Two Apple scientists published a paper detailing research into ways self-driving vehicles can better detect obstacles while using fewer sensors, in what could be the first publicly-disclosed information on the company’s work on autonomous vehicles.

Apple declined to comment on reports of the paper’s publication, which Reuters stated had been submitted to indepdendent online journal arXiv by Apple researchers Yin Zhou and Oncel Tuzel on 17 November.

The document highlighted “accurate detection of objects in 3D point clouds” as a “central problem in many applications, such as autonomous navigation”. The research also pointed to housekeeping robots, augmented and virtual reality as potential areas where detection is important.

It proposes a new approach called VoxelNet to solve the issue, from which the company had “highly encouraging results”.

According to Reuters, the news is important because the tech giant’s secrecy around such technologies is often perceived as a disadvantage for artificial intelligence (AI) and machine learning researchers.

Apple did introduce the Apple Machine Learning Journal earlier this year, but its research rarely appears outside the publication and has not featured self-driving cars.

In April, Apple filed a self-driving car testing plan with California regulators.

Although there was much speculation about Apple’s plans in this area in recent years, the company did not reveal any official line on the topic until December 2016.

At the time Steve Kenner, Apple’s director of product integrity, noted automated vehicles “have the potential to greatly enhance the human experience” by preventing “millions of car crashes and thousands of fatalities”.

In a letter to the US National Highway Traffic Safety Administration, Kenner also noted self-driving vehicles could deliver “mobility to those without”.

Apple’s research paper appeared to follow Kenner’s line: Reuters stated avoiding pedestrians and cyclists formed a key part of Zhou and Tuzel’s focus.