[ad_1]
The problem is that for an AI to learn to deal with the chaos of real roads, it has to be exposed to all the events it may encounter. That’s why self-driving car companies have spent the last ten years. driving millions of miles on the streets around the world. A few, like Cruise and Waymo, have begun testing non-human driver vehicles in a handful of quiet urban environments in the US. But progress is still slow. “Why haven’t we seen the expansion of these little pilots? Why aren’t these tools everywhere?” He asks Urtasun.
Urtasun is making bold claims for the president of a company that hasn’t just road-tested its technology, but doesn’t even have real cars. But it hopes to make building an AI driver faster and cheaper than its competitors, avoiding most of the costs of road-testing the software in real vehicles, giving the entire industry a much-needed boost.
Virtual drives
Waabi is not the first company to develop realistic virtual worlds for testing self-driving software. In the last few years, simulation has become a mainstay for self-driving car companies. But the question is whether simulation alone will be enough to help the industry overcome the last technical hurdles that are preventing it from being a viable proposition. “Nobody has yet created the Matrix for self-driving cars,” says Jesse Levinson, the company’s co-founder and CTO. zooAn autonomous vehicle startup acquired by Amazon in 2020.
In fact, almost all autonomous vehicle companies now use simulation in some form. It speeds testing and lowers costs by exposing the AI to a wider range of scenarios than it would see on real roads. But most companies combine simulation with real-world testing, typically flipping back and forth between real and virtual paths.
Waabi World plans to take the use of simulation to another level. The world itself is created and controlled by AI, which acts as both a driving instructor and stage manager, identifying the AI driver’s weak points and then rearranging the virtual environment to test them. Waabi World teaches multiple AI drivers different abilities simultaneously before combining them into a single skill set. Urtasun says that everything happens without interruption and without human involvement.
rare events
Driverless car companies use simulation to help them test how the neural networks that control vehicles handle rare events (a bike courier driving ahead, a sky-coloured truck blocking the road, or a chicken passing by) and then fine-tune them. according to this.
“When it happens infrequently, it takes thousands of miles to test it properly,” says Sid Gandhi, who works on the simulation at Cruise, which has started testing fully autonomous vehicles on a limited number of roads in San Francisco. . This is because rare or long-tailed events occur in a thousand to one. “As we try to figure out the long tail, we will rely less and less on real-world testing,” he says.
[ad_2]
Source link