＂There is a picture＂ and ＂there is no picture＂ quarreling from a distance. After all, is it still the cost pressure of smart driving？

Recently, the voice of "intelligent driving without a map" swept through the automobile circle, causing heated discussions. Before that, AITO asked the world and Tucki talked about "not relying on high-precision maps", and then the ultra-01 entered the market, and the first domestic production model that realized high-order intelligent driving by relying on "pure vision" perception algorithm officially landed.

In 2023, with the wide popularization of NOA in cities, the technical route of intelligent driving "going to high-precision maps" was pushed to the forefront. Since this year, car companies have successively entered the stage of city intelligent driving, from "light maps" to "no maps", getting rid of high-precision maps has become the key for enterprises to open national NCA plans.

Kaicheng "Marathon", the city’s smart driving vents swept through 2023.

2023 is called the first year of artificial intelligence, and artificial intelligence hot words such as AI computing power, big model, BEV+Trasnsformer are constantly emerging. The layout speed of NOA in cities has also become the annual KPI of new car companies, and the progress of intelligent driving in Kaesong seems to be a measure of the status and hard power of many enterprises in the field of automatic driving, especially in the second half of 2023, the competition among enterprises is: Kaesong numbers.

On the 1024 Xpeng Motors Science and Technology Day, the official announced that XNGP city navigation-assisted driving had covered 25 cities in November and opened to 50 cities at the end of December; As early as April this year, the more radical LI publicly emphasized that the speed of "opening the city" should be raised to three digits this year: 100; Huawei originally planned to cover 45 cities by NCA in the fourth quarter of 2023. Yu Chengdong, managing director of Huawei, CEO of BG and CEO of BU, a smart car solution, sharply increased this goal to "all cities can open" in December at the press conference of the new M7.

The competition in the number of Kaesong can be described as intensifying.

The new forces believe that the large-scale opening of the city NOA will help them stay in the competition seat in the development of high-level intelligent driving and become the key to the next stage of competition. Li Liyun, the person in charge of automatic driving in Xpeng Motors, said recently that the XNGP urban navigation-assisted driving, which can be opened all over the country, will help Xpeng Motors "enter the second half first", and the XNGP urban intelligent driving with light map scheme will not be limited by the use range and updating time of high-precision maps. Li Liyun even shouted: "The battle without maps is over".

On October 27th, Extreme Yue 01 was officially listed. Extreme Yue was the "strategic cooperation project of automobile robots" initiated by Geely Holding Group and Baidu in 2021. Geely was mainly responsible for controlling the manufacturing of the whole vehicle and building the platform architecture, while Baidu gave Extreme Yue 01 the accumulation of AI algorithm and intelligent driving. At present, Extreme Yue 01 is the first mass-produced vehicle in China that relies on "pure vision" perception algorithm to realize high-order intelligent assisted driving. Extreme Yue 01′ s advanced intelligent driving ROBO Drive Max adopts BEV+Transformer pure vision scheme, and its technical route is very close to Tesla’s. The first Occupancy Network technology ("OCC") jointly developed by Extreme Yue and Baidu is also announced for the first time, and the role of OCC is to further enhance the "pure vision" of Extreme Yue 01.

It is worth noting that the production model of Extreme Yue 01 is not equipped with lidar. Xia Yiping, CEO of Jiji and Extreme Yue, revealed that the scheme of going to lidar was only determined during the discussion at the beginning of this year, but he also stressed that going to lidar is not necessarily the key point. After the laws and regulations of L4 level are released in the future, lidar will still be needed.

According to the official plan of Extreme Vietnam, in 2023, ROBO Drive Max will open the high-level intelligent driving function in urban areas to all users in Shanghai, Shenzhen and Hangzhou. At the same time, PPA can be used in 90% of the country’s high-speed viaducts. In 2024, PPA will gradually open 200+ cities nationwide.

The voices of "going to the lidar" and "going to the high-precision map" are often mentioned recently. Among them, the advantage of "no map" is to reduce the lag of map update and reduce the cost of intelligent driving, which is also conducive to the promotion of high-level intelligent driving in different cities.

However, there are many objections to "no graph theory" in the industry.

Map manufacturers bombard, is "no map" really possible?

In the process of promoting urban NOA, due to over-reliance on high-precision maps, all car companies have been unable to expand rapidly. In August 2022, the Ministry of Natural Resources announced that six cities, including Beijing, Shanghai, Guangzhou, Shenzhen, Hangzhou and Chongqing, would carry out the pilot application of intelligent networked automobile high-precision maps. However, after more than a year, the pilot cities have not continued to develop. In other words, relying on high-precision maps, the speed of car companies opening NOA in cities across the country will become faster.

Therefore, in this era when everything is "rolled", in order to accelerate the promotion of the city NOA, car companies decisively chose to abandon the once high-precision map of Xiangbo. Xpeng Motors CEO He Xiaopeng has publicly stated that high-precision maps must be a transition. As early as 2019, Musk on the other side of the ocean angrily sprayed high-precision maps, calling them "really a bad idea". Tesla insisted on choosing visual perception technology, and launched the "BEV+Transformer" model in 2021.

It is undeniable that high-precision maps play an important role in the automatic driving perception layer, and more accurate environmental perception can be achieved through electronic maps with higher precision and wider data dimensions. Moreover, in a higher-level automatic driving system, the cooperation between high-precision maps and traffic elements and road connections is also very important. In the final analysis, car companies chose to "abandon" high-precision maps in the second half of 2023, or because of the low frequency of updating high-precision maps, high maintenance costs, complicated map review and qualifications, which did not meet the plan of car companies to "open the city" in a hurry.

Therefore, when talking about the general direction of "de-cartography", Cheng Peng, the new CEO of NavInfo, a veteran cartographer, angrily sprayed some automobile manufacturers with no map qualification, no intellectual property rights and no fear of safety. In addition, Cheng Peng said that NavInfo found that some automobile companies that claimed not to rely on high-precision maps still had map information after covering their sensors, which indicated that they still adopted the map scheme.

Of course, high-precision maps also have firm fans, including Waymo, a self-driving company owned by Google, and the development of intelligent driving systems of traditional car companies such as GM and BMW also rely heavily on high-precision maps.

We always emphasize the differences between traditional car companies and new power companies in the concept of building cars, and we can also see the cognitive differences of enterprises from their attitudes towards high-precision maps. From the current stage, urban NOA does not need the help of high-precision maps or even laser radar in most scenes, but when the author test-drives the PPA point-to-point navigation assistance of Extreme Yue 01, there is indeed a deviation in traffic light signal recognition. In addition, the environmental perception and recognition ability of pure visual perception scheme in rainy, snowy and dusty days is not good.

There is a set of interesting price data. In September this year, Tesla drastically lowered the price of its fully automatic driving (FSD) package in the United States, from the previous $15,000 to $12,000, a decrease of 20%. Before this year, Tesla officials emphasized more than once that with the continuous iterative upgrade of FSD technology, its price will only rise. Judging from the price of high-order intelligent driving schemes for three new domestic models in the last six months, Zhiji LS6 IM AD intelligent driving assistance system is provided to users as a standard in the whole department and free of charge for life. The price of Tucki G6 MAX version (XNGP) is only 20,000 yuan higher than that of Pro version (XPILOT), and the rights and interests given to users by Extreme Leap 01 include the preferential price of 6-month subscription to ROBO Drive Max and limited time ROBO Drive Max buyout of 19,900 yuan. Judging from the price trend, ordering high-order assisted driving systems is more cost-effective, indicating that car companies pay more attention to cost-effectiveness in the field of intelligent driving.

Embracing "emphasizing perception, ignoring map" or even "no map" is becoming the mainstream practice of domestic new power enterprises. Facing the cruelty of domestic new energy vehicle market competition, new power enterprises pay more attention to reducing the cost of intelligent driving, which is also caused by the big environment, but the high-precision map "backs the pot".