Dec. 13, 2021 -- A new robot that can spot two or more people standing within 6 feet of each other could play an important role in maintaining social distancing requirements during the current or a future pandemic.
The robot located, approached, and notified groups of two to six people who were standing too near each other to move apart in early testing.
The COVID Surveillance robot (CS-robot) discreetly uses an on-screen message to alert people who are too close together. The technology, developed by University of Maryland researchers, also features a thermal camera that can help detect someone hotter than others who might have a fever. The robot can then notify health or security personnel.
The study was published Dec. 1 in PLOS One.
An Element of Surprise?
How the public might react to a robot referee remains unknown. "We have mostly tested in our lab and buildings and a few public events," said senior study author Dinesh Manocha, PhD.
"Many times, humans are surprised when they see a robot moving around them or displaying such messages. Such kind of robots have not been deployed widely, so it is hard to guess public perception to use of such robots," said Manocha, a professor of computer science and electrical and computer engineering at the University of Maryland in College Park.
There are many things to consider about how people would react, said Bruce Hirsch, MD.
"I'm interested in what the psychology would be, what the sociology would be -- what does it mean to be reprimanded by a robot?" he said.
"There is already depersonalization and burnout in the health care setting, particularly with the COVID epidemic. I don't know if a robot would be perceived as helpful and supportive," said Hirsch, an infectious disease doctor at Northwell Health in Manhasset, NY.
Manocha, first author Adarsh Jagan Sathyamoorthy, and their colleagues challenged the autonomous robot in five scenarios in an indoor space that measured 4 meters by 4 meters (about 12 feet by 12 feet). They tested how well it found people, detected social breaches, and tracked people walking or standing still.
The robot also performed in non-laboratory indoor settings such as lobbies and narrow corridors with different levels of lighting, the researchers noted.
The study also compared performance of the robot alone to the robot plus input from wall-mounted closed-circuit television (CCTV) cameras. Although the robot could detect and alert people who weren’t complying on its own, use of CCTV data improved its performance in every situation tested.
On the Move
The robot can cover any blind spots in CCTV coverage through a "lawnmower strategy," where it moves to fixed locations and surveys its surroundings.
The robot uses its cameras and sensors to "lock" onto a person who isn’t social distancing to focus its movement. In a series of tests, the CS-robot's location precision was within 0.3 meters, regardless of whether the person was walking or standing still.
Interestingly, the robot also considers time in addition to proximity. The engineers wanted to avoid flagging lower-risk situations, such as when two people pass within 6 feet for a moment in a hallway.
The engineers also designed the robot to get around potential obstacles, including other people. Specifically, their Frozone collision avoidance strategy, outlined in a previous study, allows the robot's “red green blue-depth” (RGB-D) camera to track and predict where people will move soon.
The robot prioritizes its actions to attend to larger groups first and/or follow people in motion based on their movement relative to the robot.
People Remain Anonymous
Being approached or followed by a robot equipped with cameras could raise concerns about privacy.
The researchers accounted for this with standard de-identification technology such as visual image redaction for faces, gestures, and gait data. And the visual camera on the robot assigns a random number as an identifier for each person in a group.
The robot also reduces risks for humans. "Monitoring people’s temperatures remotely reduces the risk of the security/health care personnel contracting the coronavirus," the authors noted.
Still a Role for Humans
Advantages of the robot, Hirsch said, included freeing up personnel from watching people. The technology could also make monitoring more robust.
But, he said, "I'm older and I like people."
Humans are still more likely to help with education -- explaining why social distancing is important -- as well as "reading certain situations," Hirsch said.
He offered the example of an upset, crying person in a hospital who just lost a loved one.
"Is a robot going to approach that individual in a potentially sensitive moment and ask them to move 6 feet away from another individual or turn them in for additional health care monitoring?"
Although only tested indoors so far, "we expect that robots can be used in indoor or outdoor settings where humans come in close proximity," Manocha said.
The researchers might also try other ways, besides a screen message, to alert people who aren’t distancing, such as sounds or other warning signals.
Manocha noted the robot does not distinguish between strangers and people from the same household. Also, the authors said, "We need to develop better human-robot interaction approaches."
And the investigators want to study the social impact of the robots.
"We would also like to develop methods for detecting if the people in the robot’s surroundings are wearing masks," they noted.