IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Binghamton University Researchers Build Robotic Guide Dogs

Associate Professor Shiqi Zhang and two of his students say the cost, efficiency and accessibility of artificial intelligence-powered seeing-eye assistants could improve quality of life for the hearing-impaired.

A black robot dog walking on a sidewalk in the foreground with a student walking next to it in the background.
A robot guide-dog for the blind walks across campus at Binghamton University.
Photo by Ryan Yarosh
“Bearcat” will someday help a blind person cross the street safely and get to appointments on time. But unlike its canine counterparts, this four-legged friend won’t require food, affection or a pooper scooper.

For the time being, the guide-dog robot under development at Binghamton University in upstate New York knows its way around the campus engineering building and responds to tugs on the leash to stop or change direction. The bear cat is actually Binghamton University’s mascot, so at some point this AI-powered being will get a more dog-like name to respond to when its natural language functions are in place.

“That’s what we’re working on next,” Shiqi Zhang, an associate computer science professor at the school, said Tuesday. “And it won’t just know the 20 or so commands that the best dogs can learn. It will know the entire English language and actually speak it.”

Zhang and two students from the school’s computer science graduate and undergraduate departments began work on the robot in the 2022-2023 academic year. It debuted last Halloween by handing out candy to campus trick-or-treaters. At that time it was operated by remote control, the associate professor said, but when the project concludes the robot will be completely autonomous.

The team presented a paper on their creation Nov. 8 during the Conference on Robotic Learning in Atlanta. Zhang is seeking grants for continued research and development. He is also collaborating with the National Federation of the Blind’s chapter in Syracuse.

David DeFazio, a Binghamton University Ph.D. candidate who is assisting Zhang on this project, said specific input from the visually impaired will play a key role in their research.

“The other day we were speaking to a blind person, and she was mentioning how it is really important that you don’t want sudden drop-offs,” DeFazio said in a recent news release. “For example, if there’s an uneven drain in front of you, it would be great if you could be warned about that, right?”

After about 10 hours of training, Zhang said, the robot dog could recognize corridors, avoid obstacles and manipulate its machine learning capabilities to generate maps and remember places it had visited. It has also learned from mistakes, such as failing to stop at a doorway and proceeding to go left or right without first receiving a tug on the leash from its handler.

This technology will have to be trained in busy public places like malls and airports before it can be trusted to watch out for a human being, Zhang explained. “Intelligent disobedience,” where the robot dog knows when and when not to disregard human instructions, is also critically important. An example of this would be if the visually impaired handler tells it to walk into traffic.

Zhang said the advantages of seeing-eye robots over trained guide dogs are numerous. Dogs are unlikely to outlive their owners, and the costs for properly trained canines can exceed $50,000. Moreover, he said, only about half the number of dogs who enter guide training programs to serve the visually impaired end up graduating — and none of them can speak to their handlers or read a map.

Zhang said he can imagine these robots being available and shared among those who need them on a large scale, much like public rental bikes in urban areas.

“There’s a significant shortage of guide dogs,” he said. “With this, we can bring real change to their [visually impaired people] quality of life.”
Aaron Gifford is a former staff writer for the Center for Digital Education.