Baidu's contribution to the self-driving-future:
Google is not the main search engine in all countries around the globe. In China for example, people use Baidu to get their internet questions answered. “The Google of China” does not only compete with Google for search query shares, but like its American rival, the company is also pushing to get the first Baidu self-driving car on the global market.
Baidu only began investigating driverless technology in 2013. A late start, but Baidu’s focus on artificial intelligence gives them an edge to make up ground. In 2014 it partnered with BMW and modified a handful of BMW 3-series sedans with the approach to assist drivers rather than launch fully autonomous vehicles. Today Baidu is working on releasing fully autonomous vehicles that will be limited geographically, like a bus that drives the same route every day. This mix of both driver assisted and fully autonomous is unique to the industry.
The Chinese market
Like Google, Baidu is making its own 3D maps, building on the database of the very popular Baidu Maps, and feed them through their AutoBrain software to the cars’ Velodyne Lidar radar (again, like Google). Nothing new to the field, but a significant milestone for China, which is expected to be the largest market for self-driving vehicles globally. Given the high rate of accidents and traffic congestion in China, Chinese consumers are more interested in autonomous cars compared to the U.S. Baidu aims to map the majority of China’s roads within 10 years and estimates that the Chinese market for car sales, buses, taxis and related transportation services is potentially worth more than $1.5 trillion a year in revenue. But Baidu is also targeting the US (more on this below).
Autobrain on the road
Talking about the Baidu self-driving car is misleading, because Baidu is not looking to become a car manufacturer. Its partnership with BMW is to marry the automaker’s expertise with their technological strengths. Those strengths have translated into project AutoBrain. AutoBrain is the main software system that records 3D road data with the accuracy of vehicle positioning less than an inch and is currently the only driverless testing project in China that has passed the ISO 26262 safety standard for positioning, detection, smart decision-making and control of a driverless car.
AutoBrain will also be implemented in Baidu’s next big project. After testing autonomous vehicles in a 30 km loop in Beijing, Baidu now plans to sell self-driving shuttles to run loops in designated areas in China by 2018. A bus that runs the same fixed route will be able to fit a very reliable model for that route, training computer vision and other deep learnings systems through repetition. 2018 is an ambitious target and Baidu CEO Robin Leo admits the car might not fully be ‘street ready’, but it could be the first to actually become ‘sold’. Definitely if Google expects their autonomous vehicle to hit the road in 2020. Only Tesla hopes its first autonomous vehicle to be ready by 2018, but believes that it might take another two to three years for regulatory approvals.
Coming to US
Regulations could prove to be a major hurdle in the commercial launch of driverless vehicles. That is exactly why Baidu is joining regulatory discussions in the United States like Hands Off: The Future of Self-Driving Cars and will start test driving on US soil. If the company succeeds in launching their vehicles ahead of competition in China, Baidu could also capture the first movers advantage and attain a huge competitive edge over other players in the US and Europe.
On US soil, Baidu’s testing will be led by new chief scientist Andrew Ng, an artificial-intelligence scientist who conducted ground breaking research at Stanford University and was previously employed by Google. He’s also a co-founder of online-learning company Coursera Inc. Ng’s Artificial Intelligence know-how will smooth out the leftover bugs in AutoBrain. He for example recommended the development of mobile applications to facilitate communication with driverless vehicles. Today, fully autonomous vehicles like Google’s RX450h can handle a wide range of situations, but they still cannot understand a police officer yelling or process nonverbal cues from other drivers. Ng’s applications could be used by for example construction workers, to send a signal to computer-driven cars to tell them how to get around unexpected road closures.