For delivery robots, not all sidewalks are created equal - some are uneven or clogged with people and bus shelters - so researchers at Cornell Tech developed a "robotability score" and rated every street in New York City on how hospitable it would be to robots.
Their rating system is the first of its kind, researchers said, and may help urban planners and robotics companies plan for future robot deployments that won't disrupt existing sidewalk environments.
"I don't know that everybody wants robots in their neighborhood, but if they do, the robotability score can help them think about, what are the features that we've built in to help welcome robots?" said senior author Wendy Ju, associate professor of information science at Cornell Tech, the Cornell Ann S. Bowers College of Computing and Information Science and the Jacobs Technion-Cornell Institute, and the multicollege Department of Design Tech. "We're just trying to make those things a little bit more visible."
Matt Franchi, a doctoral student in the field of computer science, and Maria Teresa Parreira, a doctoral student in the field of information science, will present the study, "The Robotability Score: Enabling Harmonious Robot Navigation on Urban Streets," at the Association of Computing Machinery's conference on Human Factors in Computing Systems (CHI), April 28 in Yokohama, Japan.
Delivery robots have already hit the streets in Los Angeles and are roaming some college campuses and airports. But most communities are currently unwelcoming places for robots.
Inspired by neighborhood walkability and accessibility scores, the researchers developed the robotability score to compile multiple features affecting robot navigation into a single number.
"People love convenience," Franchi said. "We think this tool will be readily useful as people begin to envision deploying sidewalk robots in urban spaces."
To develop the robotability score, researchers first interviewed 10 experts in urban planning, robot navigation and accessibility, from academia and industry, to decide which features to include. The team then used an online survey of additional experts to determine how to weight the importance of each feature to the final score.
The robotability score includes 24 features, though the researchers used 19 in their analysis of New York City. Six features - pedestrian density, crowd dynamics, pedestrian flow, sidewalk quality, street width and density of street furniture - made up almost half of the score.
New York City is the perfect location to develop such a score, researchers said, because of the city-wide data available through the NYC OpenData site. This database holds information such as sidewalk width and condition, and the locations of bus shelters, bikes lanes and newsstands. The researchers also used about 8 million dashcam images collected around the city in late 2023, to estimate vehicle, bicycle and pedestrian traffic.
When applied across the city, areas with highest robotability were 4.3 times more "robotable" than areas with the lowest score. Franchi developed an interactive map, where users can see how different locations compare.
The team tested the accuracy of their scores by operating a trashbot - a garbage pail on top of recycled hoverboard parts that is operated with a joystick - down sidewalks at rush hour in two spots with low and high robotability in Queens and Manhattan.
"Rush hour magnified the scores," said co-author Frank Bu, a doctoral student in the field of computer science. Even at peak travel times, the trashbot had no problem rolling along on empty sidewalks in high robotability areas. "Other places with bins, high pedestrian flow, traffic flow and even with food trucks or shop stands on the side of the sidewalks - that intensified the scene," Bu said.
There were additional features the researchers wanted to include, but the data was not yet available, such as local attitudes toward robots. Future versions could include these features, and possibly dynamic features, like real-time weather reports and pedestrian traffic.
Just as the existence of walkability and accessibility scores have influenced developers and urban planners to enhance these qualities in neighborhoods, the team hopes the robotability score will do the same - but only in communities that want robots.
"We're barely starting to build cities for people - we're not trying to claim that cities should be built for robots," Parreira said. "Just seeing the urban environment through the lens of a robot was our goal."
This work was made possible through funding from the National Science Foundation and the Digital Life Initiative at Cornell Tech.
Patricia Waldron is a writer for the Cornell Ann S. Bowers College of Computing and Information Science.
