Written by Matthew Busekroos | Produced by Andrew Zucosky

MIT CSAIL PhD candidate Dominique Regli is working in the lab’s Model-based Embedded and Robotic Systems (MERS) Group with Professor Brian Williams. Prior to CSAIL, Regli grew up in Philadelphia and received her Bachelor’s degree from Johns Hopkins University in Baltimore.

At the beginning of her senior year at Johns Hopkins, Regli was torn between going into academia or pursuing graduate school. She knew she was interested in human-robot interaction and its applications to healthcare, but did not know how that fit into the context of graduate studies. Through completing her graduate applications – reading about current lab research and learning what these groups’ interests – she came to understand the forefront of current research and how her interests could align.

In the admissions process, Regli had many engaging and thoughtful conversations with different professors and students at MIT. Ultimately, she found Prof. Williams and the Model-Based Embedded and Robotics Systems (MERS) Group, and had a gut feeling that this was the right place for her.

The MERS Group at CSAIL does research in Cognitive Robotics, aiming to make robots that think, plan, and act like humans. According to Regli, leveraging human intuition in this manner means that their research is at an intersection of fields including computer science, cognitive science, autonomy, and more. At the same time, the group does scenario driven research, grounding their theoretical work in real-world application domains.

“As a first-year student, Prof. Williams has helped me really explore this field and the breadth of algorithms and methods the group has been designing over the past few decades,” Regli said. “The exciting thing for me is that my application domain of interest–healthcare–has been relatively unexplored in our lab, so the goal is to determine how methods we have built for areas like deep space exploration or autonomous driving can be iterated upon to be useful in a new setting. My lab mates have also been great to bounce ideas off of and to help learn more of the intricacies of our work.”

Regli said the goal of her research is to design human-activity recognition systems which can support eldercare and senior living communities.

“The world is currently experiencing a global aging phenomenon in which the proportion of people of the age of 65 is rapidly increasing, and we do not have the nursing support resources to appropriately respond to this,” Regli said. “Furthermore, many older adults want the ability to live independently in their homes. However, unique challenges and safety concerns arise for seniors in this setting. There is increased danger associated with emergency episodes such as a sudden fall without access to immediate help, and prolonged, unhealthy illnesses are more likely to go undetected without routine nursing care. Smart systems in the home which can recognize human activity would be able to mitigate these issues and give seniors the freedom to live independently.”

Regli said human activity recognition (HAR) is a challenging task for a few reasons. First, individuals want privacy in their homes. While camera and audio-based methods have been able to do HAR in the past, these types of sensing technologies are generally deemed too invasive for home environments. She said the first step of her project is to understand what data is both accessible and useful to solving the problem.

“Additionally, humans tend to be hard to track because we are all unique and do things differently from each other,” she said. “Many people have tried machine learning based approaches to the HAR problem, but it is hard to collect enough training data to build robust models since getting people to record and label activities they do in their homes is a nontrivial task. Alternatively, model based approaches rely on significantly less training data, but require more domain knowledge to design. I am looking to combine machine learning and model based systems, leveraging the strengths of both, to perform robust HAR.”

Regli is passionate about eldercare support and senior living because it affects everyone.

“We all naturally want independence, comfort, and wellbeing in our later years,” she said. “This field is also just beginning to gain more attention within the robotics community, so it feels like an important and exciting time to be engaging in the work. From a methodology standpoint, I love working at the intersection of engineering fields like computer science and autonomy, as well as humanities fields like cognitive science and philosophy. I not only get to think about how to design robots, but I get to explore how robots and humans are alike and not, and the implications of this advancing technology.”

As a first-year student, Regli said she is still uncertain about what she wants to do once she graduates, but hopes to either work in academia or in healthcare.

“My dream job is really just a role where I get the opportunity to help others and make a positive impact,” Regli said.