Baidu to Test Drive Autonomous Cars in the U.S.
Mar 17, 2016
Move is part of Chinese tech giant’s effort to develop a commercially viable driverless vehicle by 2018 
Baidu’s autonomous car at the company’s headquarters in Beijing. Photo: Bloomberg News
Baidu Inc.will soon start testing autonomous cars in the U.S., part of the Chinese tech giant’s effort to introduce a commercially viable model by 2018.
The move, disclosed by Baidu’s chief scientist Andrew Ng in an interview late Tuesday, is a significant step for the company, which is trying to get ahead in the race to build autonomous cars and is now calling on the resources of its Silicon Valley tech center to advance the effort. At the same time, Baidu is advocating for better coordination with the U.S. government, which the company says is necessary to get self-driving cars on the road.
Central to the push is Mr. Ng, an artificial-intelligence scientist who conducted groundbreaking research at Stanford University and at Alphabet Inc.’s Google. He’s also a co-founder of online-learning company Coursera Inc.
Late last year, Beijing-based Baidu became the latest technology company to publicize its intention to develop self-driving cars. As cars increasingly come to resemble robots, nontraditional car makers are gaining a wider berth into the automotive industry.
Mr. Ng’s lab at Stanford, where he remains an assistant professor, developed the open-source “robot operating system,” or ROS, which is used world-wide.
Baidu, a 16-year-old search-engine company, is often called the “Google of China.” Hired by Baidu in 2014, Mr. Ng leads a growing office of 160 in Sunnyvale, Calif., where many employees are dedicated to developing a self-driving car.
Ahead of Tuesday’s hearing by the U.S. Senate Commerce Committee on the regulatory obstacles to autonomous vehicles, Mr. Ng released testimony calling for governments to make small changes in infrastructure to accommodate robot cars.
Among those who testified at the hearing were representatives from Google, General Motors Co.and Delphi AutomotivePLC and Duke University researcher Missy Louise Cummings. Mr. Ng didn’t attend.
Mr. Ng recommended the development of mobile applications to facilitate communication with driverless vehicles. Some applications could be used by construction workers, who could send a signal to computer-driven cars to tell them how to get around unexpected road closures. Others could allow law-enforcement authorities to communicate with these vehicles when they are directing traffic.
Today, fully autonomous vehicles like Google’s RX450h can handle a wide range of situations, but they still can’t understand a police officer yelling through a bullhorn or process nonverbal cues from other drivers. While these so-called corner cases might seem rare, they in fact occur regularly.
“The (artificial intelligence) is good enough where the changes to infrastructure are modest,” Mr. Ng said. “Maybe in the distant future, we could make it drive like a human driver, but not in two years.”
Baidu’s aim is to roll out self-driving shuttles running a standard loop in a limited area in China by the end of 2018. A well-trod route would limit the potential for the unexpected to occur. The routes would be expanded gradually as the systems learned. Baidu is using modified BMW 3-series sedans for its tests and already has commitments with several Chinese auto makers to deploy its first autonomous vehicles, Mr. Ng said.
Baidu hasn’t decided whether its first autonomous vehicle will have a steering wheel, Mr. Ng added.
Source: THE WALL STREET JOURNAL