Baidu tests autonomous car, emerging as potential rival to Google and others

09.12.2015
Baidu has indicated that it could be a serious competitor in the market for self-driving cars by testing a fully autonomous car on a route it said had mixed roads under a variety of environmental conditions.

The vehicle, a modified BMW 3 Series,  is said to have made right, left and U-turns, slowing down if it detected vehicles ahead, changed lanes, passed other cars and  merged into traffic on the highway,  Baidu said Wednesday.

The car reached a top speed of 100 kilometers per hour (62 miles per hour) during the test runs on a 30-kilometer route, which began and ended at Baidu's Beijing headquarters near Zhongguancun Science Park in Haidian District.

Baidu, which revealed last year that it was developing a driverless car, tied with BMW to jointly research driverless car technologies. Earlier this year, Wang Jin, a company executive, said at a conference that a self-driving car would be launched in China with BMW by the end of this year.

The autonomous capabilities demonstrated by the Baidu car are also being tested by other companies aiming to offer self-driving cars, including Google. The Chinese player could, however,  be a serious contender in the autonomous car market, particularly because of its strong brand and presence in China.

Better known for its search engine, Baidu has also ventured into devices like a wearable called Baidu Eye that could rival Google Glass. The self-driving cars are being researched by Baidu's Institute of Deep Learning since 2013.

The company appears to plan to go beyond cars and also develop self-driving buses built around Baidu AutoBrain technology, which includes highly automated driving (HAD) maps, positioning, detection, and smart decision-making and control.

The company's focus appears to be on making the driverless vehicles useful in specific environments. It gave the example of training computer vision and other deep learning systems through repetition for a bus that runs the same fixed route, and later expanding the number of routes mapped by HAD mapping.

John Ribeiro