Apple Publishes Research Detailing Enhanced Obstacle Detection For Self-Driving Cars
In an uncharacteristic move, Apple has published some research on self-driving car technology, detailing new software designed to improve obstacle detection.
Apple has so far been very quiet regarding its autonomous driving plans and research, but the company is apparently ready to share some of its progress now. The research paper details what Apple calls VoxelNet, which builds on Light Detection and Ranging (LiDAR) technology to allow self-driving cars to better detect the obstacles in their surroundings.
Apple took to Arxiv, a public site that often serves as a venue for researchers to share their studies and get some feedback before publishing their research in a final, more polished form. Apple's post [PDF] on Arxiv indicates that it's making headway in the autonomous driving space and it may soon be ready to make an official announcement.
Improved LiDAR With 3D Cloud-Based Object Detection
Based on the new research Apple has published, it seems that VoxelNet will use LiDAR to make it easier for self-driving cars to detect where objects are located, as well as other essential details such as whether it's dealing with inanimate objects, cyclists, pedestrians, and so on.
It's basically a 3D object recognition system based on LiDAR, designed to detect obstacles in 3D point clouds by using neural networks that can be trained. While it's just getting started for now, the technology has great potential and it could significantly enhance LiDAR navigation accuracy when it's more advanced.
Autonomous vehicles typically rely on a combination of LiDAR and cameras to grasp details about their surroundings. This allows them to detect objects from a distance, but there's much room for improvement. The additional details that VoxelNet would add to LiDAR could notably improve safety and take things to the next level.
Apple Self-Driving Car Efforts
This paper bolsters claims that Apple is making efforts to have its own self-driving cars, albeit it will have some heavy competition. Google, Uber, and other companies are already testing their autonomous vehicles on public roads, so Apple has some catching up to do. This research indicates that its debut on the scene could be sooner than expected.
At the same time, the research confirms that Apple is no longer as secretive as it used to be regarding its research. The company announced back in December 2016 that some of its researchers would be able to publish their work in papers. Apple also released its first research paper on artificial intelligence (AI) the same month, marking a historic first for such a tight-lipped company, and it's now sharing more.