Baidu Apollo is the leader and most influential autonomous vehicle (AV) developer in China. Apollo has aggressive goals across multiple AV use cases — especially robotaxis. As a leader in AI research applying technology to AVs, Baidu’s AV research began in 2013, while the Apollo project started in early 2017.
Baidu was founded in 2000 in Beijing and became China’s leading internet–search company. Baidu went public on the Nasdaq exchange in 2005. With total revenue topping $19.5 billion and 2021 profit of more than $1.6 billion, Baidu has the financial resources to invest in AVs and robotaxis.
This column looks at Baidu’s AV activities, including a summary of its July announcement of a robotaxi–centric vehicle. The table below summarizes Baidu Apollo’s AV activities.
Apollo is Baidu’s open–source autonomous–driving tech platform that was introduced in 2017. The goal is to provide an open and safe solution to enable its partners in the automotive industry with autonomous–driving capability. Apollo is one of the world’s most active open platforms for autonomous driving, with over 700,000 lines of source code, 80,000+ developers, and 210+ industry partners.
Baidu has over 500 L4 AVs in testing or robotaxi operation. In 2021, Baidu had two AVs testing in California. In 2018 and 2019, Baidu had four AVs testing in California.
Baidu’s AV testing reached over 32 million kilometers (20 million miles) by the end of June. In California, Baidu’s AV tests reached 130,000 miles by the end of 2021.
Apollo has received 593 autonomous–driving test permits in China, including 398 test permits for passenger transportation.
By the end of March, Baidu had filed over 3,700 patent applications. Currently, Baidu ranks first in the world with over 1,000 high–level autonomous–driving patents.
Apollo is active in multiple AV use cases, with robotaxis as the most important segment. Baidu is also active in developing and testing AVs for use in mass–transit systems — often called robobuses. This includes minibuses, such as Apolong, which has been tested since 2018 in 22 urban parks in several cities, including Beijing, Guangzhou, Xiongan, Chongqing and Foshan.
The first–generation Apolong is the only AV minibus to achieve large–scale operation in China. Apolong accumulated more than 120,000 users and a total mileage of 120,000 kilometers by August 2021, at which point Apolong II was introduced.
Baidu received permission to test robobuses in Beijing in late March. Baidu is working with QCraft and SenseTime and will test eight additional robobuses in Beijing.
Baidu is also developing goods–only AVs. These AVs are the basis for special–purpose applications, such as street sweepers and similar tasks.
Apollo Go is currently available in 10 cities: Beijing, Shanghai, Shenzhen, Guangzhou, Chongqing, Changsha, Cangzhou, Wuhan, Wuzhen and Yangquan. Apollo Go has started commercialized operation in multiple cities. Users can hail a robotaxi with one tap in Baidu Maps, Apollo Go’s standalone ride–hailing mobile app, or the Apollo Go mini program on the Baidu app. In April, Baidu received the first permit in China for driverless ride–hailing services in Beijing.
Apollo Go has provided over 1 million robotaxi rides — mostly for free. Paid robotaxi rides are available in four cities: Beijing, Chongqing, Yangquan and Wuhan (added in early July).
Baidu plans to expand Apollo Go operations to 65 cities in China by 2025 and 100 cities by 2030.
Baidu has a large AV test site in Beijing called Apollo Park. The test area supports AV development, as well as 5G and C–V2X technology testing. Infrastructure to enhance AV use is a big part of Baidu’s plan to expand Apollo Go. Baidu will use 5G C–V2X as it deploys robotaxis. Apollo Park has over 200 AVs and can support all aspects of development and testing.
Baidu Apollo announced its first purpose–built robotaxi for the Chinese market on July 21. It is the sixth generation of robotaxi vehicles and the source of its name, RT6. It is a battery electric vehicle (BEV), which is a production–ready model. RT6 has a detachable steering wheel.
RT6 will be put into operation in China in 2023 on Apollo Go, Baidu’s autonomous ride–hailing service. The table below summarizes public data on the RT6.
The dedicated robotaxi design makes the Apollo RT6 distinct from earlier generations that were retrofitted for AV use in conventional vehicles. The removable steering–wheel design provides more space for unique interiors at the traditional driver seat.
The RT6 has L4 autonomous–driving capability, provided by 1,200 TOPS of computing power, with an extensive sensor architecture that includes 38 sensors. Only the number of each sensor category was included in the RT6 announcement.
RT6 has eight LiDARs. From the released pictures, it is clear that only solid–state LiDARs are used. It is unlikely that the RT6 uses any frequency–modulated, continuous–wave (FMCW) LiDARs. FMCW LiDARs are best for high–speed and long–distance views, which currently are not needed.
RT6 has six radars — probably only traditional radars without any 4D or imaging radars. One 4D–imaging radar would be useful for the forward–looking radar due to the superior interference rejection compared with traditional radars.
The Apollo RT6 exterior features an innovative look that integrates sensors on the sunroof alongside interactive lights. This provided excellent sensor integration, as seen in the below images.
Apollo RT6 is the first vehicle model built on Xinghe, Baidu’s self–developed automotive EE architecture designed specifically for autonomous driving. The RT6 uses automotive–grade electronics and has redundancy in both hardware and autonomous–driving software.
Apollo RT6 will join the Apollo Go ride–hailing service starting in 2023. However, there is no specific starting month at this time.
The production cost of the RT6 is impressive, at 250,000 RMB (about $37,000). This is half the cost of the fifth–generation robotaxi, the Apollo Moon, which is currently used in Baidu’s robotaxi services.
In March 2021, Baidu formed a joint venture with Geely called Jidu to produce BEVs and AVs. In December 2021, Jidu said it would introduce a concept vehicle in mid–2022 and deliver its first mass–produced AV in 2023. In June, Jidu introduced a prototype of its first vehicle, called Robo–1, which has considerable autonomous features. Jidu says it will cost about $30,000 and is expected to go on sale in 2023. This Wired article contains more information on the Robo–1.
It appears the RT6 could be based on the Robo–1 BEV, as the Robo–1 side–view picture in the Wired article looks similar to the RT6.
Baidu believes the RT6’s large cost reduction will lead to deployment of tens of thousands of AVs across China in a few years. It will also lead to future robotaxi rides costing half of current taxi rides. Baidu said it plans to produce 100,000 Apollo RT6 vehicles over an unspecified time period.
Baidu included a surprising and probably a controversial statement in its RT6 press release: The autonomous–driving capability of the Apollo RT6 is equivalent to a skilled driver with 20 years of experience.
The Chinese robotaxi industry continues its rapid development and is setting aggressive goals. Baidu Apollo is a leader in this trend, but there are many other competent competitors. The strong backing of the Chinese government is an important factor in AV progress. Various cities are also very supportive, and there is strong competition between major cities to lead in the road to perceived future riches at the end of the rainbow.
Baidu is at the center of the Apollo open–source AV development project, which continues to gain support from both Chinese and foreign companies. Baidu’s robotaxi service group, Apollo Go, has 500 AVs in testing or robotaxi operation, with plans for rapid expansion. The Apollo RT6 robotaxi is the key to this growth, with plans for tens of thousands of robotaxis in operation in a few years. RT6 deployment starts in 2023.
Apollo Go plans are even more enthusiastic, with robotaxi services in 65 Chinese cities in 2025, scaling to 100 cities by 2030. This would likely require at least 100,000 robotaxis by 2030 (1,000 per city). If you assume Apollo’s competitors will have about two times as many robotaxis (Apollo has 33% share), we are potentially talking about 300,000 robotaxis operating in China by 2030.
Is this a reasonable or low–probability scenario? Only time will tell, but there are a few helpful factors in China that are not present in the U.S. or Europe. The Chinese government essentially decides how fast robotaxi deployment will happen. AV safety is, of course, a major factor, but the overall crash rate in China is much higher than in the U.S. and Europe. Hence, robotaxis and AVs in general can lower the vehicle crash rates in China without matching AV crash–safety rates required or tolerated in the U.S. and Europe.
Additionally, any AV crash fatalities in the U.S. and Europe will be front–page stories and will negatively impact the AV industry. In China, not so much. The long–term goals are more important than short–term negative impact, and China is convinced that AV technology will bring major advantages ranging from economic to positive human impact.
China will lower the AV risks by only allowing robotaxi operation in parts of cities — the parts where traffic and roads are amenable to safe operation. The areas of robotaxi operation will expand with time and as AV system capability advances. The incorporation of 5G–based C–V2X will also add to AV safety in China as deployment of C–V2X for all vehicles expands quickly.