Apple researchers have published a paper outlining a potential leap in autonomous driving technology with improved detection of pedestrians and cyclists.
The Cupertino-based company has remained quiet about its autonomous driving research so far and only an application to test autonomous vehicles in California and spy pics of its self-driving Lexus RX450h fleet are all that have surfaced.
However, Apple researchers recently lodged a preliminary paper to Cornell University’s arXiv directory of scientific research – a common practice for academics before finalising findings – that outlines a new method of Lidar object recognition which improves object recognition and is more efficient.
Traditionally, autonomous driving technology for self-driving cars has required the combination of a variety of sensors such as Lidar, cameras and radar to create an accurate representation of the environment, but it requires significant computing power.
Apple AI researchers Yin Zhou and Oncel Tuzel have suggested a new machine learning code instead, called VoxelNet, that only needs Lidar data to recognise and identify types of objects such as people. The paper says VoxelNet outperforms current Lidar-based systems by a significant margin, though the authors also note it has only been tested in KITTI Vision Benchmark Suite software so far.
“Experiments on the KITTI car detection benchmark show that 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,” said Zhou.
Originating in the 1960s for use in mapping, Lidar uses a pulsated laser to illuminate objects in infrared and read its distance to the sensor. By creating a cloud map of points, Lidar can create a 3D representation of what's around it.
If the simulations are as successful in the real world, VoxelNet will allow for easier implementation of autonomous driving technology with better and safer object recognition. The lighter computer load would also consume less power - particularly useful in electric vehicles.
Apple's efforts join a divided industry working on autonomous driving technology without standardisation. GM, Ford and Google have backed Lidar technology while Tesla focuses on using radar. All systems currently require additional camera input for object identification but a GM executive told TMR earlier this year that Tesla’s system is limited.
“To think you can see everything you need for a level five autonomous car (full self-driving) with cameras and radar, I don't know how you do that,” he said.
If Apple’s claims of a Lidar-only system are true, it could give the tech company an advantage over competitors when Apple comes to market.
MORE: Autonomous Driving News