Shulou(Shulou.com)11/24 Report--

In the field of intelligent driving, a new war around "integration of travel and parking" is brewing.

In 2021, a number of chip companies launched SoC chips designed for the integrated function of travel and parking, which will ignite the flames of war. A few months later, a number of Tier 1 have followed up with the release of integrated parking solutions, aiming to produce cars in 2023. The main engine factory is even more impatient, the relevant people said bluntly: "I can't wait for mass production tomorrow." In 2022, the integration of travel and parking officially entered the eve of large-scale mass production, and the integrated scheme has also become the hottest keyword in the autopilot circle in 2022.

Recently, Horse Zhihang and Heduo Technology have announced that they will cooperate with Horizon, a provider of intelligent driving computing solutions, to launch an integrated travel and berthing scheme based on the Horizon Journey ®series chips to accelerate the landing phase of mass production.

The so-called integrated line and berth essentially means that all the sensor data are connected to a domain controller dominated by a single SOC chip, and the intelligent driving ability is realized by a set of software algorithms. In this way, whether driving or parking, the system can call the same set of algorithms and all hardware resources, including chips and sensors for processing, rather than a separate system dedicated to parking.

Travel and parking is an active impact on the simplification and iteration of automobile architecture from a subdivided track. Especially after January 2022, Tesla was praised by Wall Street as the "biggest winner" in the automotive chip shortage disaster, because they have insisted for more than a decade that as long as one more function can be integrated into fewer chips, there will be less pain of being pinned down.

Nowadays, with the continuous catalysis of the recent self-driving policy, many middle and low-end electric vehicles have also begun to speed up the scheme of integrated parking, which has accelerated from more than 300000 to less than 150000. In other words, the integration of travel and parking may replace the original "ADAS-assisted driving" in one or two years and become the mainstream of the market.

Around this phenomenon, we try to answer the following questions in this article:

1. What is the essence of the integration of travel and berthing?

2. What is the present situation of the development of the integrated travel and berthing scheme?

3. What will be the next technological breakthrough in the industry?

1ONE, cost control "travel and parking as a whole", as a variant of auxiliary driving ability, its source is Tesla. Why would you say that? Let's start with a set of data to talk about some of the current situation in the smart driving industry.

From 2020 to June 2022, the carrying rate of high-level assisted driving in the Chinese market has been climbing rapidly, especially from January to June 2022. The front loading rate of high-level assisted driving reached 26.64%. From January to October 2022, the penetration rate of L2 auxiliary driving of passenger vehicles reached 33.9%. From January to July 2022, 2.8389 million standard L2 intelligent drivers were insured, an increase of nearly 70% over the same period last year, while the average price range of models was between 15 and 200000.

Thus it can be seen that there is an urgent demand for auxiliary driving at the level of L2Comp2.5 in the market. But we all know that in the early days of autopilot, there were gradual and leapfrog routes in the industry. Some enterprises will try their best to compress the scene, and then roll the autopilot to the L3 or L4 level in this scene. However, the applicability of this scenario is so small that it is of little value to the end user.

Tesla's autopilot function begins with a single lane, then extends to the NOA of all lanes, and then extends to the city scene. In terms of security rating, these functions are still only at L2 level, but the scene coverage is very wide, most of the scenes can be driven by the vehicle itself, and the user experience is good.

Professor Wen Yi of Tsinghua University once put forward a point of view in the book the Great Industrial Revolution in China: "the market is an expensive public good". In the book, take the Westernization Movement as an example. China not only sends a large number of talents to study abroad, but also spends 50.002 million yuan on military industry, but no one pays for the steel and guns produced. As a result, there is no profit, making it impossible to achieve self-circulation.

To sum up, only the real demand of the market can accelerate innovation.

Going back to the field of self-driving, many enterprises have gradually realized that what they need most is not the L4 to replace the driver, but the L2 and L3 to help people drive, who can provide a low-cost, high-performance, highly open system, who will have the opportunity to eat more market share.

As a result, the industry is no longer blindly pursuing a higher level of autopilot, but instead thinking about how to make autopilot run on the road. And when they realize that parking is a scene that can be popularized more quickly, more and more mainframe manufacturers begin to pay more attention to automatic parking, memory parking and autonomous parking.

No matter judging from the technical path or the business path, the integrated travel and parking scheme is undoubtedly an optimized way to solve the problem.

Previously, driving and parking were two separate systems, which were too expensive. Different from the separation of traffic and parking, the integration of traffic and parking only needs a set of hardware equipment to realize the function of driving and parking, which not only makes full use of computing power, but also reduces the complexity of hardware, and improves the development efficiency of the mainframe factory, so that the cost of self-driving function is greatly reduced. Cover the main consumer models with a price of 100-200,000 as much as possible.

Photo from: automobile House not only reduces costs, but also improves the efficiency of development. The integration of line and park adopts universal underlying basic software and standard middleware, which is suitable for different platforms. compared with distributed development, it is equivalent to twice the development cycle of underlying software and middleware. The integration of travel and parking can also make the sensor reuse and enhance the driving experience. For example, AVP (autonomous valet parking) can use the forward camera of the driving system to identify long-distance objects and supplement the short-distance perception defect of the camera looking around in the parking system.

Moreover, the integration of line and park can better support the OTA upgrade of the function. The distributed architecture is replaced by the domain-controlled architecture, and the software and hardware can be fully connected. The operation and management of the whole integrated system are more stable and efficient, and the data closed loop has more advantages. An OTA upgrade can complete the function update of driving and parking at the same time.

2TWO, accelerating getting on the bus, sinking from the time point of view, the integration of travel and parking is not a very new concept, and mass-produced models also existed two years ago. In 2020, Xiaopeng P7 realized the integration of line and berth, relying on Desai Xiwei's IPU03 domain controller. In recent years, a large number of autopilot solutions are being studied.

Back in 2016, the field of smart driving is very busy. Traditional automakers such as Volvo, Mercedes-Benz and Ford have launched self-driving programs, while Google self-driving cars became an independent subsidiary of Alphabet in 2016 and changed their name to Waymo. The industry predicts that 2020 will be the first year of commercialization of self-driving vehicles. From the point of view at that time, the prospect of autopilot was indeed very good.

But since 2018, the self-driving industry has entered a bubble burst, capital enthusiasm has slowly faded, and investors have begun to return to rationality, no longer blindly pursuing the track with inflated numbers, and only focus on how much smart driving is worth in mass production.

As a result, the all-in-one track is getting hot in 2021. This year, more than a dozen local suppliers, such as Desai Xiwei, Fretek, Jimu Intelligence, E-Airlines Intelligence, Intelligent driving Technology, and so on, announced the launch of integrated parking solutions, and successively got the pre-installed production points.

At the end of April 2022, the first production car equipped with E-Airlines intelligent NOA parking program has been officially launched. Compared with the listed NOA scheme, the cost of this scheme is reduced by more than 50%, and the functions such as automatic overtaking, automatic road network switching and memory parking can be realized, and the NOA can be covered to less than 150000 yuan through a flexible hardware configuration scheme.

Then, the light boat Zhihang announced the cost of as low as 10,000 yuan of car-level autopilot program, and completed the pre-loading mass production. Wang Ping, chief executive of Cambrian Xingsong, also revealed that two self-driving chips will be officially released at the China Electric vehicle 100 people Forum (2022). Among them, SD5223 is oriented to the L2 + self-driving market, and its maximum computing power exceeds that of 16TOPS. A single SoC (system-on-chip) can realize the function of integrated travel and parking, supporting the coverage of the autopilot system to entry-level models of about 100000 yuan. In addition, the integrated NOA solutions of many enterprises, such as Zhixing Technology and Heduo Technology, will also achieve mass production.

Picture from: the side of Yiou net, the boarding of the integrated parking scheme is also accelerating.

Last year, the tank 500, equipped with the HPilot 2.0 driving assistance system, was officially launched, which is also the first model of the upgraded HPilot 2.0 integrated parking system. On October 31, Euler's "Super streamlined Pure Electric sedan" Lightning Cat went on sale, also equipped with the HPilot 2.0 solution, an optional package called ORA-PILOT 3.0 Intelligent driving Assistance package, which costs 20,000 yuan.

In addition, as the vanguard of the domestic auto market, BYD will also begin mass production on the integrated parking program this year. The SDA architecture of Changan Automobile's intelligent car super digital platform will also build the electronic and electrical architecture of the "central + area" ring network, and deploy the integrated parking and cabin solutions.

At least two years ago, intelligent driving domain control was exclusive to high-end models, and the accelerated landing of these intelligent driving solutions, which are said to be higher than ADAS, may mean that China's automobile market is not far away from the era of truly realizing a smart car.

3THREE, challenges and difficulties although the integrated parking market is very popular, it has been learned that most of the integrated parking products in the market are still designed and integrated based on the 1.0framework. To put it simply, the driving subsystem (driving SoC) and parking subsystem (parking SoC) are combined into one box, and the driving and parking systems are essentially independent of each other. But the real line-to-park solution should be to share a SoC and sensor hardware.

Therefore, the implementation process of the integrated travel and berthing scheme is not achieved overnight. based on the existing software and hardware level, the traveling and berthing integration scheme is divided into two kinds: lightweight integrated berthing domain control scheme and big computing integrated berthing domain control scheme. The biggest difference in design between the two is whether more computing power is pre-embedded to provide room for upgrading in the future. In terms of computing power, the computing power of lightweight schemes generally ranges from a few TOPS to dozens of TOPS, while high-power schemes can reach more than 200TOPS.

Which is better or worse? Guo Enqing, CTO of Intelligent driving Technology, believes: "the advantage of the big computing solution is that the OTA can bring more functions in the future, but the cost is higher, while the lightweight solution with basic functions can achieve large-scale mass production at this stage, but it is almost impossible to add additional functions, let alone function upgrades that produce generation disparity."

Both lightweight schemes and big computing schemes are faced with a series of challenges and difficulties.

For one thing, it lies in arithmetic. After the functional integration and fusion, each camera needs some supplement and coordination, and the algorithm will pay attention to fusion and adjustment for different scenes, so as to achieve the expansion of function and further improve the recognition accuracy, reliability and other indicators.

The second is the reduction of power consumption. How to maintain low power consumption while having high performance is a difficult problem for the integrated line-to-park scheme at present.

Finally, the strategies and algorithms adopted by autopilot software are different in different scenarios, and the integrated travel and parking scheme requires different scenarios to share a set of algorithm model. it puts forward higher requirements for the scheduling of the software and the generalization of the algorithm model, and the scheduling and algorithm model of the software should be able to meet the needs of driving and parking scenarios at the same time. How to design a set of reasonable software architecture and scheduling system is another difficult problem to realize the integrated travel and berthing scheme.

In this regard, the relevant person in charge of the Baidu intelligent driving business department believes that the ability of research and development of the whole stack is the key to break the situation. For example, in terms of functional development, the computing power of chips is generally insufficient, so it is necessary to optimize heterogeneous computing with software and hardware to ensure the experience; through self-research and development of hardware, we can better optimize computing power in software and maximize the value of software; suppliers need to do hardware adaptation in the process of cooperation with car factories. On the other hand, the self-developed hardware does not need to adapt to many different hardware domain controllers, has strong technical stability, shortens the development cycle, reduces the workload, and reduces the risk at the implementation level.

In the long run, intelligent driving operators who are unable to take root in the underlying data logic and are difficult to strengthen the competitiveness of their products according to market demand will be eliminated in the reshuffle period in the future.

4FOUR, what is the end of the future of the self-driving industry? How should the commercialization of self-driving companies go? At present, no one can explain.

However, at present, the landing time of mass production of L4 self-driving is unknown, and whether it has the function of auxiliary driving is bound to become a must for intelligent driving companies in the next stage, so enlarging the L2 auxiliary driving cake, complying with the development path of NOA parking, and constantly improving the intelligent driving ecology may be the best choice for progressive intelligent driving solutions.

[full text reference]

[1] Why is the all-in-one solution so popular? China Automobile News, Ma Xin and Zhang Yahui

[2] "the first year of 2022 Noa parking", Yiou Automobile, Zhang Yuzhe

[3] "the integration of people in different ways has finally been made clear", the Heart of the car, Tian Xin

This article comes from the official account of Wechat: che Bai think Tank (ID:EV100_Plus). Author: Zhou Shuangjiang.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.