Robot Guide Dogs Expected to Make Blind People's Lives More Convenient

Engineers from the Department of Computer Science at Binghamton University in New York have successfully developed a robotic guide dog for visually impaired individuals. This robotic dog is capable of responding to leash pulling.

Assistant Professor Shiqi Zhang, along with doctoral student David DeFazio and junior student Eisuke Hirota, have been working on the development of this robotic guide dog to improve accessibility for visually impaired individuals. They conducted a live demonstration where the robotic dog confidently and attentively led a person through the laboratory corridor.

Shiqi Zhang explained some of the reasons behind initiating this project.

He said, "We were surprised to learn that only a small percentage of the visually impaired and blind community have access to a real guide dog. We looked at the statistics and found that only 2% of people are able to achieve this goal."

One of the reasons for this shortage is the high cost of a real guide dog, which is around $50,000, and the two to three years of training required. Only about 50% of dogs are able to complete the training and serve visually impaired individuals. Robotic guide dogs offer significant advantages in terms of cost, efficiency, and accuracy.

This is one of the initial attempts to develop a robotic guide dog following advancements in quadruped technology and cost reduction. After approximately one year of effort, the team has developed a unique leash pulling function using reinforcement learning.

Shiqi Zhang stated, "After about 10 hours of training, these robots are able to navigate indoor environments, guide people, avoid obstacles, and detect leash pulling actions."

The leash pulling function allows users to pull the robot in a specific direction at intersections in the corridor, and the robot will turn accordingly. Despite showing significant potential, DeFazio mentioned that further research and development are needed before the technology is ready to handle specific environments.

He said, "Our next plan is to add natural language functionality. Ideally, I can have a conversation with the robot to seek assistance based on the situation." He added, "Additionally, intelligent disobedience is an important feature. For example, if I, as a visually impaired individual, tell the robot to walk into traffic, we want the robot to understand that. In that case, we should override human intentions. These are some of the future directions we are researching."

In order to obtain direct and valuable feedback from members of the visually impaired community, the team has been in contact with organizations for the blind. DeFazio believes that this will help guide their future research.

DeFazio said, "A few days ago, we had a conversation with a blind individual who mentioned that avoiding sudden obstacles is crucial. For example, if there is an uneven drain in front of you, it would be great to have a warning in advance, right?"

Although the team has not set any limitations on the application of the technology, their feedback and intuition lead them to believe that robots may be more useful than real guide dogs in specific environments. Due to their ability to store maps of difficult-to-navigate areas, robots have the potential for higher efficiency in guiding visually impaired individuals to their destinations.

Shiqi Zhang stated, "If everything goes smoothly, in the next few years, we can have these robotic guide dogs in shopping malls and airports. It would be as simple as people using shared bicycles on campus."

Although still in the early stages, the team sees this research as a promising step towards improving accessibility for the visually impaired community in public spaces.

The team will be presenting their research paper at the November Robotics Learning Conference.