A driverless taxi (robotaxi) without a safety officer, carrying Li Zhenyu, vice president of Baidu group, and Bao Xiaofeng, host of CCTV, drove nearly 700 meters in Beijing Shougang park.
This car is a test car from Baidu’s automatic driving project. Although the journey of several hundred meters is short, it reveals an important message that Baidu will become the first autonomous driving enterprise in China with no road test.
On the day of the exercise, baidu Apollo got the first “unmanned automatic driving” road test license in China. This “road test certificate” in the field of automatic driving also makes Baidu full of confidence and ambition.
Baidu CEO Robin Li said on the same day, “in 5 years, China’s first tier cities no longer need restriction and restriction, and 10 years later, they can solve the problem of traffic congestion.” So, starting with Baidu’s Robotaxi, is the domestic autopilot industry really like Robin Li’s saying that it will come into reality in 5 years?
1. Is robotaxi a “pseudo concept”?
Robotaxi has always been regarded as the main direction of landing and commercialization of autonomous driving.
Starting from Baidu’s unmanned rental car, people from all walks of life began to declare an atmosphere: “this year is the year for the full launch of robotaxi in China.”
In view of the current level of automatic driving technology and the current situation of the industry, Kungfu motor believes that robotaxi is still in the stage of “pseudo concept”.
At present, the experience products being developed all over the country are not robotaxi in the real sense, which is far away from the reality.
Why is robotaxi a “pseudo concept” in China?
First of all, many domestic robotaxi take fixed routes. If the route, starting point and destination are fixed, why not use unmanned buses with more passengers and better economy?
At the same time, some robotaxi, although the route is not fixed, can only operate in a small closed factory or park, which is more like a “ferry car” than a taxi.
Secondly, if the use environment of robotaxi is really similar to that of ordinary passenger cars, it will undoubtedly raise the single car intelligent technology or vehicle road collaboration technology to “hell level difficulty”, including safety, reliability and maturity.
In addition, if robotaxi really realizes driverless, it will not be a big problem to carry driverless technology on ordinary passenger cars.
At this time, where can I use robotaxi?
Overall, robotaxi’s problems and challenges are obvious.
2. Self driving “into reality”? Five years is not enough
Although Robin Li believes that driverless technology will achieve large-scale business in 5 years, Baidu has not given a clear answer to what business model and how to benefit.
Behind this, there are still many fundamental challenges in the field of automatic driving.
The first is sensors.
Autopilot uses a variety of sensors to “observe” the surrounding environment and help the system detect objects such as pedestrians, other vehicles and road signs.
But as far as the current technology is concerned, bad weather, heavy traffic and road signs with graffiti will have a negative impact on the recognition ability of the sensor.
Take Tesla as an example, its “autopilot” hit other parking vehicles in July this year.
This fact further proves that there is still a long way to go for sensors to cope with all-weather driving scenarios.
Secondly, it is the standard of artificial intelligence learning judgment.
In the field of automatic driving, most of them use artificial intelligence to learn to process the data from sensors, and then make specific decisions about the next action.
In the future, machines may have more efficient object detection and classification capabilities than human drivers.
But at least for the moment, the machine learning algorithm used in the car is still lack of sufficient security basis. For example, the whole automatic driving industry has not reached a consensus on how to train and test machine learning algorithms.
Moreover, in terms of regulatory requirements, there are no adequate standards and regulations.
The existing standard assumption of vehicle safety only requires the driver to take over immediately in case of emergency.
For autopilot, the regulation only provides for certain special functions, such as automatic lane keeping system.
As for the automatic driving system including autopilot, although some international standards have been set up, some problems related to sensors, machine learning and behavior learning have not been solved for the time being.
Therefore, as long as there are no universally recognized regulations and standards, automatic driving vehicles, whether safe or not, have no right to drive on open roads.
Finally, there is the social acceptance of autonomous driving.
Previously, Tesla’s automatic driving function has caused a number of accidents. These accidents lead to the gradual decline of social recognition for automatic driving.
As a result, consumers need to participate in the introduction and adoption decisions of autopilot. Without this link, the technology may be rejected by the people.
Obviously, only by solving two problems can the automatic driving industry have the opportunity to overcome the last two obstacles.
At present, all parties in the industry are striving to become the first company to launch fully automatic driving vehicles.
However, if the industry fails to reach a consensus on all aspects of the field of automatic driving, including standards and safety, the automatic driving vehicle will still be in the “testing stage” for a long time in the next few years.
3. Write at the end
Although Robin Li’s remarks are slightly “inflated”, it is undeniable that Baidu’s attempt is promoting the development of the domestic automatic driving field.
And the future space of automatic driving market is large enough.
The wind started at the end of Qingping. The core of whether the domestic automatic driving industry can be stable and far-reaching is the comprehensive progress of technology and standards.
For the current Baidu, the development of a feasible and safe AI automatic driving system is the primary goal. It may be too early to talk about landing in full.