Sensor and AI Innovations from NC State ECE and CS Helping to Build Better Working Dogs

A recently published Science magazine article highlights the urgent need for new scientific and technological solutions that enhance working dog and guide dog training and selection. These solutions are being developed at NC State’s College of Engineering.

A recently published Science magazine article highlights the urgent need for new scientific and technological solutions that enhance working dog and guide dog training and selection. In particular, the story emphasizes how sensor and data-driven approaches can help address the global shortage of highly trained service animals. This coverage explores the limitations of traditional, largely subjective behavioral assessments for dogs, and points to emerging research that uses objective measurements and analytics to better understand canine behavior and performance.

... these objective assessments of trainee guide dogs help support the early identification of temperament traits. Essentially: better data, better training, better dogs.

The article features images of sensors developed at NC State University, technology that is currently in use at Guiding Eyes for the Blind, one of the leading guide dog schools in the United States. Research teams led by Alper Bozkurt, ECE Distinguished Professor, and David Roberts, CS associate professor, are at the forefront of these innovations. Their interdisciplinary work has led to the development of wearable sensor systems and AI algorithms that quantify canine motion, physiological and environmental signals; these objective assessments of trainee guide dogs help support the early identification of temperament traits. Essentially: better data, better training, better dogs.

The technological advances developed at NC State underpin the research shared in the Science article. The traditional method of training guide dogs is dependent on experts and behavior scoring. This creates a “supply chain” issue, where the number of trained dogs is limited to the number of expert, professional guide dog trainers working in the industry. This expert-dependent behavior scoring can be supplemented or even replaced by quantitative, replicable, sensor-based evaluation frameworks. These quantitative measures are especially valuable, because they are not person-dependent, and can be scaled to address global needs. By integrating wearable hardware, embedded sensing and machine learning, the research being done at NC State doesn’t just support improved guide dog training outcomes, it advances the broader innovation landscape in working dog science.

For more about this initiative, see: Transforming Guide Dog Training with AI and Sensor Technology.

To read the full Science article, see: Grimm, D. (2026, February 12). “Can science build a better working dog?” Science. https://www.science.org/content/article/can-science-build-better-working-dog