Apple computer scientists revealed a new method that self-driving could use to detect pedestrians and cyclists in a recently-published research paper, giving a rare glimpse into the U.S. technology giant's work in the field.
In the paper from November 17 published in online journal arXiv, Apple researchers Yin Zhou and Oncel Tuzel discussed the advancements they have made in being able to detect 3-D objects.
Current object detection systems rely on "LiDAR" technology. This works by shining light onto a surface and measuring how long it takes to return in order to figure out how far an object is and create an image of its shape.
But LiDAR alone can make detecting small objects that are far away difficult without an additional camera. Apple's researchers argue a LiDAR and camera-based solution could be difficult to deploy in many situations and it could be "more sensitive to sensor failure."
Instead, Apple's computer scientists propose a LiDAR-only detection method that they dub "VoxelNet". The solution uses complex computer vision and artificial intelligence to carry out this function.
Apple tested this on a computer simulation rather than a real car but the researchers claim that VoxelNet "outperforms the state-of-the-art LiDAR-based 3-D detection methods by a large margin."
The iPhone maker has been relatively silent on self-driving cars beyond a few vague comments from Chief Executive Tim Cook. In an earnings call in August, Cook said creating autonomous systems is the "mother of all AI projects."
He also said that Apple is focused on such systems from a "core technology point of view," which appears to ring true with the latest piece of research. In other words, Apple may not be building an actual car, but instead creating key software that will help these vehicles operate without a driver.
Earlier this year, Apple was granted a permit by Californian authorities to allow it to test self-driving cars.