Baidu to Launch Self-Driving Cars, Buses by 2020

Baidu, one of China's leading internet search engines, says it has carried out successful tests of an autonomous car under mixed road conditions.

The tests are a milestone in the company's efforts to develop self-driving vehicles, which include plans to develop an autonomous bus in the next three years and mass-produce them within five years, according to a Reuters report.

Baidu joins Google and Apple, among other technology companies, in the race to develop sensing, positioning, infotainment and other software to support driverless vehicles.

Baidu's self-driving car—a modified BMW 3 Series—demonstrated full autonomy over a 30-km route on Beijing streets. It reportedly executed right, left and U-turns; decelerated when it detected vehicles ahead; changed lanes; passed cars; merged into traffic from on-ramps; and exited from off-ramps. The car speed reached 100 km per hour during the test runs.

Baidu's AutoBrain, the core of its autonomous driving technology, includes positioning, smart decision-making and control, detection and highly automated driving (HAD) maps. The HAD maps record 3D road data to "within a few centimeters of accuracy" of vehicle positioning, according to the company. With Baidu's object recognition and environment perception technology, the car can detect, recognize and follow other vehicles; recognize road lanes; and accurately gauge distance and velocity.

Baidu says its autonomous car approach is to advance incrementally through different environments, rather than through different levels of driving autonomy. For example, the company says that a bus that runs the same fixed route will be able to fit a reliable model for that route. In doing so it will train computer vision and other deep-learning systems through repetition, and further expand the number of routes mapped.

Within 5-10 years, Baidu says most of China's roadways could be charted with HAD mapping.