Bucklands Beach Intermediate student Doyul An has built a self-driving robot that learns the way a child does, using nothing more than a normal camera instead of expensive laser sensors.
The project began as a small, short-term build, based on an earlier project Doyul made called Erik AI.
Doyul is developing the robot as an independent project, aiming to create a machine that learns from its own experiences and can navigate on its own, the same way people do.
The robot builds and uses maps of its surroundings, and learns from what it’s encountered before to get better at finding its way around.
Along the way, Doyul realised he may be the first person in his age group in New Zealand to build a robot that learns like a child and drives itself.
He’s currently getting feedback on the project from the University of Arizona, Arizona State University, the University of Queensland, and other universities.
How does it work?
Most self-driving robots use a technology called lidar, which scans the surroundings using light, a bit like radar but with lasers.
Instead, Doyul’s robot uses an ordinary camera, turning what it sees into a system that helps it avoid bumping into things.
He fed the robot hundreds of photos to build a kind of picture dictionary, so it can recognise objects in front of it, work out their size, and steer around them.
Doyul says this kind of technology could one day be used in things like self-driving cars or robot vacuums.
Funding trouble
Sadly, Doyul says funding is tight, and he may not be able to keep working on the project.
Even so, he’s hoping to keep buying parts when he can, so the project can carry on as a legacy for others to build on.
SMART Prep | Quick Question 
