SUSTAINABILITY

Autonomous Cars

DEEP LEARNING

AUTONOMOUS CARS

Safer Roads, Cleaner Air:
Autonomous Cars Could Drive a Better Future

Fueled by advances in artificial intelligence, cars are getting smart enough to begin to drive themselves. But autonomous vehicles will do more than change how we get around. They have the potential to dramatically reduce the number of car crashes, shrink carbon emissions, and provide mobility to people who can’t drive.

“The contribution of self-driving cars to society is, arguably, incredible,” said Jen-Hsun Huang, NVIDIA CEO and co-founder, during his keynote address at the 2016 Consumer Electronics Show.

Even if just 10-20 percent of cars worldwide have some self-driving capability by 2025, the technology could eliminate driver error — and save 30,000 to 140,000 lives a year, according to a report the consulting firm McKinsey & Company. It estimates that autonomous vehicles could cut fuel use by 15-20 percent and reduce carbon emissions by as much as 300 million tons, equivalent to roughly a quarter of all carbon emissions from petroleum.

Driverless cars weren't possible until researchers adopted an AI technique known as deep learning, which relies on powerful GPUs, access to vast troves of data, and sophisticated algorithms for deep neural networks to solve complex problems.

Deep learning is essential for autonomous vehicles because no one can write software that anticipates every possible scenario a self-driving car might encounter. With deep learning, the car’s computer can learn, adapt, and improve.

Autonomous Driving Could Save Lives

Nearly every major carmaker, along with Google and other companies, are developing cars that can drive themselves some or all of the time. And that could be a good thing for auto safety.

Human errors account for more than 94 percent of U.S. car accidents, according to the National Highway Traffic Safety Administration. The World Health Organization estimates that more than 1.2 million people worldwide die yearly in car accidents.

“That far exceeds the number of people killed in war,” said Gill Pratt, CEO of the Toyota Research Institute, during his keynote address at NVIDIA’s 2016 GPU Technology Conference.

Controlled by supercomputers, driverless cars would avoid such human errors as speeding, drunken driving, and running red lights. Plus, computers don’t get distracted by phone calls or texts, and they never get drowsy.

Toyota aims to make its self-driving cars safer by using GPU-powered simulation and deep learning. These technologies will allow autonomous vehicles to experience and train on numerous real-world situations before they’re on the road, Pratt said.

Real-World Testing

To test its cars in real-world conditions, Volvo is putting 100 self-driving cars in the hands of customers on selected roads in and around its hometown of Gothenburg, Sweden.

The company will equip the cars with the NVIDIA DRIVE PX 2, a supercomputer for vehicles that uses deep learning, sensor fusion, and surround vision to understand a car’s surroundings and navigate a safe path forward.

“Nobody should be killed or seriously injured in a new Volvo car,” said Henrik Lilnk, senior technical leader for the Volvo Car Group, during a talk at GTC. “Long term, new Volvos shouldn’t crash.”

Potential to Cut Fuel Use, Emissions

Driverless cars can also help save fuel and reduce the greenhouse gases that contribute to climate change.

Transportation, particularly cars and trucks, is the second-largest source of greenhouse gas emissions in the United States, according to the Environmental Protection Agency.

Autonomous vehicles use sensors and software to eliminate unnecessary acceleration and braking. They can relay traffic data to other cars to enable the most energy efficient routes. And they could drop off passengers and then park themselves, minimizing the considerable time and fuel people spend searching for parking. These features would reduce energy use in individual autos, as well as ease road congestion.

“Traffic jams can be made a thing of the past,” German carmaker Audi says on its website. “And smooth-flowing traffic means less environmental impact.”

Audi is using NVIDIA’s Tegra X1 processor to provide intelligence needed for “piloted” driving, with an anticipated launch next year.

Freedom for People Who Can't Drive

Everyday activities like shopping for groceries, going to a doctor's appointment, or meeting a friend for dinner can be an enormous challenge for those who can't drive. Driverless cars could provide mobility to elderly people who have given up their licenses, disabled people, and those without a license.

With self-driving cars, "people could have the independence, the dignity, the autonomy themselves to move around in the world to have a high quality of life regardless of their infirmity, regardless of their age," said Toyota's Pratt.