Editor's note: "He Xiongsong takes you to read the commercialization of autonomous driving scenarios", takes you to understand the commercialization of different scenarios in the field of automatic driving, and objectively and rationally interprets the evolution of investment strategies behind commercialization.
This column is jointly produced by He Xiongsong, executive general manager of Chentao Capital, and the heart of the car, and is updated every Saturday, and the content is exclusively authorized to be released by the heart of the car.
Today is the first lecture in this series, and then we will enter the commercial scenario of autonomous driving.
Let's take a look at the ups and downs of the automatic driving industry for more than 12 years, why is the real automatic driving still not landed in batches? How far away is it from us? With these questions in mind, let's explore them together.
01
The industry has experienced several ups and downs
First, let's briefly review the development of the autonomous driving industry.
In 2009, Google began to formally develop autonomous driving technology, which is an important symbol of the industry's start. After that, the industry has not yet entered the public eye. Until 2014, AI technology, especially image recognition, made a major breakthrough, allowing everyone to see the hope of automatic driving landing.
In 2014-2016, there was a small climax in autonomous driving. Startups have sprung up, and many self-driving companies such as Zoox and Pony were founded at this stage, and Tesla launched Autopilot 1.0 version at the same time.
Traditional OEMs have also made many moves, such as GM's acquisition of Cruise, BMW's formation of an autonomous driving research and development alliance with Intel and Mobileye.
After several years of rapid development, the continuous rapid progress did not appear, 2018-2019, the development of automatic driving has entered a low tide period, the domestic Roadstar due to infighting stopped operating, foreign trunk automatic driving company Starsky declared bankruptcy.
Why?
On the one hand, the previous industry expectations were too optimistic, while the actual landing progress was not so fast, and the capital level was tightened, and the financing of various companies was affected to a certain extent.
Until 2020, especially the second half of the year, the industry picked up rapidly, with several factors:
First, benefit from the development of Tesla and the listing of related companies such as Tucson;
Second, the capital side is picking up;
Third, some scenes, especially specific scenes, have gradually seen the dawn of landing.
Because of these factors, the industry has once again shown a thriving scene.
The above is the general development of the autonomous driving industry in recent years. Next, we will further discuss when the automatic driving will land, in order to answer this question, we will analyze the main factors affecting the landing.
02
Landing is influenced by security, business models and right of way
Both NHTSA and SAE have specific definitions of autonomous driving, and it needs to be made clear here that the "autonomous driving" we mention here and beyond is limited to L3 levels and above.
In addition, when we say landing, we mean that we have reached a stage that can be copied in batches, which is different from the progress of the Demo nature. We believe that there are three important factors that affect the landing of autonomous driving.
The first is whether security is sufficient.
No product is absolutely safe, and autonomous vehicles are no exception. So, we need to set a standard, which is very important, but at the moment no one can accurately define what kind of safety standards are acceptable to everyone.
For self-driving vehicles in open scenes, if the accident rate of autonomous vehicles is lower than that of people, we will be more assured of the safety of autonomous vehicles, according to Musk, when the accident rate of autonomous vehicles is one-tenth of that of human drivers, safety concerns can be released.
For specific scenarios, the definition of security can be more flexible.
For example, in mines, if the operation of autonomous vehicles is isolated from workers to ensure that even if an accident does not lead to casualties, the standard of safety can be greatly reduced, and only the economic loss caused by safety needs to be controlled.
The second is whether the business model is reasonable.
In the commercial scenario, the essence of autonomous vehicles is production tools, and the core of production tools to be promoted is whether they can achieve cost reduction and efficiency, so the economic benefits of products should be considered.
According to industry practice, if the payback period is less than two years, the product will have the basis for large-scale promotion.
The main economic value of autonomous vehicles lies in saving drivers through autonomous driving kits. If the more drivers can be saved, and the higher the driver's salary, the greater the economic value of autonomous driving.
In scenarios such as mines and ports, a vehicle needs to be equipped with 2-3 drivers. Once unmanned, the business model is very reasonable.
There are two trends that are critical to judging business models:
First, the cost of autonomous driving kits will continue to decline as the industry matures.
Second, the mainland's demographic dividend will gradually disappear, and labor costs will continue to increase.
Under this circumstance, the commercial value of autonomous driving will become more and more obvious.
Finally, there is the right of way.
Driverless vehicles cannot be on the road at will. On July 5, 2017, Baidu CEO Robin Li received a fine shortly after video broadcasting the live broadcast of a driverless car driving onto Beijing's Fifth Ring Road.
The state has strict regulations on motor vehicles, and the legality of driverless vehicles on the road needs to be confirmed by legislation.
The issue of right-of-way is more prominent for self-driving companies in open scenarios.
But we are generally optimistic, mainland policies are very supportive of the development of autonomous driving technology, Shenzhen, Beijing and other places have been actively legislating, and when the technology matures, we believe that the right of way will not be a limiting factor.
On the other hand, for self-driving companies in specific scenarios, the right of way has little impact. This is also the main factor that we are optimistic about the landing of specific scenarios first.
This is our analysis of the factors affecting the landing of the autonomous driving industry, which are safety, business models and right of way.
03
Specific scenes are easier to land on
Based on the above analysis, we can draw some conclusions:
1. It is difficult to carry people than to carry goods.
Compared with the load, the passenger needs to consider the comfort of the passenger, the control algorithm and other requirements are higher, and in order to take care of the safety of the people on the car, there will be higher requirements in the stability and structure of the body.
In contrast, the control algorithm and body safety requirements of autonomous vehicles are much lower.
2, high speed is difficult than low speed.
From a safety point of view, the technical difficulty of high-speed autonomous driving is much higher than that of low-speed.
High-speed vehicles require longer braking distances, so autonomous vehicles need to perceive longer distances and need to make more complex motion predictions about surrounding objects.
At present, the sensing distance of sensors in autonomous vehicles is limited, and the ability to perceive distant objects is greatly reduced compared to nearby.
Motion prediction is affected by time, the closer the time, the more accurate, the longer the time, the accuracy inevitably decreases, and safety will be challenged.
3. Open scenes are difficult to specific scenes.
The landing of specific scenes is faster than that of open scenes. The participants on the road of a specific scene are controllable and relatively limited, and there are fewer long-tail scenes, which is much less difficult to achieve technically. In addition, autonomous driving in specific scenarios has fewer obstacles in terms of right-of-way.
From the above qualitative analysis, it is not difficult to conclude that specific scenarios such as mines, ports, sweepers, and unmanned logistics cars are easier to land than scenarios such as trunk logistics and taxis.
04
Actual landing situation and outlook
What is the actual progress?
1. Mines. Do not involve the right of way issue, the road can be planned by themselves, the environment is relatively simple, the fault tolerance rate is very high, even if the overturned car may not have too much loss, no major safety hazards, the entire landing environment is very friendly.
2. Port. The road is highly structured, the environment is simple, and the right of way is relatively easy to solve.
3. Sweep. Slow speed, low technical requirements, sanitation worker recruitment and safety is a pain point, unmanned is an inevitable trend.
4. Unmanned logistics trolley. Since 2021, the industry's top companies such as White Rhinoceros, JD.com, Cainiao, etc. have removed security officers and are currently in the stage of gradual release.
The above scenarios are expected to be implemented in batches in 2-3 years.
5, Robotaxi runs on an open road, the environment is very complex. The maximum speed is more than 100 kilometers per hour, the braking distance is long, and the pre-judgment ability of the car and the perceived distance of the sensor put forward high requirements.
There are many long-tail scenes, limited by the sensor and braking distance, many problems are difficult to solve, and it is difficult to land in the short term.
At present, there is no guarantee that a clear landing node can be determined. What to do? Technical difficulties can be reduced by reducing the dimensionality of the environment (so that it is expected to see landing within 5 years):
1) Reduce the speed: from more than 100 km / h to 40-50 km / h, the technical requirements are exponential decline;
2) Restrictions on the road: Pick out the road section where unmanned driving is relatively good to land, and try it first. If there is a good platform, enough users can even operate unmanned vehicles in a hybrid way.
6. Trunk line transportation. No one can guarantee a completely unmanned time. Heavy trucks have longer braking distances, higher accident hazards, extremely low fault tolerance and higher technical difficulty.
The advantage is that the road environment on the highway is relatively simple, so it has its own advantages and disadvantages compared to Robotaxi. It is more difficult to achieve L4 in the short term; when the scene really realizes L4, the market space for the trunk line is larger in the long run.
The domestic trunk line heavy truck stock is 7-8 million units, and there is a trillion potential market space of trillions from the operational point of view, which is also the reason for the high valuation of related targets.
It is expected that in the future, the market will not be all-you-can-eat, and several points can also make each family share nearly a trillion market.
The above is my introduction to the actual progress of the landing.
Okay, I'll stop here with that. In this speech, I mainly shared the actual 6 landing scenarios in the field of automatic driving, and looked forward to the time when they expected to land in batches, and I think that specific scenarios will be the first to land.
Here is also a preview, the next lecture I will introduce the opportunities and challenges of the autonomous driving industry.
Finally, I would like to invite you to talk about what you think about the landing of commercial scenarios in the field of autonomous driving? Which landing scenario are you more optimistic about the commercialization of the field of autonomous driving? We look forward to your sharing in the message area.
That's all for this talk, and we'll see you next.
Draw the point
· Safety, business model, and right-of-way are the main factors affecting the landing of automatic driving;
· Specific scenarios such as mines, ports, sweepers, and unmanned logistics cars are easier to land than scenarios such as trunk logistics and taxis.
Produced in this issue
Speaker: He Xiongsong Producer: Zhu Shan
Editor: Ye Fang Later: Zhu Shan
Design: Chen Xiyang Operation: Lighthouse
Audio platform
Himalayas | microcosm
Podcasts | Little Goose Pass
Search for the heart of the car on the above platform
to listen to all the shows
"The Heart of the Car · The connoisseur said" trailer
Thursday, March 10, 20:00-21:00 pm, E-Cafe CTO Bai Junbo will be a guest car heart · The connoisseur said, "What rhythm will the hot skateboard site land?"