Waymo, Uber Propose AI Techniques to Improve Self-Driving Systems

Original article was published on Communications of the ACM – Artificial Intelligence



By Venture Beat

June 30, 2020
Comments

Waymo's fully self-driving Jaguar I-PACE electric SUV.

During a workshop on autonomous driving at the Conference on Computer Vision and Pattern Recognition 2020, Waymo and Uber presented research to improve the reliability and safety of their self-driving systems.

Credit: Waymo

Researchers at Waymo and Uber presented new technologies at the Conference on Computer Vision and Pattern Recognition 2020 that aim to make their self-driving systems safer and more reliable.

Waymo’s Drago Anguelov demonstrated ViDAR, a camera and range-centric framework created in partnership with Google Brain.

ViDAR uses motion parallax to learn three-dimensional geometry from image sequences, like frames captured by car-mounted cameras.

The technology spurred the creation of models that, for instance, estimate depth from camera images and predict the direction that obstacles, like pedestrians, will travel.

Meanwhile, Uber’s Advanced Technologies Group’s Raquel Urtasun highlighted the V2VNet system, which allows autonomous cars to share information, such as timestamps and location, over the airwaves.

An artificial intelligence model helps compensate for time delays and selects only relevant data, like LiDAR sensor readings, from the data sets.

V2VNet’s error rate was 68% lower than that of single vehicles.

From Venture Beat
View Full Article

 

Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA


No entries found