In the paper, Yin Zhou and Oncel Tuzel propose a software program called “VoxelNet” to help computers detect three-dimensional objects. That’s relevant to autonomous cars because pedestrians, cyclists, and other potential obstacles tend to be three-dimensional.

The software could potentially boost the effectiveness of sensors used in self-driving cars to identify obstacles. Most companies developing autonomous vehicles use a combination of cameras, radar, and lidar (which works like radar, but with light), to create a complete picture of the environment. Each type of sensor has its own weakness, which is why all three are needed.

But the software described in the Apple researchers’ paper has shown the ability to spot pedestrians and cyclists with lidar alone, at least in computer simulations.  The system hasn’t been tested on an actual car in the real world yet.

Apple recently received approved to test self-driving cars on public roads in California, and its Lexus RX development mules have been spotted a few times over the past few months. But the company’s specific goals remain unclear. CEO Tim Cook has called autonomous driving the “mother of all AI projects,” but has hinted that Apple will develop autonomous-driving systems only, rather than complete cars.

The post Apple Releases Self-driving Car Research appeared first on 涡轮增压电控执行器 - 深圳亿昇动力科技有限公司.