How far is the real unmanned driving?
On August 18th, Baidu CEO Robin Li officially released Apollo "automobile robot" and unmanned car travel brand "Carrot Run" at Baidu World Conference 2021.
According to Baidu, car robots already have L5 autonomous driving capabilities and can realize fully autonomous driving.
As the domestic "leader" of autonomous driving, when will Baidu Apollo achieve commercial mass production?
Where is the difficulty?
Can it revive the public's confidence in autonomous driving?
1
Baidu releases Apollo "automobile robot"
Possesses L5 autonomous driving capability
At Baidu World Congress, which is an important stage for Baidu to show its technological breakthroughs, the autonomous driving Apollo has always been the highlight, and this year is no exception.
On August 18th, Baidu CEO Robin Li officially released Apollo "automobile robot" and unmanned car travel brand "Carrot Run" at Baidu World Conference 2021.
According to reports, Baidu's car robot has L5 level autopilot capabilities and multi-mode interaction capabilities such as voice and face recognition. It can analyze the potential needs of users and proactively provide services.
In addition, car robots also have self-learning and continuous upgrading capabilities.
From the experience of the live guest, the appearance of the car robot is an automatic gull-wing door, a full glass roof and external sensors. There is no steering wheel, pedals and other mechanical control facilities in the cab. It has a large curved screen and intelligent control. Intelligent configuration such as table, dimming glass, zero-gravity seat, etc., the whole operation is controlled by sensors and voice.
During the driving, the car robot drives automatically throughout the entire process, smoothly coping with complex scenes such as unprotected left turns, avoiding pedestrians and other vehicles.
Li Yanhong said: "This section of the road is already quite complicated. Auto robots not only have to abide by the traffic rules themselves, but also deal with other people's violations."
At last year's Baidu World Conference, Baidu Apollo demonstrated the ability to go to safety officers, based on 5G cloud driving technology for fully driverless, and completed the world's first fully driverless live broadcast.
However, in last year's live broadcast, cloud drivers still need to take over driving frequently when encountering complex road conditions.
In this year's live broadcast, almost no active intervention was seen.
According to Baidu, 5G cloud drivers have been able to take care of multiple unmanned vehicles by one person.
Li Yanhong pointed out: "Baidu has been working on this technology for 8 years and it is still iterating. It is no problem on open roads, but various license plates are still needed. In the future, it will enter the era of automotive robots."
According to the "Automotive Driving Automation Classification", the ACC, ABE, parking assist and other functions that many cars are currently equipped with can only be regarded as L1/L2 level automatic driving.
Among the mass-produced models that have been released, only the Audi A8 is in L3, and Tesla's Autopilot is between L2 and L3. There is no mass-produced car of L4 and above.
In June of this year, Baidu launched a new generation of L4 self-driving unmanned vehicle Apollo Moon, further reducing the cost per kilometer by 60%. The cost price is 480,000 yuan, which is close to the price range of mass-produced passenger cars.
2
Commercialized level of autonomous driving:
Parking scene L4, driving L2+
Baidu is the first player to enter the autonomous driving track in China.
In 2013, Baidu's unmanned car project started; in 2017, Baidu launched the Apollo autonomous driving platform and the Apollo smart car overall solution based on this platform.
At present, Baidu Apollo has grown into the world's largest open platform for autonomous driving.
After 8 years of development, Baidu Apollo has entered the stage of large-scale commercial operation from the technical verification stage.
Judging from the newly released Apollo "automobile robot", Baidu theoretically already possesses L5 autonomous driving capabilities.
However, Baidu is still mainly engaged in research and exploration of L4 autonomous driving technology.
The second quarterly report shows that Apollo L4's accumulated autonomous driving test mileage exceeds 12 million kilometers.
Baidu Apollo has obtained 278 autonomous driving test licenses.
Li Yanhong has previously clarified the three business models of Baidu's Apollo business: providing Apollo autopilot technology solutions for host manufacturers; Baidu making cars; and sharing unmanned vehicles.
In terms of OEM empowerment, as of now, Baidu has the ability to provide autonomous driving solutions to various OEMs, including L4 autonomous driving in parking scenarios, autonomous parking AVP and L2+ autonomous driving in driving scenarios, as well as urban Pilot assisted driving ANP, and reached cooperation with 70 car companies and more than 600 models.
Leading ANP for assisted driving, Baidu claims that it is a product of L4 autopilot technology released to the field of assisted driving by dimensionality reduction. In fact, it belongs to the L2 level autopilot under the national standard.
The autonomous parking AVP, although the caliber in the propaganda is "L4 autonomous driving in limited scenarios", it is not equal to true L4 autonomous driving.
The Red Star Capital Bureau found that Weimar Motors empowered by Baidu was also marked on the official website of the Zhixing Auxiliary System.
Baidu Apollo, these vehicles known as L4 autonomous driving, some operate on fixed routes/semi-open routes and limited scenarios, and some operate on open roads, but they are still unmanned commercial demonstration projects and open test projects. And it will be taken over remotely when encountering road problems.
Currently, only pilot assisted driving ANP and autonomous parking AVP are the only autonomous driving technologies that are truly commercialized and mass-produced. Therefore, Baidu Apollo has not yet achieved true L4 autonomous driving in terms of commercialized mass production.
3
The roadside of the vehicle and the road should also be smarter
Artificial intelligence is not enough to support full-scenario unmanned autonomous driving
Level 4 autonomous driving further restricts a more stringent driving environment than L3, and requires the system to customize the related closed road environment.
The former requires high-precision maps, and Baidu is one of the few domestic manufacturers with map surveying and mapping qualifications; the latter requires road-end detection to support vehicle-end, that is, intelligent vehicle-road collaboration.
The current consensus in the industry is that bicycle intelligence can only solve 90% of autonomous driving problems, and the rest needs to be solved by road sections.
This requires that the more expensive sensing devices, artificial intelligence, and driverless technology on bicycles be transferred to the roadside, and public services may pay for them in the future.
The latest version of Baidu Apollo released last year-Apollo 6.0, which includes the V2X vehicle-road collaboration function.
On June 19, 2019, Baidu Apollo disclosed a closed-loop solution for autonomous driving pure visual city roads-Baidu Apollo Lite.
Apollo Lite is currently the only L4-level visual perception solution for urban roads in China, and it has been tested in many places and sections such as Daoxiang Lake in Beijing.
On April 9, 2020, Baidu Apollo officially released the "ACE Intelligent Transportation Engine", which is known as the first full-stack intelligent transportation solution integrating vehicle, road and line at home and abroad.
In May of this year, Baidu and Tsinghua University Intelligent Industry Research Institute released the Apollo Air plan, which will use pure roadside perception capabilities to achieve closed-loop vehicle-road coordination of L4 autonomous driving on the open road continuous road network.
Li Zhenyu, senior vice president of Baidu Group and general manager of the intelligent driving business group, said that Baidu Apollo is committed to realizing the intelligentization of all elements of transportation and "the roadside should also be smarter."
At the Baidu World Conference on August 18, Baidu Apollo demonstrated the effectiveness of its AI intelligent information control system in Baoding, Hebei.
Baidu deployed an AI intelligent signal control system at 84 intersections in Baoding, forming a "green wave belt" that shortened vehicle travel time by about 20% on average.
At present, Baidu's ACE intelligent transportation engine is cultivating multiple tracks of intelligent network connection, intelligent traffic management, intelligent highway, and intelligent parking. The related achievements have been implemented in nearly 30 cities including Beijing, Guangzhou, Shanghai, and Chongqing.
Industry insiders told Red Star Capital that at present, because artificial intelligence is not enough to support full-scenario unmanned autonomous driving, the business model has not formed a closed loop, policies and regulations have problems such as liability presumptions, and the popularization and implementation of L4 autonomous driving are still facing difficulties.
Chengdu Commercial Daily-Red Star News reporter Wu Danruo