Apple is known for its secretive approach toward future products, but the company’s researchers have disclosed for the first time details about their study on self-driving cars. The company’s computer scientists posted their research paper on November 17 to online journal arXiv, providing a glimpse into how autonomous vehicles can be improved to better detect pedestrians and other objects from a distance with the use of a few sensors.
According to the paper’s authors, Yin Zhou and Oncel Tuzel, a new software tool called "VoxelNet" can assist computers in identifying three-dimensional objects by using just the data from LiDAR sensors. While the Apple scientists claimed that this software approach was able to surpass other methods for spotting 3D objects using only LiDAR, the study was conducted with just computer simulations with no actual road tests.
Autonomous vehicles work to combine typical two-dimensional cameras with LiDAR sensors to scan their surroundings. However, the ability of self-driving cars to recognize small and distant objects is limited by the low resolution of most LiDAR units. Nonetheless, a normal camera tethered to those units can help to better detect faraway cyclists or pedestrians, the research suggests.
Apple’s recent disclosure about its self-driving car research is supposed to help advance the development of artificial intelligence and machine learning research initiatives. Previously, the tech giant remained mum on the kind of self-driving car it was working on despite Chief Executive Tim Cook’s confirmation last June that the Cupertino giant was working on software for driverless cars. Additionally, the company received a permit in April to test self-driving cars in California.