Job Recruitment Website - Job information - Robotaxi is already available. How far will it take for a Meituan guy to turn into a robot?

Robotaxi is already available. How far will it take for a Meituan guy to turn into a robot?

Written by?/?Tu Yanping

Editor?/?Niu Monkey Shang

They talked about Robotaxi, On the afternoon of August 11, 2020, at a roundtable forum at the 2020 12th China Automotive Blue Book Forum hosted by Sohu Technology Director Han Xiao.

These people are Wei Junqing, CTO of Didi Autonomous Driving Company, Wu Gansha, co-founder, chairman and CEO of Uisee Technology, Zhang Zhenlin, vice president of technology of Zhongzhixing, Gu Jianmin and Zongmu, CTO of Valeo Group China Technology founder and CEO Tang Rui.

Part of the reason for talking about this topic is because Robotaxi, which once felt far away, seems to have come to people.

About 4 months ago, on April 20, 2020, Baidu announced the full opening of the Apollo Robotaxi service. Changsha users can open Baidu Map to enter the taxi service or search for the "Dutaxi" applet on Baidu APP. Can be called with one click.

Then on April 27, AutoX autonomous vehicles were connected to AutoNavi and launched an experience recruitment campaign in Shanghai. Shanghai citizens open the Amap map and call AutoX’s driverless cars after successful registration.

On June 24, WeRide WeRide connected to AutoNavi and officially launched the Robotaxi service. Guangzhou citizens can call autonomous vehicles with one click through AutoNavi.

Then on June 27, Didi officially opened the Robotaxi service function to the public. Users can make an online reservation in the Didi APP, and after passing the review, they can take a test drive on the Shanghai autonomous driving test section.

McKinsey once predicted that China is likely to become the world's largest autonomous driving market in the future. By 2030, the order value of travel services based on autonomous driving will reach approximately US$260 billion.

At this China Automobile Blue Book Forum, as CTO of Didi Autonomous Driving Company, Wei Junqing said that 30,000 people have registered for Didi’s Robotaxi trial operation in Shanghai.

For the public, autonomous driving with safety personnel in a small area may be the first autonomous driving they can experience. In Shanghai, there are also commuters who use our autonomous driving network every day. ”

Wu Gansha, co-founder, chairman and CEO of Uisee Technology, believes that Didi will gradually reach more consumers through mixed order distribution and make corrections during the use process. Making mistakes and learning is the right approach. Even if Robotaxi technology has matured, mixed dispatch may still be needed.

He said: "Maybe in three to five years, robotaxi will be basically the only ones in some places. But once there is bad weather or major bugs, human-driven taxis can make up for it."

Of course, this roundtable discussion around the theme of "How autonomous driving grows in consumer perception" is not limited to Robotaxi.

Gu Jianmin, CTO of Valeo Group China, said that for B-side customers, highly automated driving equipment or vehicles are already operating in semi-enclosed scenarios. "For C-side customers, we should not overemphasize discussions at the L3, L4, and L5 levels, but should return to scenarios and functions related to the personal experience of C-side customers."

He believes that earlier The autonomous driving that C-end customers can enjoy includes parking and TJP (Traffic Jam Pilot, autonomous driving on congested road sections) on structured roads.

Wei Junqing said: "Small-scale autonomous driving on fixed routes and fixed areas is expected in the next few years. For large-scale autonomous driving to be realized, it can handle any place, environment, and scenario, and it can be handled by everyone. Within the acceptable cost range, I think it will take more than four to five years.”

There is no doubt that the safety and reliability of technology and the reduction of costs are the only way for large-scale implementation of autonomous driving applications.

Zhang Zhenlin, Vice President of Technology of China Intelligent Banking Corporation, believes that at the current stage, we must be cost-aware, but we cannot be constrained by cost. It can already be seen that "in the absence of large-scale commercialization and implementation in the entire autonomous driving industry, the price of sensors is even falling faster than Moore's Law said."

Wu Gansha said: "The technology will mature in the next few years, and it will soon be possible for configurations like Robotaxi to drop below 100,000 yuan."

Then, in the next few years, For autonomous driving applications in the passenger car field, what will the mainstream product forms look like?

Tang Rui gave this view: "L2 will be the main one on the high speed, and L3 and L4 on the low speed. These two systems may be merged into one system, which will be widely used by consumers." "

The following is the content of this roundtable discussion compiled by Automotive Business Review, which is slightly abridged.

The B-side has been implemented, and the C-side should focus on scenarios and functions

“

When we talk about it, we often confuse unmanned driving with autonomous driving. In fact, what we do more is autonomous driving, that is, L4. Our general judgment is that the B side will be faster than the C side, but consumers may not be able to feel the B side.

——Wu Gansha, co-founder, chairman and CEO of Uisee Technology

”

Han Xiao (Sohu Technology Director): Our topic today It is "How autonomous driving grows in consumer perception." Now is a critical moment. Some applications at various levels have begun to make consumers aware. There are two types of applications of autonomous driving in the transportation field. , one is for the transportation of people, and the other is for the transportation of items.

I would like to ask you a question first. What do you think will be the most effective way for consumers or users to use it in the next few years? Which direction is the application of autonomous driving with deep perception?

Wei Junqing (CTO of Didi Autonomous Driving Company): In the next period of time, I feel that true autonomous driving will be widely used in commercial applications within two to three years. It is still difficult, whether it is a breakthrough in technology, business, or policies and regulations.

We have launched a trial operation of Robotaxi in Shanghai, even with safety officers. The public is also very curious about driving, and has provided us with a lot of suggestions and encouragement. I think that for the public, autonomous driving with a safety officer in a small area may be the first autonomous driving they can experience. , commuters in Shanghai use our autonomous driving network every day.

Han Xiao: How many people use this application every year?

Wei Junqing: This data is not convenient yet. However, more than 30,000 people in Shanghai have registered for the whitelist and are using our application.

Han Xiao: Mr. Wu, what do you think of this issue?

Wu Gansha (co-founder, chairman and CEO of Uisee Technology): This question is quite difficult to answer. First of all, when we talk about it, we often confuse driverless driving with autonomous driving. In fact, we do more of autonomous driving. Human driving is L4. We have a rough judgment that the B-side will be faster than the C-side, but consumers may not necessarily feel the B-side.

Han Xiao: Will the demand for the B-side be more difficult?

Wu Gansha: From the perspective of replacing drivers, the B-side is more beneficial. The B-side is faster, and logistics is even faster.

Of course, from the perspective of autonomous driving, autonomous driving means driving with a driver. In fact, we have already experienced this product, such as Autopilot. Now some Tesla owners have enjoyed Smart Summon (smart summoning function). You can summon it when the car is parked from a distance.

I believe that an automatic valet parking product will be released in the next two years. This product will be suitable for those who really hate parking and have poor parking skills. For good consumers, the perception may be stronger.

Han Xiao: Mr. Zhang, what do you think? When we talked before, you said that L4 would be coming soon.

< p> Zhang Zhenlin (Vice President of China Intelligent Technology Technology): In fact, according to the international classification of autonomous driving, or China’s classification of autonomous driving, from L0 to L1, L2, L3, L4, and L5, in my opinion, it is the first to be perceived by consumers. The one is in L2. For high-level automatic driving assistance functions like Tesla's Autopilot, the driver still needs to bear the main responsibility. There may be many consumers who are already experiencing such autonomous driving.

Second, I am not particularly optimistic about L3. I jump directly to L4. L4 is the first to be felt. As the two bosses just mentioned, it may be able to work under specific working conditions and specific areas. It is first felt by consumers. For example, Didi is very impressive. It held L4 autonomous driving experience in Jiading, and Gansha always operates completely driverless in areas such as airports. I think overall it is better to look at different levels, so that the discussion may be more precise.

Han Xiao: Mr. Gu, what do you think?

Gu Jianmin (CTO of Valeo Group China): My general views are similar to those of the previous three guests. Let’s look at the client first. In fact, for B-side customers, highly autonomous equipment or vehicles are already operating in semi-enclosed scenes in ports, mining areas, and logistics parks. AGV is a L4 autonomous vehicle that has been operating in the park for a long time.

Looking at C-side customers, we should not overemphasize discussions at the L3, L4, and L5 levels, but should return to the scenarios and functions related to the personal experience of C-side customers. If we look at scenarios and functions, our driving assistance or highly automated driving products have already appeared.

I think there are two scenarios for autonomous driving that will be easier or earlier for C-side customers to enjoy.

One scene is parking. From parking assistance to automatic parking to remote control automatic parking, remote control automatic parking means that people can choose to be in the car or outside the car and let the vehicle automatically find a parking space to complete the parking. There is also memory parking. Generally speaking, From fixed parking space parking to AVP automatic valet parking, this route is very clear.

You can see that many vehicles, not necessarily high-end vehicles, have mass-produced parking assist. For example, Changan CX75 has mass-produced remote-controlled automatic parking. The advertisement here uses Valeo products. It can be seen that consumers are already enjoying this feature.

Another scenario is structured roads such as TJP on highways.

The traffic scene on the highway is relatively simple, and coupled with the restrictions of traffic regulations, users can completely let go of their hands and feet below a certain speed, which has been achieved on many vehicles.

In these two scenarios, C-end users are already enjoying a certain degree of autonomous driving, so in one sentence, don’t be completely obsessed with which level, L3, L4, or L5, but what you should focus on is scenes and functions.

Han Xiao: Where is Mr. Tang? Just now, Mr. Gu and Mr. Gansha both talked about automatic parking. You have been working in this area for many years, so you should have similar views, right?

Tang Rui (Founder and CEO of Zongmu Technology): The previous guests have discussed this issue more thoroughly, so I will simply add something. Gansha always mentioned that there is a distinction between autonomous driving and driverless driving. In fact, if it is further subdivided, there are assisted driving and autonomous driving, L2, L3, L4, and L5.

In the next few years, whether it is Robotaxi or logistics, if we look at passenger cars, I think the mainstream product form will be L2 at high speeds, and L3 and L4 at low speeds. The two systems may be merged into one. Such a slightly mismatched combination will become the main form of the product and be widely used by consumers.

Han Xiao: I summarize the answers just now: In some specific scenes, such as low speed, closed, and some parking scenes, we can quickly or already be able to clearly understand the situation. Perceive the convenience that autonomous driving or driverless driving brings to us, and the progress of science and technology.

Hybrid dispatch is a reasonable business path

“

We are slowly and more deeply aware of the importance of people in (autonomous driving) , I think our drivers will definitely not be replaced in the future, but they will provide higher value services

——Wei Junqing, CTO of Didi Autonomous Driving Company

”< /p>

Han Xiao: If this question is extended one step further, C-end users may be more concerned about large-scale perception such as true autonomous driving. This issue has actually been discussed for a long time. At the half-time moment, when can this thing really enter our lives? For example, will Didi have no drivers in the future? Meituan guy turned into a robot? When will these arrive?

Wei Junqing: Didi’s original intention of doing autonomous driving is to improve the safety and efficiency of the entire transportation. In the process of doing it, we have slowly and deeply realized that people are inside (autonomous driving). I think our drivers will definitely not be replaced in the future, but they will provide higher value services.

In addition, Didi’s evolution of the autonomous driving industry or the entire industry will create more and better job opportunities, including autonomous driving testers and safety officers, including future remote assistants and controllers. member. From a timeline perspective, small-scale autonomous driving on fixed routes and fixed areas is expected in the next few years. I think it will take more than four to five years before large-scale autonomous driving can be realized, in any place, in any environment, and in any scenario, and within a cost range that everyone can accept.

Han Xiao: You just talked about cost. When will the cost of autonomous driving and the cost of manned driving reach a balance? At present, the cost of autonomous driving is still relatively high.

Wei Junqing: I don’t think there is a very good answer in the industry. The predictions made by all car manufacturers and technology companies in the past 10 years have been constantly revised. Since the DARPA competition was held in 2004, I thought that autonomous driving would be popular all over the world in five years. However, after 16 years of work, we still found that there are many difficulties. And, the closer we get to productization and the ultimate goal, the greater the difficulty becomes.

From a cost perspective, I feel that we are still in the middle of the S-curve of technological evolution, and we are still making rapid progress. If we take the Robotaxi application as an example, it is not yet time for us to start tightening costs. We are still considering how to solve this problem more safely. The industry is still at this stage.

Han Xiao: Mr. Wu, what do you think?

Wu Gansha: As Jun Qing said, every prediction in history has been slapped in the face. It's really hard to say. I think that from an industry perspective, we will not stagnate just because the technology has not yet fully matured. We must find ways to get it used first. There is a saying in this technology industry called “Law of Exposure”. If you let your users use it every day, they will gradually like you more and more, and gradually become more tolerant of your mistakes. This is called the "law of exposure".

I think it is the right way to gradually expose more consumers to consumers through mixed order dispatching like Didi, and to correct mistakes and learn during the use process. And I think that even if Robotaxi is technically mature one day, it may still need a mixed dispatch situation.

Because there are two situations where it is inevitable to stop. One is extreme bad weather, and the other is the discovery of a very important bug. From the perspective of being responsible for consumers, all cars must be recalled. If it is not a mixed dispatch, your service will be in shock and your APP will be useless.

So, I think hybrid dispatch is a good way to operate technology through grayscale deployment while also meeting extreme service interruption scenarios.

Now everyone sees that mixed dispatch has appeared. Maybe in three to five years, it may be basically Robotaxi in some places. But once bad weather or major bugs occur, human-driven taxis can make up for it. This is a very reasonable future business path.

Han Xiao: What I just mentioned refers to the application on the C side. What if it is on the B side?

Wu Gansha: I think the B-side has already started this year. For example, our driverless luggage tractor at the airport has been running for 8 months, and it has shouldered the responsibility well when the epidemic was very serious. responsibility for the lack of manpower. There is no problem at all in these scenarios.

Han Xiao: Do ??these scenarios also require a hybrid deployment?

Wu Gansha: At present, it is still a mixed deployment. We believe that in half a year or a year, unmanned deployment can be achieved on a single line.

Han Xiao: In addition to airports and factories, what are the next better applications?

Wu Gansha: In fact, there are many scenarios. In addition to Robotaxi, another big market is trunk logistics. In addition to the logistics at airports and factories just mentioned, there are mines, sanitation, terminal distribution, ports, and many more. Unmanned driving in specific scenarios like this should be put into operation in the next few years. Let B-side customers experience cost reduction and efficiency increase.

I don’t think you need to worry too much about cost. In the technology industry, as long as your technology matures and your volume increases, cost will never be an issue. The key now is that the technology is not yet mature and the volume cannot be increased, so you will encounter the chicken or the egg problem. I believe that as the technology matures in the next few years, it will soon be possible for configurations like Robotaxi to drop below 100,000 yuan.

The demand for transporting goods is greater than the demand for transporting people

“

If this road takes a long time to go, do we have some other options? Assistive methods, or some alternative methods. Therefore, we have proposed a technical route for L4 autonomous driving that integrates 5G+AI bicycle intelligence and vehicle-road collaboration.

——Zhang Zhenlin, Vice President of Technology at Zhongzhixing< /p>

”

Han Xiao: I would like to ask Mr. Zhang, you have focused on L4 before. Every time I interview you, you always talk about how good L4 is. I hope you will do it again. Tell us a little bit, you are an optimist about L4. Why are you so optimistic?

Zhang Zhenlin: According to the classification of autonomous driving, our idea of ??L4 is relatively complete autonomous driving. In fact, everyone currently talks more about fully autonomous driving, and more about pure autonomous driving represented by Waymo. Intelligent fully autonomous driving of bicycles. Zhongzhixing is a very young company, established in 2018, but the team had very rich experience in the autonomous driving industry before that.

How can we realize the implementation of autonomous driving as soon as possible, like the timetable you just mentioned. Relying purely on bicycle intelligence, let's compare these data with Waymo. We just talked about Waymo? Hundreds of vehicles have been continuously evaluated since 2012, but as of today, it is very difficult for Waymo to rely purely on data to solve all conventional scenarios. .

When Zhongzhixing was founded, we were thinking about what kind of technology we should use. If this road takes a long time, do we have any other auxiliary methods or some modifications? Way. Therefore, we proposed a technical route for L4 autonomous driving that integrates 5G+AI bicycle intelligence and vehicle-road collaboration.

We have been working on bicycle intelligence. We install sensors on the vehicle, constantly do road tests, and constantly train our models. Recently, we announced our cooperation with China Telecom Group on vehicle-road collaboration and 5G. We will integrate the intelligence of bicycles with the high bandwidth, low latency, and security features of 5G to solve our need for large amounts of data accumulation to cover those scenarios. .

In March this year, China launched the concept of "new infrastructure". A very obvious feature is the vigorous deployment of 5G and its efforts on smart roads. China’s advantages in artificial intelligence, combined with China’s large investment in 5G and infrastructure construction, give us sufficient confidence that when the two technologies are integrated, through smart cars, smart roads, and 5G cloud, everyone can easily We will soon be able to enjoy L4 fully autonomous driving in certain areas.

Han Xiao: We just talked about 5G. Are you satisfied with the current deployment speed of 5G? Do you feel that you are lagging behind or unable to catch up?

Zhang Zhenlin: If you asked me this last year, because last year was the true first year of 5G, I might feel unsatisfied or lagging behind. However, this year, various operators are conducting pilot deployments of 5G networks. China Intelligent Technology Co., Ltd. is very lucky to receive strong support from the government and operators. We are also deploying 5G networks with independent main networks in many areas based on business needs. . At present, our cooperation is developing very rapidly, and overall we are moving forward rapidly, which also gives me confidence. But if 5G is fully covered, it will take a certain amount of time for all areas to need 5G networks. But from the current point of view, 5G is not an obstacle.

Han Xiao: Let me ask Mr. Gu, I saw a very interesting thing at CES before. Valeo showed an electric unmanned delivery vehicle. What attracts me more is that this project is in cooperation with Meituan. What is the current progress of this project?

Gu Jianmin: At the 2019 CES show, Valeo and Meituan signed a strategic cooperation agreement to cooperate on the development of the last-mile unmanned delivery platform, because Valeo already has an electric drive system , there are autonomous driving systems, and some other auto parts systems. A year later, at this year's CES show, the two parties launched a prototype electric unmanned logistics delivery vehicle.

The chassis, electric drive system, autonomous driving sensors and controllers of this prototype were all developed by Valeo. Meituan is responsible for the entire body distribution cabinet and APP areas. It is a relatively new company in this regard. Good at. This is a prototype car. One year seems to be a long time, but during this process both parties need to communicate, negotiate basic conditions, including mutual coordination, until the final prototype is manufactured and shipped to the United States. In fact, the real development There are only a few months left.

The original plan was that after the CES show in January, we would transport the car from the United States to Europe and then to China for subsequent display and communication development work. Unfortunately, the outbreak of the epidemic has delayed the original plan, including the postponement of the Beijing Auto Show, and restrictions on the entry and exit of personnel. Currently, the two sides are discussing what to do next.

Han Xiao: I originally wanted to ask if this project is likely to be applied on a large scale. When do you think similar projects will be applied on a large scale?

Gu Jianmin: What is on display at CES is a prototype car, and large-scale application are still two different concepts. This is a prototype car and a platform. At the same time, Valeo is also cooperating with new travel models or start-up companies other than Meituan. I think the issue of large-scale mass production should be left to Meituan or other travel service providers. Valeo provides such technology, and of course we hope that large-scale mass production can come soon.

Han Xiao: I noticed that you repeatedly mentioned in some speeches that the demand for transporting goods is actually greater than the demand for transporting people. How do you explain this problem?

Gu Jianmin: There are two major clients, B-side, such as ports, mining areas, logistics, parks, etc. You can already see that there are many applications and the demand is already there.

Let me give you an example. There are many mine cars in the mine to transport slag or mineral materials. The working conditions for these drivers driving mine carts are very difficult. First of all, it is relatively remote and remote. Secondly, the roads are completely different from those in urban areas. Many drivers have occupational diseases. We often say that taxi drivers already have a hard time, but drivers in mining areas have even more hard times. It is difficult to recruit this kind of mine truck driver. No one is willing to do this job even if the price is very high. In this case, the demand is high.

There is a more important issue in the mine, which is safety. Vehicles are a small problem. If the vehicle is damaged, it is a matter of amount. If there are casualties, the red line may be triggered, which will have a negative impact on the mine operation. direct impact.

From this perspective, the demand on the B side is there.

If we look at logistics, what many listeners may have personal experience with is express delivery or takeout in logistics. There is a very interesting figure. Meituan released a data yesterday. Meituan’s order volume in the past few days has exceeded 40 million orders. How did this data come from? In 2017, it exceeded 10 million orders, in 2018, it exceeded 20 million orders, and in 2019, it exceeded 30 million orders. The daily order volume increased by 10 million orders almost every year. The incremental market is particularly large.

Have you ever thought about the number of courier boys that have increased so much? There are other couriers, there are other operators out there. Therefore, the demand for express delivery in the entire market is there. I believe that companies like Meituan are very interested in using unmanned logistics to solve this part of the incremental market. This market exists, which does not mean that there is no market for transporting passengers. However, in many cases, transporting goods may be more practical than transporting passengers.

Safety-led and cost-conscious

The exponential decline in cost and price is the trend