Researchers at Binghamton University, State University of New York have developed a talking robot guide dog system using large language models, including GPT-4 with voice commands. The system determines an ideal route, safely guides users, and offers real-time feedback. Shiqi Zhang, an associate professor at the Thomas J. Watson College of Engineering and Applied Science’s School of Computing, noted that real dogs can understand around 20 commands at best, while robotic guide dogs have stronger language capabilities. The robot provides plan verbalisation before departure and scene verbalisation during travel, offering information about routes and surroundings, which is important for visually impaired or blind people due to limited situational awareness. To test the system, the team recruited seven legally blind participants to navigate a large, multi-room office environment. The robot asked users where they wanted to go, presented possible routes and travel times, and guided them to a conference room while verbalising obstacles. Following the test, users completed a questionnaire rating helpfulness, ease of communication, and usefulness; a combined approach with planning explanations and real-time narration was preferred. A simulated study also showed success. The team plans to conduct more user studies, increase autonomy, and have robots navigate longer distances indoors and outdoors. Study participants were enthusiastic about the technology’s potential.