A Smarter Collar
Wearable data collection systems being developed in the College are empowering guide dog trainers and making a costly, time-consuming process more efficient
Posted on Wednesday, May 13, 2020 | Filed Under: Research
Lorraine Trapani’s husband, Michael, was a hard-charging, happy person. So, when he became partially blind, he worked to live with it, learning a scanning technique that made the most of the vision that remained.
“He said the hardest thing was not being able to see my face,” Lorraine Trapani remembers.
Seventeen years after his death, Trapani honors her husband’s memory as a volunteer puppy raiser, providing a loving home and teaching foundational skills to puppies who will one day provide their recipients with independence, safety and companionship.
“I want to raise an army of guide dogs in honor of Michael,” said Trapani, who was paired with her seventh puppy trainee, Jackie, in 2019.
As a volunteer for the nonprofit guide dog school Guiding Eyes for the Blind, Trapani will take a puppy for about 16 months. During that time, she’ll keep the animal with her at all times, exposing her to different social situations and constantly observing how she reacts.
Not every puppy is cut out for this type of work. The best guide dogs can make hundreds of snap decisions a day while remaining calm in environments that can be noisy and stressful. Guiding Eyes can invest $50,000 and two years in a dog, whether the animal is ultimately trained and placed with a blind person or not. Fifty to 75 percent of puppies that start the training don’t complete it, so the earlier they can determine whether a dog is cut out for the work, the more resources can be put toward training the right dogs and placing them in a position to help.
Technology developed at NC State, combined with Watson artificial intelligence applied to data collected and stored on the IBM Cloud, has led to a smart collar that gathers valuable insights during Jackie’s training period that will help determine if life as a guide dog is the right path.
It’s a tool that will help improve the training process, giving volunteers like Lorraine Trapani a chance to train that army of dogs in memory of her husband.
A Mutually Beneficial Relationship
Candidate dogs begin evaluations as young as 71/2 weeks and are scored on a number of traits, including fear of strangers, separation anxiety and excitableness.
The drawbacks are that it takes years of experience for a trainer to be able to score an animal accurately, and it requires a high level of time and attention for volunteers.
Collecting that data on dozens of dogs with smart collars and using Watson AI technology to predict what it means, Guiding Eyes will be able to tell earlier on which animals are fit to continue training as guide dogs. The data could also offer valuable information for the organization’s genetics program by helping to estimate the heritability of temperament traits between generations of dogs.
In a 2019 paper, researchers from NC State’s Departments of Computer Science (CSC) and Electrical and Computer Engineering (ECE) showed that by outfitting dogs with a data collection system that records electrocardiography and other measurements, they could predict 29 behavioral scores with approximately 92 percent accuracy over 11 distinct tasks.
Our technologies are creating impact in real life applications, for two critical populations in this case: guide dogs and their humans.”
—Dr. Alper Bozkurt
The research is part of a long partnership between Dr. Alper Bozkurt, professor in ECE, and Dr. David Roberts, associate professor in CSC. The two met in a faculty orientation workshop after arriving at NC State and have collaborated on several projects involving animals, sensors and artificial intelligence, including work on an interactive harness for search and rescue dogs that monitors their health and allows them to communicate with handlers in a disaster zone.
Projects in Roberts’ Canine Instruction with Instrumented Gadgets Administering Rewards (CIGAAR) Lab study how to improve upon the centuries-old mutually beneficial relationship between man and his best friend.
The CIGAAR Lab smart handle project focuses on the non-visual communication of canine heart and respiratory rates. The main goal is to communicate canine physiological data to a blind handler in real time to give him or her a better idea of the animal’s current physical and emotional state. A scent discrimination project collects physiological and behavioral data from a dog during scent-detection tasks to identify patterns that correlate to the presence of a target odor; it can be used to develop an automated training system that can capture and reinforce a desirable behavioral response upon detection of scents.
Bozkurt’s research involves using engineering to better understand biological systems, be they human, animal or plant.
In work with the National Science Foundation Advanced Self-Powered Systems of Integrated Sensors and Technologies (ASSIST) Engineering Research Center led by NC State, he is helping develop wearable, self-powered systems that track a patient’s health data and environmental exposure.
Bozkurt has studied the potential for turning Madagascar hissing cockroaches into cyborgs for use in disaster zones and, in collaboration with the Plant Sciences Initiative in the College of Agriculture and Life Sciences at NC State, is looking at how sensors can be used to improve plant health for agriculture applications.
“We have been working on wearable technologies for dogs for the last 10 years,” Bozkurt said. “We are now very glad to see that our technologies are creating impact in real life applications for two critical populations in this case: guide dogs and their humans.”
Tech for Good
Trapani’s first puppy, Merrick, was successfully trained and placed, but even before he became a guide dog he served as a therapy dog. He was the first thing Trapani allowed herself to love after her husband passed away. Her second puppy completed training and was matched with a partner. Before graduating, however, he was released, as the stress of actual guide work was too much for him.
The setback made her want to work harder and wonder what could be done to ensure more successful trainings and placements.
Guiding Eyes has been collecting data on its dogs since 1995 and had begun using IBM Cloud to store it, then Watson to help predict whether a dog in training would ultimately make a good guide dog. The company learned about the research going on at NC State from a paper that Roberts and Bozkurt had published together.
An IBM donation is putting smart collars on 350 puppy trainees and is sponsoring dogs being trained by Trapani, who works for the company in Government and Regulatory Affairs. The puppy she worked with before
Jackie was named TJ after Thomas J. Watson, a former IBM CEO and the company’s AI platform’s namesake.
Trapani’s work includes frequent trips to Washington, D.C., and her trainees tag along. She’s constantly watching for a calm emotional state, smooth movements and as little barking as possible. Learning when it’s appropriate to eliminate is also part of the process, especially not when in a senator’s office.
“Jackie is walking the halls of Congress with me, advocating for the potential of AI to benefit society,” she said.
Original article from College of Engineering News