Chapter 157 Come Prepared
Upon hearing this, Cao Shen looked at the tall man in charge of the industrial port and said:
“Whether high-precision maps are necessary depends on how far our country hopes to develop autonomous driving technology.
The tall man said to the slender eyes:
“The Society of Automotive Engineers divides autonomous driving into six levels from L0 to L5.
L0 is an auxiliary driving system without automatic driving. It has reminder and warning functions but does not interfere with manual driving.
L1 is interventional assisted driving, such as detecting obstacles ahead and providing braking, maintaining a safe distance from the vehicle in front, etc.
L2 is considered semi-autonomous driving. It can start autonomous driving under good traffic conditions, but the human driver is the main driver and the automatic car is supplemented.
L3 is interactive driving between humans and vehicles, autonomous vehicle driving, and human participation in command. The car is the main driver, with human assistance.
L4 is highly autonomous driving, which is autonomous driving under non-special circumstances or non-restricted roads.
L5 is fully autonomous driving, and the system completes all driving tasks under any circumstances, without limiting conditions and environments.
Today's most advanced ADAS assisted driving systems can support up to the L2 level, and they also use ADAS-level maps.
But if you want to go to a higher level, I think you must use high-precision maps. "
Cao Shen felt that the tall man had a very good sense and was worthy of being a leader and was professional.
After all, it is only 2012. In the original world, the penetration rate of new car ADAS pre-installation was less than 50% by 2020, and it is expected to reach more than 90% by 2030.
Now in 2012, the penetration rate of ADAS is still in single digits, and the highest penetration rate of various types of ADAS installations is below 3%.
The frown on his slender eyes did not relax as he continued:
“Why must we study autonomous driving?”
Cao Shen smiled and said:
“Generally speaking, there are four major benefits.
First, it can reduce traffic accidents. 1.2 million people around the world are killed in automobile traffic accidents every year, more than 90% of which are caused by humans.
Second, it can reduce congestion.
We have calculated that on a busy road section, the first car will start immediately even if it only stops for two seconds, and it will take at least fifteen minutes for the last car at the end of the queue to start.
We have simulated through the system and found that the use of autonomous driving can significantly reduce the phenomenon of "ghost congestion".
Because in the entire autonomous driving platform, we can automatically adjust the speed of each vehicle to maximize smooth driving on the road and avoid the entire traffic flow from being jammed.
Everyone must be familiar with the traffic situation in Beiki City. During the morning and evening rush hours, it takes at least an hour to travel half of the Third Ring Road.
As my country's urbanization process accelerates and people's economic and living standards improve, the number of car ownership must be rising.
Traffic restrictions and purchase restrictions are indeed a solution, but we may also be able to maximize the balance between the city's carrying capacity and people's living needs from a technical perspective.
Of course, reducing congestion can also reduce environmental pollution.
Third, as serious as congestion is the parking problem.
We have calculated that using autonomous driving can reduce the space reserved for people by 10 centimeters on each side, saving 1.95 square meters per parking space.
Moreover, the height of the parking lot can also be designed according to the height of the car.
If it is an automatic parking and traditional parking shared garage, each garage can save 28% of the area.
If all parking is automatic, the area of each garage can be saved by more than 60%.
Not only does it save land, it can park more cars, but it also saves time in accessing and retrieving cars.
Fourth, it can reduce driving time. In the United States, which has a relatively high car penetration rate, the average commuting time per capita is more than 50 minutes a day.
If you adopt autonomous driving, you can use this time on more interesting things.
In addition, in addition to the above values, unmanned driving can also have special functions in the military and some special industries.
For example, mining and firefighting can reduce human participation in high-risk scenarios. "
military……
His slender eyes narrowed.
The handsome uncle's eyes also lit up, but he didn't say anything. He had always adhered to what he said when he first came here, but he just didn't listen and spoke.
The tall man said: "Autonomous driving technology is definitely something our country wants to develop vigorously. Cao Shen, have you ever calculated how much it costs to make a high-precision map of the country's main roads?"
Cao Shen: Can I say that it only cost 500 yuan...
"Calculated!" Cao Shen took a sip of tea and continued:
“The accuracy of ADAS-level high-precision maps is about 50cm, and the accuracy of high-precision maps is about 10cm.
Therefore, we must have professional high-precision mapping vehicles.
A standard car requires three 360-degree panoramic cameras and one front-facing industrial camera.
Two rotating lidars are used to scan 700,000 points on each target within a 90-meter range.
Although lidar has high accuracy, its wavelength is much smaller than millimeter-wave radar. Compared with millimeter-wave radar, which can easily exceed the detection range of 200 meters, lidar performance is difficult to reach 150 meters.
Moreover, there is a high possibility that the laser radar will fail when encountering special weather, such as haze, heavy rain, etc.
Therefore, our collection vehicle will also be equipped with two millimeter wave radars. Especially for car following scenarios on highways, millimeter wave radars can do a better job.
In addition, there are gyroscopes and GPS systems for calculating spacing. The car is also equipped with at least a 1TB hard drive for storing data, as well as a computer for running analysis software.
There are also some temperature and humidity sensors.
In general, the cost of a professional mapping vehicle is about 10 million yuan.
If all the main roads in China are run down and meet the update requirements, we have calculated that at least ten vehicles are needed.
Of course, the data collected by our own vehicles can only serve as the "base" of the entire high-precision map.
In the future, a large amount of "crowdsourcing" collection will be required to improve the map, especially to meet the update requirements.
In terms of "crowdsourcing" collection, we will lower the equipment requirements and equip it with cameras of hundreds of yuan and millimeter-wave radars of thousands of yuan.
Only with low cost can there be more crowdsourcing forces to complete this task together.
Of course, the data sent back at such a low cost must have poor scanning accuracy, which puts high demands on our later algorithms.
For crowdsourcing vehicles, of course, the more the better. In addition to paying for their equipment and equipment installation costs, we will also pay corresponding remuneration for each piece of data.
The total cost should be similar to that of our professional vehicles, or even higher, after all, the quantity is large.
However, this is only the preliminary data collection work.
For high-precision map production, the later maintenance and update is the core competitive point.
The coexistence of dynamic and static information in high-precision maps determines that the later update and maintenance will occupy a larger workload.
According to the definition proposed by Bosch in 2007, the local dynamic map data required in the era of unmanned driving can be divided into four categories according to the update frequency:
Permanent static data, update frequency is 1 month,
Semi-permanent dynamic data, update frequency is 1 hour,
Semi-dynamic data, update frequency is 1 minute,
Dynamic data, update frequency is 1 second.
Compared with the current update frequency of one to two months for ordinary electronic navigation maps, the update frequency of high-precision maps is extremely high, and the difficulty is naturally extremely great.
This means a lot of work for us.
However, Xishen Technology is currently operating well, and we can afford this part of the cost, please rest assured, leaders. ”
The tall man nodded kindly:
“It seems that Mr. Cao has really done a lot of preparation for autonomous driving, which is great!
He is also willing to invest heavily in the advancement of national technology, which is not bad. ”
In the eyes of the tall man, autonomous driving is a long way to go.
If it can't be done, the money and energy will be wasted.
If it is not for the large investment of the country, most companies are unwilling to do this kind of thing.
But what the tall man doesn't know is that because there is no Tesla in this world, this industry is not very hot.
In the original world, because of the effect of Tesla, autonomous driving is the "new entrance" that everyone is vying for.
Among the several large domestic manufacturers of high-precision maps, AutoNavi Maps was acquired by Alibaba.
Baidu Maps is doing it itself.
The boss behind NavInfo is Xiao Ma's father.
It can be said that the shadows of BAT stand behind high-precision maps.
Cao Shen touched the back of his head a little embarrassedly:
“To do something, you have to be fully prepared. We have been studying for a long time. It depends on whether the country can open up the qualifications for us to implement these studies. "
The tall man looked at the thin-eyed and the sharp-tongued old man and raised his chin and smiled:
"We are prepared, Director Cheng, Engineer Wang, what do you think?"
Old man Yu also smiled at the two and said:
"Good stuff, if you are worried about confidentiality, see if Xiao Cao can come up with any compromise.
As far as I know, the United States is very open in this regard, and it has attracted a large number of talents.
We are already a generation behind in the automotive industry. If we fall behind again, we will not be able to catch up. "
The dark-faced man also urged: "What else can we think about? This is not okay, that is not okay, and we are not allowed to develop!
And it is only open to Cao Shen's family, not everyone! "
Cao Shen looked at the two and said: "How about the country encrypting the original map, and then issuing authorization codes to car owners for decryption?"
The thin-eyed nodded:
"We will study this method internally. As for whether we can open the surveying and mapping qualifications to you, I also have to ask the superiors for instructions.
After all, this is related to major security issues, so we must be cautious.
In addition, I don’t know what Mr. Wang means. ”
The old man nodded:
“The things are really good and very practical.
Since there are entrepreneurs willing to spend money to do it, we don’t need to obstruct it within the scope of safety.
After all, to promote the progress of national technology, the upgrading of industries really depends on a lot of social forces.
The times are different, and many companies have the ability to master more advanced technologies. We must support and cooperate. ”
This is considered a relief. Xiao Gang nodded to Cao Shen, indicating that he would help Cao Shen with the follow-up matters.
The tall man also winked at Cao Shen, and it seems that he will help in the future. The tall man continued:
“We assume that the problem of high-precision maps has been solved.
I now want to know the next steps.
Autonomous driving requires the system to complete the three steps of perception, decision-making, and execution, and all of them are executed and dispatched in a very short time. This requires powerful computing power and extreme control over each part.
Can the existing electronic and electrical architecture of automobiles achieve this kind of control?
The existing central processing unit may not be able to meet such performance...
In other words, the structure and hardware may not be able to keep up.
Moreover, for car manufacturers, in addition to opening up many interfaces and control permissions, it is hard to say who is responsible for problems when a third-party "autonomous driving" controls their cars.
Have you considered the practical problems in these aspects? "
As expected of a leader in charge of the industrial sector, this question is already quite professional.
Cao Shen sat up straight, here it comes!
Today's really important things are about to begin!