Nvidia’s self-driving safety report aims to prove just how safe self-driving cars are

Tom Bedford October 24, 2018

Legislation is the biggest hurdle self-driving cars need to overcome before they can hit the streets. As a major player in the development of self-driving cars, Nvidia feels it’s its place to forward this cause with the announcement of a self-driving safety report for autonomous cars.

Nvidia’s self-driving safety report aims to prove just how safe self-driving cars are

This self-driving safety report opens up the technology and systems in place to ensure automated cars are as safe as possible. It explains how these cars conform to industry guidelines and how companies are constructing security failsafes. It also posits four “pillars of safe autonomous driving”, rules by which automated cars should be designed.

The first pillar is “an artificial intelligence design and implementation platform”. This means a vehicle must be, in effect, panoptic, sensing, computing and resolving situations in real time. One of Nvidia’s GPUs for autonomous vehicles , the AGX Pegasus, can run 320 trillion operations per second, translating to quicker reactions and a safer driving experience.

Pillar two is a “development structure that supports deep learning” – not only must a car be able to process information quickly, but the data from it and all other cars should be collected and corralled. To this end Nvidia has created infrastructure to store and process all this data, to help the fleet as a whole learn and improve.

The third pillar is “data centre solution for robust simulation and testing”, or the ability to test the automated AI in simulated settings. This doesn’t mean Nvidia plugs its AI into Grand Theft Auto V, however; a specially designed driving simulator program is used to throw all sorts of problems at the car, such as different weather conditions, obstacles and lighting issues. This tests autonomous cars in ways that wouldn’t typically be possible in real-world tests.

Finally is the “best-in-class pervasive safety program” through which the automated vehicle must adhere to various regulations and safety regulations. These rules come from government and industry guidelines, to ensure a vehicle is safe for its driver and other drivers.

These pillars build on Nvidia’s plans for driverless cars’ driving licenses. Nvidia’s pledge to the safety of automated cars will hopefully spur regulatory bodies into supporting the burgeoning industry.

Many companies have taken steps to prepare for driverless cars becoming a reality, with London taxi firm Addison Lee preparing to launch driverless taxis, Tokyo trialling self-driving taxis before the Olympics, and Stagecoach planning to trial automated buses by as early as the end of 2018. Before any of these ventures become publicly usable, however, more regulations need to be passed.