1. Home
  2. >
  3. Car News
  4. >
  5. New Energy Vehicles
  6. >
  7. Xiaopeng Motors Xia Heng:...

[YesAuto News] From June 5th to 6th, the 2018 (Ninth) Global Automotive Forum co-sponsored by the China Council for the Promotion of International Trade Automobile Industry Committee, the China Chamber of International Commerce Automobile Industry Chamber of Commerce and the Chongqing Auto Show Organizing Committee was held at Chongqing Yuelai Held at the International Conference Center. The theme of this forum is “Shaping the future of the industry in an era of change”. Guests from the automotive industry from all over the world gathered together to discuss new changes and trends in the automotive industry under the new generation of technological revolution. The following is a partial speech record of Xia Heng, co-founder and president of Xiaopeng Motors:

Intelligence and AI are the real driving forces of the automotive industry

Dear guests, good morning! When I was in the audience just now, I was actually communicating with President Su. Our topic today is to promote the next round of development of new energy vehicles.

In fact, when it comes to electric vehicles, it is not a very new product. The development of electric vehicles has been around for a long time. Nowadays, a very big bottleneck restricting the industrialization and mass commercialization of electric vehicles is cost. The cost of our batteries and the cost of our complete vehicles have always been a very important part of the competition between electric vehicles and fuel vehicles. So for Xiaopeng Motors, if we think about the next round of new energy changes purely from an electric perspective, the space we can manipulate is actually very small. Why? In fact, from the perspective of the current technology, the maintenance of pure electric drive shows that our electric vehicles are relatively inconvenient to use, with mileage constraints, charging facilities constraints, and current maturity of electric drive systems. The second is the cost we are talking about now. It is indeed to be promoted now. There is a cost constraint to allow people and private users to use electric vehicles in large quantities. So when Xiaopeng Motors started to do this in 2014, we were not thinking about the next round of the automotive industry from the perspective of electric vehicles. What do we think will be the real driving variables in the next round of the automotive industry? We think it is intelligence and AI. So if we look at the next round of opportunities from the perspective of a broader industry and technology community, how do we think about it?

The first dynasty hardware, the second dynasty PC+software, the third dynasty PC Internet, and the fourth dynasty mobile Internet. Every dynasty gave birth to some great companies, such as hardware companies IBM, Lenovo, PC Internet companies Alibaba, Google, mobile Internet giant Apple, now Xiaomi, which is about to go public, and UC created by our chairman, Mr. He Xiaopeng. We think AI+Internet+hardware will be a huge opportunity in the follow-up, so we believe that AI capability is a new height for new energy vehicles.

In the early days of new energy vehicles, like Apple mobile phones, many people had a lot of complaints when they first used Apple mobile phones, but Apple mobile phones brought some new things. Last time some of our Internet classmates recalled the entire development history of WeChat. There were a lot of complaints when WeChat was first used, but WeChat brought new things. I just talked about new energy vehicles. As a new thing, if compared with traditional fuel vehicles, there are some shortcomings, but the empowerment of AI will definitely lead to new heights. Just like the transition from feature phones to smart phones, the addition of AI will transform new energy vehicles and smart cars. Speaking of AI capabilities, smart cars are now a field of divergent opinions, and everyone is talking about this topic. We understand that the most important AI capability in the early stage of a smart car is that it can be upgraded and grown into autonomous driving and connected capabilities. How to understand? Although traditional cars now have a lot of intelligence and networking capabilities, it is often difficult to learn by themselves and to evolve and grow quickly. If you want to learn and grow quickly for the entire vehicle, just graft some smart functions on a traditional car, just like when you configure a car, you need to configure a leather sofa and a large-size tire. , We think it is very difficult to have prospects.

If we really want to infuse AI capabilities into the car, we must make the entire electronic and electrical organization of this car, the layout of the entire vehicle and the thinking of developing the hardware platform, all based on an upgradeable and growing intelligent autonomous driving capability and service. We believe that a combination of software and hardware must be used to develop smart cars so that the AI capabilities of smart cars can be reflected.

Four thoughts on AI Car

On AI Car, Xiaopeng Motors has four very core thoughts on this.

First, we must use high-performance chips + independent innovative software development to develop our entire autonomous driving system. In other words, in terms of chips, we still have to use the best products, but we can do a lot of innovative work in the software field. Xiaopeng Motors has established a very large intelligent networked development team, now there are about 100 people, and strives to establish a Sino-US joint intelligent software development team of more than 600 people in three years.

The second is massive Chinese scene big data. We think that there is a great opportunity for intelligent network connection and intelligent driving. The reason why China's Internet can be so successful is to a large extent. Data is very important. Chinese drivers’ habits are different, China’s road environment is different, China’s traffic signs are different, China’s traffic and tourism are different, and the city layout is different, giving autonomous driving and intelligent networking a huge opportunity. Work based on massive data development.

The third is six-dimensional perception. Six-dimensional perception, including peripheral perception and precision perception, greatly enhances the ability to perceive the external environment and makes the car smarter.

The fourth is the safety design of the system. Autonomous driving involves personal safety. How to implement redundant design for functional safety and safety of people and vehicles is a part of our automatic driving considerations.

Specifically, the first product of Xiaopeng Motors also uses Internet thinking: we must first find out one point and make the product experience the ultimate. Which point? It is automatic parking. How to use what users are willing to use, easy to use, suitable for Chinese people, I have done a few points. One is the parking space recognition of the whole scene. Our common automatic parking recognition is the recognition of a side position and a parking space. If there is a parking space, an oblique parking space, or even some free parking spaces, this has been solved. Fully automatic parking does not require manual intervention. Key voice remote control parking, 40cm parking distance, now in the benchmark design, the best parking distance in the whole industry can be achieved. The memory of parking spaces, we have parking spaces and parking spaces that we often park, how to remember them to achieve a better parking experience, this is what we do.

In addition to automatic parking, another pain point we hit is charging. As mentioned at the beginning, the popularization of electric vehicles and new energy in large quantities, and the cruising range charging is a core pain point. Since it is the core pain point, it must be refined. Xiaopeng Motors needs to implement a 5-10 km fast charging network layout in Shenzhen, Guangzhou and the Pearl River Delta. It will realize 1,000 overcharging stations and 10,000 overcharging piles in 3 years and put them on the Internet platform to strive for the owners of Xiaopeng Motors. Can have a better charging experience.

Autonomous driving layout: from a single scene to multiple scenes, from low speed to high speed development

Xia Heng: In fact, apart from the automotive field, we can find that the current overall integration of artificial intelligence and industry is a very big trend. For example, from some minor aspects, all hotel parking lots now basically use intelligent recognition of license plates. It was actually relatively rare two or three years ago. Why is it now implemented? It is actually a combination of a very simple intelligent technology and some execution technology. It uses visual perception to judge your license plate number. After the system makes a judgment, open the lever. This is the chain of artificial intelligence from your perception to decision-making. In fact, there are more and more functions like this in our daily life, but we think that the car is a scene where AI is integrated or AI has the best effect in the industry. Why? Because the entire car driving process is the most complicated, it will be used more extensively in scene perception, positioning, and execution of the on-board vehicle system. The application scene itself is very complicated.

Second, President Su mentioned to other friends in the automotive industry that the next step in shared travel must be a very big trend for us. Because we found a very interesting phenomenon. Young people in first-tier cities, including young people in many large foreign cities, are now buying fewer and fewer cars, because the cost of using cars is getting higher and higher. Not only buying cars, but parking spaces are all It is a very scarce resource. Car-sharing can greatly reduce costs. Now Didi and Uber have limited management, and now they have reached the bottleneck. The next step is the combination of artificial intelligence and cars. The biggest difference between the application of artificial intelligence in cars and other cars is safety, and safety must come first. Because of the judgments I just mentioned, if the license plate is not recognized, if it doesn't work, you can find someone to drive or stop at another place. But once the eyes of the car do not detect pedestrians or obstacles, it is human life, so this is a matter of life. Therefore, in the field of autonomous driving, Xiaopeng Motors would rather be awed and conservative at all stages of implementation, rather than step into some too radical areas and directions at once. Therefore, Xiaopeng Motors believes that it is best to expand from a single scene to multiple scenes, instead of doing fully automatic driverless driving all at once. Because it's not just about algorithms, in fact, the development of autonomous driving algorithms has been similar to the present. It is more about the matching of scenes, so we must control the number of scenes in the early stage. This is the first one. The second is that our Xiaopeng Motors will expand from low speed to high speed. Because the development and application of low-speed automatic driving are relatively safe and reliable. So I think Xiaopeng Motors' view of autonomous driving must be safe in the early stage, and gradually move from a single scene to multiple scenes, from low speed to high speed. But in the long run, we are full of confidence in autonomous driving technology, because autonomous driving is not only based on “eyes” cameras, “nose” radars, and “ears” connected to the car. In fact, it has more dimensions of perception, as mentioned in the speech just now. “Six-dimensional” perception and positioning of high-precision map matching, etc., so we are full of confidence in this in the long run, but we must be cautious at this stage and ensure safety.

Talking about the revolution of automobile design: gradual “being upright is surprising”

Xia Heng: Yes, the design of smart cars is definitely different from traditional cars, but I personally think that there are more of the same from the perspective of cars. We often say that keeping upright is surprising. The part of keeping upright must adhere to the basic principles and basic ideas of traditional automobile development. In fact, the design of a car is not only an engineer’s subjective creation, but also involves the entire industrial system, especially the supply chain system. From this perspective, whether it is its own site system or electrical and electronic architecture, its development is definitely not transformational, it is definitely gradual. It involves the collaborative development of all parts companies, not a company's business. So when it comes to car design, I think it’s surprising that there are more than 10,000 parts and components assembled into 200 or 300 systems, which are basically developed by the industrial collaboration chain, and use mature technology and mature technology as much as possible. For the wire control technology mentioned earlier, there are many very excellent suppliers in the market, very reliable and long-term proven solutions, this one must be solved in cooperation with these best-in-class suppliers.

In the surprising part, do we completely follow the traditional car development? Certainly not. If this is the case, there is no need for a new company to emerge in 2014. Changan did this part of it decades ago, and Mercedes-Benz did it more than a hundred years ago. Where is the surprising part? The equipment is in the development of automotive software. In the development of automotive software, there are traditional embedded developments that use micro-processes for development. In fact, many software developments such as autonomous driving, deep learning, and intelligent networking are not applicable. They are based on rapid iteration and Internet-based product thinking. For the development of software, we must inject some new thinking and new ideas into our automotive software development. The second is the development of the entire product life cycle of the car, how to achieve upgrades, this is the point to be innovated. Because traditional cars are sales-oriented, after I sell this car to users, it actually means the end of the work of the entire project director and R&D director, but Internet products are different. After the Internet products are launched, the work of the product manager has just begun, so how do we a R&D worker continue our R&D work after product interaction? This is relatively rare in traditional car development. Because in the past, after-sales personnel gave feedback to sales, sales feedback to production, production feedback to R&D, and then iterated in the next step. Now it is different. Now all data is required to go directly to R&D. There are a large number of sensors in the car, which can detect how many kilometers the car is driven per day, how long it has been charged, how many people are there when using the car, and there are gravity sensors, etc. Is the glove box often used? The big screen APP is used for a day how many times. According to these results, through remote OTA upgrade, the car will undergo many iterations and upgrades during its life. Of course, in this process, how to ensure the safety of upgrades and how to ensure the safety of product development is a topic we want to discuss, but we must do it. Therefore, I think that smart cars are a very challenging and exciting thing for automotive practitioners. There are things that stick to the traditional thinking in this field, and there are also things that we innovate. How to integrate the two is our biggest challenge.