FLINT, MI – Traditional school bus safety might be a thing of the past.
A group of undergraduate computer science students at Kettering University are developing a system to implement cheap, reliable artificial intelligence systems on school buses.
Through a partnership with Navistar, a truck and school bus manufacturer that employs Kettering alumni, students have used this term to develop systems that would alert the driver how many students are on the bus and whether the bus door is safe to close in real time.
The “No Child Left Behind” program is finding ways to improve safety on school buses.
With a running count of everyone on the bus and a notification for when it is safe and unsafe to close the door, Kettering students are implementing that safety.
Using refurbished cell phones -- in this case, the Google Pixel 4 -- students can use the technology in the devices to communicate data to the driver, using a client and server setup.
Essentially, the driver will have one of the phones placed in a safe position on the dashboard that will show student counts and a flag for whether it is appropriate to close the door or not.
Using the Pixel’s camera, students can measure the depth of children getting on to the bus to indicate whether they are in the “pinch zone” or not, which is about a five-foot section before and after the doorway.
The Pixel is then able to process that data and communicate with the other phone via Bluetooth which signals the flag on the driver’s dashboard display -- green means safe to close, red means do not close.
While the two processes are basic ideas, the same AI that students are training and learning how to use in the senior design classroom can be applied in various applications.
Phone cameras could detect smoke and alert the driver. They could eventually even recognize human behavior to detect possible bullying. These models, hypothetically, could be used outside of the school bus as well.
“When we were working on this project, we figured out that we created modes that can be deployed for other applications too,” Electrical and Computer Engineering Professor Mehrdad Zadeh said. “We are creating the artificial intelligence model that can act and notify people and do some decision making in real time.”
The models they are creating in class are scalable to be used with bigger projects.
Since the technology would be used primarily in school districts, the group chose to use refurbished phones to keep the cost low.
Typically, these kinds of systems would operate under one computer. But if you can have the two phones communicate with each other, you can get double the computing power for half the price.
It’s an area of research called edge computing where multiple processors communicate with each other to complete a task.
To keep a running count of people on the bus, students are training artificial intelligence to recognize the fronts and backs of all people in all different circumstances. When optimized, the AI could detect the smallest of elementary students wearing a white jacket in front of a field of snow.
When MLive-The Flint Journal visited the classroom on Jan. 24, the team of students working on the live count had asked Professor Zadeh for permission to use the Navistar school bus at the General Motors Mobility Research Center located on Kettering’s campus.
The team needed to train the AI to differentiate students that are getting on and off the bus from other bystanders that may be on the sidewalk or near the bus stop who are not getting on.
Using the Mobile Research Center and its accompanying testing track, students can simulate situations to train the artificial intelligence without actually picking up real students.
The program just began its fifth year of a five-year partnership with Navistar, who provides education opportunities and some scholarship funding.
Navistar Senior Technical Specialist Teik-Khoon Tan is the company’s liaison with the researchers at the university.
He described the partnership as a “perfect marriage” in a Kettering University news release.
“This is thinking outside of the box,” he said. “We need innovation, so our strategy is to partner with universities. Without it, we lack other innovative perspectives. (With) Kettering students -- the learning and growth are very high. I’m very impressed.”
In the first few years of the program, students worked on autonomous sensors and behavioral learning.
Read more at The Flint Journal:
Davison High principal stars in behind-the-scenes snow day parody video
Town hall on UM-Flint strategic transformation scheduled for Wednesday
Prahl Center at Mott Community College slated for $25M renovation project