Robotics Expert Builds Bots to Be Caregivers

They are designed for people with Autism, the elderly, and others who could use a helping hand

6 October 2015

IEEE Fellow Maja Matarić imagines that within a decade robots will not only help care for stroke patients, the elderly, and the disabled in their homes but also engage with children who have autism or teens who suffer from anxiety.

A professor of computer science, neuroscience, and pediatrics at the University of Southern California in Los Angeles, Matarić runs the school’s Interaction Lab, where she works to develop smart, socially adept robots to assist people with physical, cognitive, and social needs. These robots could potentially provide in-home therapy that would complement the shortage of human caregivers.

“The United States in particular has a high rate of cognitive, social, and physical disorders coupled with a lack of affordable and available care,” she says. “Socially assistive robots could help fill this gap.”


Matarić is developing autonomous robots that use body language, gestures, and facial expressions of their own to motivate people to perform specific physical, social, or cognitive behaviors. She develops her helper bots on new or existing robotics platforms by programming them to interact with people in various ways. The platforms include the 0.6-meter-tall Nao humanoid robot from Aldebaran Robotics and Bandit, a robot with an expressive, humanlike face from BlueSky Robotics. The crux of Matarić’s work is developing the clever software that drives them.

“It’s the algorithms that enable the robot to react to and motivate people to perform desired actions,” she says.

The robots could, for example, demonstrate physical therapy and rehabilitation exercises to an elderly person or stroke patient and then offer feedback and praise to aid with motivation. Or they could develop an autistic child’s social interaction skills by making eye contact or playing a game. Her studies have shown that in the presence of robots, many children with autism—who generally have communication difficulties—become more social and vocal and begin to interact first with the robots, then with their adult caregivers.

In addition to relying on complex machine vision, signal processing, and machine learning techniques, Matarić’s algorithms draw on models and theories in the neuro, social, and cognitive sciences.

Matarić’s team is programming the bots to have ever more sophisticated skills. For example, they are creating algorithms to interpret a person’s expressions, actions, and changes in voice pitch so a robot can make an intelligent decision and be more responsive to moods. For example, if a patient backs away as the robot nears, the robot would be programmed to, say, keep a certain distance or orient its body in a non-threatening way.

In a 2014 study, the group showed that by programming an occupational therapy process called graded cueing, the robot could improve a person’s skills at common tasks by giving increasingly specific prompts. This could, for example, help adults with disabilities or brain injuries relearn how to use their arms and hands. Or it could help kids with autism improve their ability to copy others by playing an imitation game with the robot; learning to imitate even a robot is a useful social skill for such children.

For stroke patients, her hands-off, fully autonomous robots act as physical therapists, prompting and guiding patients to perform actions such as reaching for an object and providing feedback and encouragement as they go along. Matarić’s group has found that a patient’s performance improves over time as the robot responds verbally or by shaking or nodding its head.

She now has a project to develop relatable, empathetic robots to reduce anxiety in teens and help them cope with depression and being bullied. The group is looking into metrics to evaluate anxiety and is researching robot capabilities that could help diminish the condition. Matarić will also be working with mechanical engineers at the University of Pennsylvania, in Philadelphia, to develop a modular robotics platform with which other engineers could develop socially assistive robots of their own.


Matarić moved to the United States from Yugoslavia when she was 16 years old. Her uncle, an engineer, persuaded her to study computers “because they’re the future,” he told her. She found she had an aptitude for computer science and, what’s more, she loved it. “I really like the structure and logic of programming. It’s very appealing,” she says.

She earned a bachelor’s degree in computer science from the University of Kansas, in Lawrence, in 1987, and received her master’s degree and Ph.D. from MIT in 1990 and 1994, respectively. She taught computer science at Brandeis University, in Waltham, Mass., before joining USC in 1997.

As a graduate student at MIT, Matarić built robots that navigated independently around obstacles and worked together as a group. Over the last 15 years, she has become intrigued by human-machine interaction—specifically, how interactions with robots can motivate people with special needs.

Matarić is also committed to getting preuniversity students, particularly girls, interested in engineering. As her school’s vice dean of research, she leads several outreach efforts, such as developing robotics curricula for elementary through high schools and mentoring programs for teachers and students. She has also been instrumental in developing The Next MacGyver competition, a contest to develop TV show scripts that feature a woman engineer as the main character.

“It’s still difficult to get girls into engineering,” she says, “We need to do a much better job of making them know why engineering is relevant and exciting.”

Learn More