Driving with automatic driving can make Weilai better?
In the survival knockout, the automatic driving schedule of the new car company began to accelerate. Xiaopeng Automobile just said that it will launch the Leve 3 autopilot function in 2020, and Weilai announced that it will mass produce Level 4 autopilots in 2022.
On November 5, Weilai and Intel’s autonomous driving technology company Mobileye announced a strategic cooperation. The two sides will build an L4-class autonomous driving model based on Weilai’s second-generation vehicle platform. At the same time, Mobileye will also purchase autonomous self-driving cars specially designed for its smart travel service.
Mobileye is an Israeli autopilot technology company that was acquired by Intel in August 2017 for $15.3 billion, making it a key player in the autopilot field.
Mobileye announced the landing schedule for cooperation with Weilai in the official Twitter: In 2022, Weilai launched Level 4 autopilot. Based on this, Mobileye will also consume in 2022. The level of self-driving cars is placed in the Chinese market and provides travel services. When the news came out, Wei once rose 37%, sweeping the stock price cloud that lasted for weeks.
Intel Response
But this collaboration is still full of challenges.
“This is the first Level 4 autonomous vehicle for the consumer market in the same category of products, and will be produced in accordance with automotive quality standards, cost and volume production requirements.“Intel described this collaboration in its official website.
Indeed, Level 4 autopilots currently in operation in the market are demonstration vehicles equipped with expensive laser radar and industrial computer, and L4 autopilots that are actually produced according to automotive engineering standards. Not yet available. Even the self-driving leader Waymo is also working in Michigan after purchasing Jaguar i-Pace and Fitte Pacifica.The batch “assembly” of the factory did not realize the pre-assembly mass production in the true sense.
The head of an autopilot product from a car company told the L4 that the production car is very different from the demo autonomous car, “demonstration and research autopilot The main consideration is the hardware performance of the algorithm, and In addition to considering the system development, the car company should also consider the cost trend (optimal cost point), durability, after-sales cycle, insurance …etc.”
Li Tianshu, assistant vice president and second-generation platform leader of Weilai Automobile, accepted the interview. He admitted that Weilai’s requirements for the mass production time of the second-generation platform are urgent, after considering many autopilot companies. I chose Mobileye. “Maybe many companies can implement L4 technology, but in terms of mass production hardware and experience, Mobileye has more accumulation.”
In the Mobileye announcement, it will provide an L4 autopilot development kit for Weilai, including three major parts, camera, millimeter wave radar, laser radar, GPS and other hardware. These hardware will run on the EyeQ5 chip. This hardware solution will also be designed according to low-cost and low-power solutions, especially the deep learning algorithms and performance required for autonomous driving;security Standard Model RSS; Driving Policy Driving Policy.
In the official description of Weilai, Mobileye is responsible for providing system chips and related software, and Weilai is responsible for vehicle-level autopilot system development, integration and large-scale models. Mass production.
“ Algorithms, chips, and peripheral systems, both parties will participate in the selection and deployment. For example, the camera solution and laser radar should be designed together, and no one will be distinguished. “Li Tianshu said that although Mobileye is a company that is known for its camera solution, in the level 4 perception, redundancy is required, so both parties will consider lidar applications.”
After Tesla’s localization, Model 3’s biggest selling point is concentrated on FSD and AP systems, and from the comprehensive cost performance, Model 3 and Weilai’s ES6 overlap in consumer population and price. The place. This is enough to make Wei are vigilant. Only by accelerating the landing of autonomous driving technology can we build a moat for the follow-up products.
However, it is worth noting that the cooperation between Weilai and Mobileye has not been strongly bound by paper agreement,this is the rush of time. Said to be a potential unrest factor. Previously, Mobileye had already had experience in autopilot cooperation with BMW/Nissan/Volkswagen in Israel/Europe. How to keep up with the progress has become the next difficulty for Wei.
“This time there was no signing of an exclusive agreement,” Li Tianshu said. “But the two sides have a lot of details on the timetable, delivery quality, etc., and we have focused on the first batch of China. And our products are mainly based on this kind of agreement, and the effect of exclusive cooperation can be formed in a short time.”
According to Weilai’s plan, the first L4 model will be launched in 2020. In fact, the L4 level of autonomous driving requires the vehicle to travel during the process of without the driver’s intervention. The vehicle can make an independent judgment based on the road conditions. This requires the vehicle to be in various situations. The environment must be recognized, and there are opinions in the industry that it is difficult for L4 to land in 2023.
Li Tianshu also admits that everyone in the industry is now challenging the timetable. This timetable is the result of the negotiations between the two parties. The basic conditions of Weilai L4 are available, but all ODDs (limited areas of travel) in 2020 It is absolutely unrealistic to fully open. “Automatic driving has a variety of ODD elements such as urban loop, intercity highway, underground garage, and up and down ramps. It must be gradually introduced in a limited area. This is on the Weilai ES8, there are already many ways to launch features in stages.”
The idea of Weilai is very good, but the automatic driving is not the awakening of the Wei family. Prior to this, research on global autonomous driving technology has formed an alliance. In the camp, apart from car companies, the car companies formed a combination of Ford and Volkswagen ARGO AI, Honda and GM’s Cruise Alliance, as well as technology companies, including Waymo and FCA/Jaguar Land Rover, Uber and Toyota. / Volvo, in the domestic and Baidu’s Apollo and a group of car companies (FAW / Changan / Great Wall / Volkswagen, etc.) alliance.
In the face of challenges, Li Tianshu said that Weilai is also improving its output efficiency by aligning with Mobileye, diluting algorithms and technology investments, “Our goal is to be the first company in the world to deliver L4 autonomous models.”
The following is an interview with Li Tianshu:
Q:What characteristics of Mobileye will attract Weilai?
A: Objectively evaluate the overall situation. Many companies can also do a good algorithm in the early stage. The demo looks good, and the timetable we can basically satisfy, but Perhaps the determination of foreign strategy in this regard should be said to be the most determined. Inside, I feel that we try to avoid comparison in today’s discussion, because it is very likely to hurt friendly forces, but in general I think mobileye’s determination and strength in mass production must be the strongest. Whether it is said that he is at level 2 or even the penetration rate behind, or the determination to invest in the future.
In fact, he and Intel actually did a big transformation, that is, from the simple to provide the chip plus algorithm tier 1 and even the role of tier 2, he hopes to be able to do it on the value chain A transformation, not only to provide such a low-level chip, may also involve deep participation in the application layer. The wisdom of traveling is what they think may be a direction for the next step. In fact, at present, there are many travel giants in the world who are looking for opportunities in autonomous driving in the reverse direction, and technically accumulating. Then, in this regard, mobileye has a natural huge advantage, it is a prior art. To force such a job, whether it is travel or consumer market, he has actually solved the premise of a huge threshold for him.
Q: It is invested in this part of mass production. What factors should be considered different from those considered by waymo?
A: I think that everyone is also a dimension from visual to fused to complementary computing. In general, there is not much difference in the logic of everyone, but the most attractive thing for mobileye is its mass production. In fact, the mass production I said refers to mass production, and one refers to efficiency. In fact, mobileye in general, its chip is actually from the perspective of computing power, including the deployment of the cloud, its The entire input level is not that big, because it actually uses a very efficient calculation to achieve professional, more efficient calculations, so in this case, its efficiency is actually Relatively high.
In the future, if you do real mass production, there is actually a large part of the cost. On the one hand, the cost of development, as well as the cost chip of the hardware platform itself, development.The cost of the algorithm, including the cost of the cloud deployed to the future cloud location, and the cost of the data, is that his level of leanness is extremely high.
I don’t think it’s too much to describe his technical details, but I think that the display of such strengths like this may also be an important advantage for his overall efficiency improvement. In the future, I feel that it is not only a matter of calculus, but what is more critical is that it will be cheaper, more efficient and faster after mass production.
Q: Actually, Mobileye is not so dependent on deep learning. That is to say, in the system we are cooperating with, in fact, the application of deep learning algorithms is not based on the city’s mentality, or a safe and stable one. Mathematical calculation of traditional algorithms?
I think the visual perception scheme is actually the best in efficiency. Now everyone is doing machine vision based on the camera. On this basis, everyone’s understanding may be different.
For example, like waymo, the number of sensors deployed is very large. It starts early, and may have some limitations in all aspects of hardware selection. It takes a lot of sensors. Ways. With the future development, in fact, we still have to consider the optimization of autonomous vehicles. Now, if you look at it, the visual solution is definitely more efficient.
We have included Lidar in this program. This is a direction we believe. The specific technical details will be slowly disclosed later.
Q: What kind of relationship is your alliance now? Can other alliances join in the future?
A: In the global context, after all, he has already done some small robo taxi demos with the public on the Israeli side. In Europe, BMW’s EyeQ5-based chips have also been in deep cooperation. This time it was the first application for both parties and it was the first partnership.
From a global perspective, our second-generation platform will be deployed simultaneously, so this cooperation is based on China, but with global cooperation, then the cooperation process must not emphasize the so-called Exclusive and exclusive. The alliance must be to take advantage of the technical side and the brand side to improve its efficiency in the overall data output and algorithm polishing.
Q: So I didn’t sign some exclusive agreements, right? Because if there is a strong binding, it will allow everyone to spend more on the project itself.
A: There really is no. There are many ways we can do this, including the timetable, the quality of delivery, and the priority of the actions of both parties. In fact, there will be many details of the agreement and detail frame design. From a big logical point of view, we hope that both sides can focus and then strive to achieve the firstBatch, then mainly in China, with our products as the mainstay.
Q: The progress announced by Weilai is to land in L4 in 2020. Is this progress too radical?
A: Actually everyone in the industry is challenging the timetable. The reason why our timetable is released is the consensus formed by both parties. It is a single judgment. In addition, the basic conditions of our Level 4 are actually available, but it is absolutely true that all ODD regions in 2022 are fully open to do. Reality, so it is still conditional to put in a limited area.
In fact, at present, like ES8, there is definitely a way to gradually introduce and open in stages, but what may be required for robotaxi is to be fully open in a limited area, each with its own requirements. But as long as it has an ODD concept in it, how much he can open in 2020
Q: Equivalent to the fact that the L4 class vehicle may have been launched in 2022, but this function is not a completely open concept that we understand.
A: Yes, but L4 can be applied to underground garages, urban road traffic lights, Shangyu Road, urban ring roads, highways, and many other elements of geofencing. The concept of launch, but once we deliver the vehicle, we say it is L4, he must not be a L4 but just a market thing, we definitely do not want to do this thing.
Q: Is the alliance only a relationship between the two parties, or will they jointly promote industry standards?
A: There will be joint promotion of industry standards. At present, in the United States and Europe, in the process of setting standards for legislation like autonomous driving, China is just beginning. Regardless of standards and regulations, this aspect must be hand in hand to develop some standards, including China and global. This also requires a lot of other oems, including deep participation in some standards-setting committees.
Q: Will other OEMs or Tier 1 be introduced next?
A: Tier1 I don’t think it’s the point. Maybe it’s like OEM. In fact, MOBILEYE is already working together, but our goal is to focus on whether we can deliver the first or the first batch worldwide. The joint development of the autopilot set is our focus.