Bicyclists have it rough. Unlike the automobile driver, the bike rider travels without protective shell or seatbelt, the body vulnerable to any impact. To avoid collisions, new cars have cutting edge technology to enlighten blind spots and warn of approaching vehicles. Bikes have no such thing.
Travis Chan, a junior at Johns Hopkins University majoring in electrical and computer engineering, was thinking of such issues while riding a scooter two days before the HopHacks hackathon he planned to attend.
“I noticed it was so hard to see cars coming from behind me,” he said. As bicyclists outnumber scooterists, Chan decided he would use the coming hackathon to create a tool to help them. “And, from a technical standpoint, it’s easier to mount it on a bike.”
Chan teamed up with Thomas Keady, a robotics master’s student, and, with three backpacks and several tool boxes of gear, got to work. “Thomas and I ran up to one of the classrooms, go our own corner, and set up shop. We were there the entire time,” Chan said.
In addition to the soldering irons, microcontrollers, and LEDs available at the hackathon, the team brought their own LIDAR unit. The device, which senses objects at a distance using a laser, was essential to their invention.
So was a good deal of code. “From a software perspective, part of it was dependent on found machine learning algorithms others had made for identifying pedestrians and cars—open source stuff,” Keady said.
The hackathon lasted 36-hours and in that time the pair was able to cobble together code and hardware into a working system and mount it on a bike. “It was a huge time crunch,” Keady said. “We 3D-printed parts and had to hope they would print in time. We didn’t finish till two hours before we had to stop working on it.”
But finish it they did. Unfortunately, they did not have time to take the prototype outside for a test run. So the first and only YOLObike, as they are calling the product, was used to identify the humans milling around the hackathon instead of vehicles.
Regardless, they proved the viability of using neural networks to detect enemy traffic for a bicyclist—and with all processing done on board the bike, not relying on external power nor cloud connections.
Chan and Keady’s efforts won them the grand prize at the hackathon—and its award of $1,024. They hope to use their winnings to further their product.
“One idea we had was to add a GPS unit onto it and create a ‘dangerous quotient,’ ” Chan said. The data from the GPS would measure not only automobile traffic, but how close cars drive to bicyclists, and what time periods are most dangerous on any given road. The YOLObike could then offer the safest route from point A to point B at any given time.
Another possible innovation would be to add an accelerometer. “If someone tries stealing the bike when it’s off, it could sense that,” Chan said. “It would be pretty cool to turn on a buzzer and take a picture of the person trying to steal the bike.”
At the moment, no one could even attempt to steal a YOLObike-equipped bicycle. The pair had to disassemble their device at the hackathon’s conclusion. “We needed some of the hardware we were using for classes, so we could get back to work on our homework,” Keady said. “But we are going to try to put the whole system together again and give it a shot outside to see what it’s like on the road.”
Michael Abrams is a technology writer based in New Jersey.
Recommended for You: MakerHack Magic for DUMBO