machine Learning Mataric's entire research career has been in AI and
Machine Learning (ML), from her MS thesis on learning spatial representations,
learning by imitation, and learning through
human-robot interaction, to the work on understanding human activity for human-robot interaction. Her work in socially assistive systems has focused on developing personalized diagnostic and assistive and therapeutic interactions in a variety of domains (early child development, autism therapy, pain management,
cerebral palsy therapy, ADHD and anxiety support, stroke rehabilitation, and dementia detection and support) by studying and developing models for understanding and supporting key behavioral capabilities and predictors, including personality, engagement, and motivation. Her work has spanned deep learning & small and large models, continuing to contribute to the rapidly evolving AI landscape. the Scientific Advisory Board of the
Max Planck Institute for Intelligent Systems, the
DARPA Information Science and Technology (ISAT) Study Group, the US Scientific Advisory Council,
Nature and
Scientific American (Springer), and the AAAS Leshner Leadership Institute, among others. == Mentoring ==