Grade 10 Lo-Ellen Park Secondary School student Brendon Matusch said he was “completely blown away” when he took home top honours at the Canada-Wide Science Fair last week.
“I was not expecting this at all,” Matusch said.
“I was thinking maybe I would be able to do better than last year (when he received a silver medal for placing in the top 30 in his age category), but this was completely out the blue.”
Matusch not only won the Platinum Award for Best Intermediate (Grade 9/10) Project, he also won Best Project Award, meaning his project was judged to be the best in the entire fair.
He won $3,500 in cash, a fully funded trip to compete in the 2018 European Union Contest for Young Scientists in Dublin, Ireland, which takes place in September, as well as entrance scholarships to several universities.
“All three winners demonstrated an exceptional level of innovation,” said Canada-Wide Science Fair national judge-in-chief, Jeff Hoyle, in a press release.
“The judges were especially impressed by the winners’ originality and creativity. The winner of the best project award further exhibited a high level of enthusiasm and attention to detail combined with a willingness to pursue all avenues of investigation to realize his ideas that made his project stand out.”
“All of this year’s winners developed innovative solutions to real-world problems that affected them directly, at a local level, and yet are all globally significant challenges,” said Reni Barlow, executive director, Youth Science Canada.
“Their innovative thinking and positive solutions generation creates continued hope for our future.”
Matusch's project is entitled "Development of an Autonomous Vehicle Using Machine Learning." He investigated techniques for the use of convolutional neural networks for autonomous driving.
Check out his vehicle in action below:
For this, Matusch developed an autonomous vehicle using components for vision, path planning, speed control and stop sign detection.
As part of his project, these variables were tested in a custom 3D simulation, as well as on a physical vehicle based on a modified electric go-cart.
“A neural network is based on a biological brain,” Matusch explains.
“Basically it's a mathematical construct made up of artificial neurons, and they're stacked up into layers.
“The idea is that you can train them on a set of training data that a human can collect. In my case, I trained it on a set of images and corresponding steering angles that I collected while driving by myself.
“The idea is that the neural network can learn to replicate what I do.”
While Matusch initially used computer simulations to test his work, things didn't exactly go well when he finally took his go-cart for a test drive.
“It was sensitive to stuff like shadows and glare, and I was actually never to get it to drive for more than 73 meters,” he said.
“What I had to do after that was basically scrap everything I had and rebuild it from the ground up, detecting the (road) lines and using that to steer. Once I did that, I had another successful series of tests in mid-October.”
He said he received some help from his dad, Steve Matusch, the owner of the local engineering company Ionic Engineering, in the construction of the go-cart itself.
Matusch said he also wants to thank some of his teachers, as well as Laurentian University professor Ken Westaway, for their advice.
But Matusch said because this is “90 per cent a software project” in an emerging field, he was on his own more than people realize.
“I read papers online,” he said.
“Machine learning is still a relatively rare field. There are not many people you can talk to about it. Most of this was researching stuff online, reading existing papers and then figuring it out as I go.”
Matusch was one of six students from Greater Sudbury who took part in the Canada-Wide Science Fair, and he wasn't the only one who came home with honours.
His fellow Lo-Ellen student Kerry Yang won a gold medal, and Nethra Wickramasinghe of Lockerby Composite School and Rochelle Larivière of Ecole St-Paul both won silver medals.