Innovation

Apple's autonomous car software uses neural networks to improve navigation, object detection

In a research paper, Apple scientists explain how new software can help self-driving cars identify pedestrians and cyclists by improving LiDAR.

The years of secrecy are over: Apple scientists published a research paper detailing new software that could help autonomous cars better detect pedestrians and cyclists using fewer sensors.

The paper, submitted on November 17 to independent online journal arXiv, explains how a software called VoxelNet helps computers detect 3D objects. Autonomous cars, housekeeping robots, and augmented/virtual reality projects often struggle with this, the researchers noted.

Autonomous vehicles use a combination of 2D cameras and LiDAR (a remote sensing method that uses light in the form of a pulsed laser to measure distance ranges) to navigate the world around them. However, LiDAR is not always effective at detecting small, distant objects on its own.

SEE: Survey: What is your organization doing with autonomous transportation?

However, Apple's research claims that VoxelNet software allows LiDAR to better detect such objects—including cyclists and pedestrians—on its own.

VoxelNet also outperformed state-of-the-art LiDAR-based 3D detection methods "by a large margin" in experiments on the KITTI car detection benchmark, the researchers noted.

"Furthermore, our network learns an effective discriminative representation of objects with various geometries, leading to encouraging results in 3D detection of pedestrians and cyclists, based on only LiDAR," the researchers wrote.

Apple has been notoriously secretive about their autonomous vehicle development plans in recent years, and this paper represents the company's first public acknowledgement of its work on self-driving cars. In June, CEO Tim Cook said that the tech giant was focusing its efforts on "autonomous systems," including by not limited to those in self-driving cars, but also said that Apple would not reveal what it was doing in the space in terms of product development.

The paper seems to confirm rumors that Apple had nixed its plans for developing an actual self-driving vehicle to compete with Tesla, and pivoted instead to working on autonomous software platforms.

The paper marks Apple's first official steps in the autonomous vehicle market, joining companies including Google, Tesla, Uber, and Intel. This move does not come as a surprise, Bryant Walker Smith, assistant professor of law at the University of South Carolina and expert in the autonomous vehicles space, told TechRepublic amid earlier Apple self-driving car rumors.

"Companies like Apple, Google, and Amazon want to be ubiquitous in our lives," Walker Smith said. "For better or worse, driving is a big part of many people's lives, so I'm not surprised that Apple is exploring the car as another device or at least as another space. We'll likely see some interesting partnerships in the next few years as companies combine their strengths."

istock-692832436.jpg
Image: iStockphoto/chombosan

The 3 big takeaways for TechRepublic readers

1. Apple scientists recently published a research paper in the journal arXiv detailing new software that could help autonomous cars detect objects in the road with fewer sensors.

2. The software, VoxelNet, allows the LiDAR sensors in autonomous cars to better detect objects including cyclists and pedestrians on its own, without additional cameras.

3. The paper represents Apple's first official detailing of its plans in the autonomous car space.

Also see

About Alison DeNisco Rayome

Alison DeNisco Rayome is a Staff Writer for TechRepublic. She covers CXO, cybersecurity, and the convergence of tech and the workplace.

Editor's Picks

Free Newsletters, In your Inbox