While Apple (NASDAQ:AAPL) has been arguably the most dominant tech company of the past decade, some regard it as a laggard in the race for self-driving cars. Its Project Titan initiative has been overhauled a few times, and competitors Waymo, a division of Alphabet (NASDAQ:GOOGL) (NASDAQ:GOOG), and Tesla (NASDAQ:TSLA) are thought to be leaders in this exciting field.
But on November 17, a new research paper on self-driving technology was published by Apple researchers Yin Zhou and Oncel Tuzel, signaling how Apple is moving forward on self-driving tech. The announcement was notable for two big reasons. One, it was the first paper on the subject published by Apple, which shows the company is serious about leading in self-driving technology despite its late start. Two, the research was published in a public journal called arXiv, a very uncharacteristic move for the secretive iPhone giant.
In the paper entitled, VoxelNet: End-to-End Learning for Point Cloud-Based 3D Object Detection, Zhou and Tuzel write that they have developed a software program called VoxelNet which greatly improves a car's ability to detect far-off small objects, such as pedestrians and cyclists, using only LiDAR data.
LiDAR, as I previously wrote, is a preferred depth-sensing technology for self-driving cars, though it is not the only type of self-driving sensor, and has limitations (in addition to being hugely expensive). For instance, due to its low resolution, especially in fog or rain, LiDAR can have trouble detecting smaller objects at a distance. That is why LiDAR is often paired with passive visual cameras synced up with the LiDAR system, adding complication and cost.
VoxelNet apparently greatly improves a car's ability to detect these small, far-away objects with fewer sensors and only LiDAR data. The researchers wrote:
VoxelNet outperforms the state-of-the-art LiDAR-based 3D detection methods by a large margin. 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.
Zhou and Tuzel appear to believe they have invented a technology superior to what is out there. That is somewhat surprising, as it is widely thought that Waymo has the most advanced LiDAR system and the most test miles driven by far. Tesla doesn't even use LiDAR, instead opting for a combination of cameras, ultrasonic, and radar sensors. If Apple were to have hurdled over the other two, that would be quite an achievement, given the late start and well-publicized problems during Project Titan's early days.
Opening the curtain?
With the publication, Apple also showed signs of changing its secretive ways -- at least for certain divisions, such as artificial intelligence and self-driving tech. This is likely because scientists who work on these technologies want to be able to share their work, as sharing one's research gives the individual scientist recognition in the industry, and the feedback between researchers can speed up the innovation process. Many leading tech companies are utilizing this open-source approach more and more.
In fact, this July, Apple launched the Apple Machine Learning Journal, which allowed the company's AI and machine learning scientists to publish research in a public blog for the first time. However, the VoxelNet article is the first publication on self-driving tech and was published in a non-Apple-branded research journal, which is a significant departure.
The reason for the policy change is likely that Apple realizes in order to win in new technologies, the company must attract the best talent, and top tech talent has more and more options these days. While Apple is not exactly an open-source company just yet, these new initiatives show the iPhone giant is adapting to this new world.
Don't declare victory yet
While the paper was no doubt a positive development for Apple, the race to self-driving supremacy is still relatively early. Moreover, the research for VoxelNet was tested in simulation, not real-world deployments. In addition, it is unclear if a more open-source approach will detract from Apple's historically closed (and profitable) ecosystems.
Nevertheless, the announcement is clearly positive for Apple shareholders, who may have just glimpsed an Apple future beyond the iPhone.