Three students in the School of Engineering built an r/c lawn mower for their capstone research project. Then they taught it to steer itself.
What if you could mow the lawn while sitting on your porch, in the shade, on a chaise lounge, with your feet up, your book open and only your drink sweating? Wouldn’t that be great?
“How could it not be?” asked Nick Drake, who recently graduated from Penn State Behrend with a degree in computer engineering. “Lawn care takes too much time. It’s summer. It’s nice out. There are a hundred things I would rather be doing, instead of pushing a mower back and forth.”
He found a better way. As part of a capstone research project, he and two classmates, Zachary Ley and Nicholas Self, designed and built a remote-control lawn mower. With funding from the Internet of Things (IoT) Research Lab at Behrend, they bought a Greenworks 13-inch electric push-mower. They took off the parts they didn’t need – the handle, the clippings bag, and the rear wheels, which they replaced with 2-inch pivoting casters – and then added several “smart” components, including a Raspberry Pi single-board computer.
This lawn mower, built by students at Penn State Behrend, can be operated with an Xbox controller. The left joystick steers it, and the right trigger spins the cutting blade.
Credit: Penn State Behrend
The team programmed the Pi, which serves as a control unit, to transmit signals to and from the mower. That allowed them to operate the system with an Xbox controller. The left joystick steers the mower. The right trigger spins the cutting blade.
“It’s just like driving an r/c car,” Ley said.
Or an Autobot. The Pi gave the mower “eyes.” Using different gradations of green, the team programmed the mower to recognize objects in its path – a ball, for example, or a dog – using a computer vision algorithm. When the mower senses an object, it stops and reverses its path, much like a Roomba vacuum.
“That way, it won’t drive itself into the road, or into the side of a house,” Drake said.
The team’s adviser, Abdallah S. Abdallah, assistant professor of computer engineering at Behrend and the founder of the IoT research lab, was impressed by how quickly the students incorporated that function, which they first discussed in a computer-vision course taught by Fethi Belkhouche, associate professor of electrical and computer engineering. The feature made the mower partially autonomous.
“Autonomous means it’s smart enough to make decisions,” Abdallah said. “It’s an incremental approach to product development, with additional features being added according to an agile project-development framework.”
Abdallah envisioned the project as an ongoing effort, with multiple teams refining the mower’s design over a two- or three-year period. The first version of the mower would be remote-controlled. The second would be autonomous, like a Roomba. The third will be fully intelligent and connect to the Internet cloud, similar to Amazon’s Alexa or a Google Assistant.
The prototype team managed the first two steps.
“They worked really hard on this,” Abdallah said, “and they got further than any of us had expected. They started to implement features from the second generation. They essentially built a functional house-based utility robot.”
Using a computer vision algorithm, students at Penn State Behrend programmed the mower to recognize objects in its path -- a ball, for example, or a dog -- and automatically steer away from them.
Credit: Penn State Behrend
It isn’t perfect. With a full charge, the mower can operate for only 24 minutes, so it’s better suited for a small lawn.
The team considered adding a 30-watt solar panel. That would add four minutes of cutting time, but it also would increase the mower’s price. One goal of the project was to build a prototype that could be sold for less than $750.
A second team of students will continue the project, adding new features – GPS navigation, maybe, or a smartphone app.
“You could lie in a hammock and watch the mower on your phone,” Ley said.
He also has an idea for an artificial-intelligence neural network, which would be used to map a lawn during cutting. That would be more efficient than the prototype mower, which moves in random directions any time it detects an object.
The prototype was built on a modular design, which will allow subsequent teams to add or alter components without rebuilding the entire system. That could lead to different applications, especially in Erie, where lawn mowers spend long winters stored in garages.
“We designed this in a way that allows you to add different attachments,” Self said. “Maybe you fit it with an auger. Maybe, instead of a lawn mower, you have a remote-control snow blower.”