The "strong wind" in the autonomous driving industry is coming again.
On June 21, Xin Guobin, vice minister of the Ministry of Industry and Information Technology, stated at a regular policy briefing of the State Council that he would launch a pilot program for the access and on-road traffic of intelligent networked vehicles. Commercial application of driving functions". The Ministry of Industry and Information Technology has previously revealed that the "Guidelines for the Standard System of Intelligent Connected Vehicles" will be officially released and implemented soon.
News is coming out frequently, and the autonomous driving industry chain has another wave of restlessness. But in the secondary market, things are a bit different. NavInfo, an industry cutting-edge Tier 1 with "intelligent networked vehicle data compliance + high-precision maps + localized chips + intelligent driving software and hardware" as its value anchor, has been jumping up and down in performance and its stock price has fluctuated sideways for eight years In the context of the current situation, it has been subject to controversy among investors.
As early as April, Beijing has opened the approval of high-definition maps for the whole region, accelerating the release of the value of high-precision maps for intelligent driving. But also from the Shanghai Auto Show at that time to the present, a group of car companies and autonomous driving companies have announced that they will gradually strengthen the "emphasis on perception, light on maps" route, and even realize "off the map". This has virtually put NavInfo, which uses high-precision maps as the cornerstone of its business, into a dilemma for public opinion.
However, we cannot ignore that in the statement of the Ministry of Industry and Information Technology, high-level autonomous driving not only requires vehicle intelligence, but also depends on the coordination of vehicles, roads, and clouds. It requires the support of a large number of information infrastructure to form an orderly transportation system. . In other words, the perception of bicycle intelligence does not seem to be the ultimate way out.
Recently, NavInfo's advanced driving assistance map covering 120 cities in 30 provinces has been approved by relevant departments, becoming the first company in the country to obtain the national urban open road advanced driving assistance map approval number. By 2025, NavInfo also hopes to increase this number to 150. These data will become a part of NavInfo's "four-in-one" (navigation map, assisted driving map, high-precision map, parking lot map) OneMap platform, forming a highly three-dimensional intelligent driving system, which happens to be compatible with The orientation of the top-level design has formed a certain resonance.
In contradictions, there must be misunderstandings. How to evaluate the "perception dream" of autonomous driving and NavInfo's Tier 1 road?
Differences in starting points lead to misunderstanding of high-precision maps
High-precision maps are entering a kind of "extreme" public opinion: one party believes that high-precision maps are superfluous in the current autonomous driving environment, while the other firmly believes that it is indispensable.
Some investors bluntly said that the implementation of NOA (Navigate On Autopilot, automatic assisted navigation driving or navigation assisted driving, which belongs to an application of high-level assisted driving functions) vigorously promoted by various car companies this year is essentially bad for NavInfo, because car companies In our planned NOA application, we need to gradually abandon high-definition maps, which have long been NavInfo's biggest highlight and the foundation of its business.
(Mainstream car companies' "off-map" plan. Source: Essence Securities)
But it ignores an important standpoint-for car companies, cost reduction is the top priority in the development of smart driving solutions. They can't wait to have the same frequency as Tesla, and even remove the lidar together, and only use the visual solution.
On this core issue, the original problem of high-precision maps is high cost. Yu Chengdong once explained that Huawei spent two or three years collecting high-precision maps in Shanghai, and finally covered 22,000 kilometers of roads. However, if it wants to cover the whole of Shanghai, this number will reach 36,000 kilometers. For car companies that need to enter the national market and gradually establish a sustainable business model, choosing a high-precision map is equivalent to taking the initiative to carry a heavy burden, and this does not include subsequent operations such as updates and maintenance.
Therefore, there seems to be a recognition in today's market that car companies are "going high-precision maps", and this is an unavoidable strategic adjustment for the development of intelligent driving solutions.
So, is the logic in it necessarily correct? Maybe not quite.
On the one hand, starting from the core cost issue, the high cost of high-precision maps is actually a misunderstanding. A good high-precision map solution can reduce costs.
At this stage, lidar is still the most expensive part of smart vehicle hardware, but due to the needs of perception solutions, most car companies can only reduce its quantity and configuration, but cannot completely cancel it. For example, in Huawei ADS2.0, the lidar is reduced to one, and the millimeter-wave radar and high-definition camera are also reduced to a certain extent. Compared with its 1.0 version, the cost has been reduced by nearly half.
However, no matter how fast the scale effect of hardware grows, it is also affected by factors such as raw materials and configuration costs. Therefore, hardware cost reduction has reached a bottleneck, and high-precision maps have naturally become the object of cost reduction. Especially at the current stage when car companies use L2-level assisted driving solutions, it is not necessary to adopt high-precision solutions for low-level assisted driving. From a marketing point of view, car companies will not brag about their advantages without adopting a high-precision map solution.
But when it comes to higher-level autonomous driving in the future, car companies need to consider more about safety redundancy and the overall operating efficiency of the transportation system, and the profitability of autonomous driving solutions has increased, and high-precision maps may still return to the "C position" on the stage . At that time, the wind direction will inevitably change.
Starting from the characteristics of high-precision maps, it is capable of greatly reducing the perception cost of car companies, but unlike the perception scheme, users can "train" themselves to achieve faster deployment. In fact, the comprehensive cost of perception hardware can reach tens of thousands of yuan, but the additional cost per vehicle of NavInfo's "high-precision map + perception" solution can reach one to two hundred yuan. The high-precision map solution does not need to purchase new hardware, and can meet the needs of car companies for low prices, good functions, and fast on-boarding.
Meng Qingxin, Senior Vice President, CMO and Secretary of the Board of Directors of NavInfo Group, once mentioned in an interview with Sohu Automobile: "The sensors on the market now cost hundreds of thousands. We have to consider the cost of car buyers. End users hope that both high-quality experience, and there are also preferential prices... Tens of thousands of yuan can be directly reduced to a hundred yuan, and the application is good, and the cost is also reduced, which is quite lethal." For example, the latest scene map released by NavInfo, based on the data of crossings across the country, With a cost of 20 US dollars, the ability to reduce the hardware cost of the vehicle by 20,000 yuan is achieved.
On the other hand, abandoning high-precision maps means adding additional costs, but this is also ignored in the context of low-level assisted driving, and the problem will become prominent in the future sooner or later.
From the perspective of hardware cost, the pure perception solution means increasing dependence on various information sensors and computing power, including but not limited to GPS, wheel speed sensors and so on. The more complex the hardware architecture, the higher the probability of problems, which will also increase the later maintenance costs.
From the perspective of decision-making costs, the scene in which the vehicle is located is purely based on perception, and it is necessary to maintain a state from "ignorance" to "knowledge" at all times, because before obtaining surrounding information, the vehicle has no idea what will happen in the next ten meters. In the future, after L3 level and above automatic driving enters the actual application, the responsibility will be assigned to the car companies, and the car companies will consider retaining more safety redundancy.
(Users are still controversial about autonomous driving. Source: Hupu comment area)
Finally, high-precision maps have essentially become the "base" of NavInfo's autonomous driving solutions. As a part of data, maps play a systematic role rather than a simple tool. In the past, the value of maps lay in navigation, which was for people to see, but now maps are spatial information complexes, which are for machines to see. This change hides a mystery.
Striving to become Tier 1, NavInfo is entering the "systematic" range of perception
"We provide a solution, not just a picture." Meng Qingxin's words express the essence of NavInfo as a new type of Tier 1.
In addition, Jian Guodong, senior vice president of NavInfo, also mentioned in an interview: "Autonomous driving is inseparable from a highly reliable perception system. Among the various sensors that make up the autonomous driving perception system, the autonomous driving map has high , over-the-horizon and other key features, can solve the problem of reliable positioning and planning when other sensors fail."
(High-precision maps contain multi-dimensional information. Source: Huaxi Securities)
"Therefore, core businesses such as autonomous driving maps, high-precision positioning, cloud service platforms, and automotive-grade chips for ADAS and autonomous driving may become the basic elements for creating a 'new Tier1 smart car'." He said.
Therefore, relying on a misunderstanding of the way out of high-precision maps to deconstruct the future of NavInfo's smart ecosystem from the perspective of expectations is obviously biased.
For example, in the performance briefing meeting at the end of April, almost all the focus of the interactive session of institutional questions fell on the "smart core" chip business. From the perspective of performance, in 2022, NavInfo's smart driving business revenue will grow by 2144.4% year-on-year, and it has won orders from major customers such as BYD. At this year's Shanghai Auto Show, NavInfo presented a series of full-scenario solutions covering ADAS to L2+. NavInfo is not on the branch of the supply chain, but has entered the main road.
In addition, a direction that is easily overlooked is, in a normal social environment, is the ultimate goal of autonomous driving just to liberate manpower? The answer is no. The reason why the Ministry of Industry and Information Technology is trying to promote the integration of vehicles, roads and clouds is to improve the overall efficiency of road traffic, that is, the overall efficiency of the transportation system. However, the self-awareness of bicycle intelligence can only meet the needs of vehicles and cannot meet the needs of the transportation system. Unless all car companies get through the data, this is extremely difficult from a commercial perspective.
However, starting from the high-precision map, NavInfo provides a reference idea. Meng Qingxin mentioned: "For data, more decisions are made based on the location of the car, which can greatly reduce the "short horizon" of cameras and radars, because the map is a 'full scene'. That is to say, if we put The self-driving maps of highways and key cities across the country have been collected, and the data can be used to restore a real world, laying a layer of car brain memory on the overall solution of the vehicle."
The greatest value of this actually lies in the "reproduction" of the transportation system. It successfully integrates the individual perspective of bicycle perception into the overall perspective of the transportation system. In this way, in scenarios such as complex intersections and ramps, the significance of autonomous driving for traffic management has been greatly improved.
Therefore, from the final perspective of traffic, high-precision maps are not necessarily the ultimate answer to perception, but they are valuable adjustment systems. And these values are brought into play based on the systematic solution of "cloud + terminal + data + computing power" provided by NavInfo.
epilogue
Zhiyun, smart cabin, smart core, and smart driving, the four-dimensional integrated NavInfo is the complete body and can afford the dream of Tier 1. The automotive industry is undergoing tremendous changes, which will bring about a reshuffling of the supply chain. The old Tier 1 falls, and the new Tier 1 rises again. Only by mastering the core ability can we remain invincible. In this round of ups and downs, NavInfo will move forward together with the industry. Those controversies about it will eventually disappear in this process.
Source: Songuo Finance