天天爽天天看-日本少妇xxxx动漫-亚洲av成人精品毛片-欧美一区二区三区激情啪啪-91精品国产成-欧美牲交a欧美牲交aⅴ-欧美日韩国产免费观看-九九99精品久久久久久综合-欧美理论影院-国内老熟妇乱子伦视频-www在线观看av-黄色裸体网站-少妇高潮惨叫久久久久久-国产偷国产偷av亚洲清高-国产午夜精品理论片小yo奈-高大丰满熟妇丰满的大白屁股

Source: China Automotive News  22^nd Aug 2025

"By 2027, the choice between the two routes may no longer be an issue!" Recently, He Xiaopeng, the chairman of Xpeng Motors, made such a statement, once again stirring up the debate between the pure vision route and the LiDAR route. The dispute over the two major environmental perception technology routes has a long history but has never been settled. Nowadays, technology, cost, and the popularity of assisted driving are all different from before. Can the dispute over the intelligent driving route really be settled as He Xiaopeng said by 2027?

Both routes have their supporters

If previously the pure vision approach was merely the "persistence of a minority", lidar was the "choice of the majority". Now, both technical routes have gained many supporters and are evenly matched. There are new entrants choosing to join the lidar camp, and there are also carmakers choosing to "defect" to the pure vision approach.

At present, many automakers are still laying out the LiDAR route. The newly launched Tank 500 is equipped with LiDAR and is fitted with the third-generation Coffee Pilot Ultra autonomous driving system of Great Wall Motor. The Ideal i8 is equipped with 1 LiDAR, 12 cameras, 12 ultrasonic radars and 1 millimeter-wave radar, totaling 26 intelligent sensors. Xiaomi Auto is also firmly committed to the LiDAR route, with the Xiaomi YU7 series all equipped with LiDAR and 4D millimeter-wave radar. Even the traditional multinational automaker Audi, in collaboration with Huawei, has equipped its strategic pure electric product Q6L e-tron with a perception system consisting of "2 LiDARs + 5 millimeter-wave radars + 13 cameras + 12 ultrasonic radars".

Moreover, lidar is no longer exclusive to mid-to-high-end models. The Zero Run B01, which was launched in July, offers two versions equipped with lidar: the 550 Lidar Edition and the 650 Lidar Edition, with starting prices of 113,800 yuan and 119,800 yuan respectively. Previously, Changan Automobile also disclosed that it plans to install lidar on models priced at around 100,000 yuan in August. Data released by the GGII Institute for Intelligent Automotive Industry shows that in the first half of this year, the pre-installation and delivery volume of lidar in China's passenger car market reached 1.0439 million units, representing a year-on-year increase of 83.14%.

However, compared with the past, on the pure vision route, Tesla is no longer the only "brave warrior". Instead, more and more automakers and ADAS suppliers have emerged. In March last year, DJI's subsidiary, Zenutro, released the "Chengxing Platform Upgrade Version" and "Chengxing Platform High-End Version" for urban high-level intelligent driving. Among them, the Chengxing Platform Upgrade Version is equipped with 7V + 100TOPS, relying solely on 7 cameras and a Qualcomm chip with 100TOPS computing power to achieve no-map, pure vision urban intelligent driving functions. Nowadays, automakers such as Volkswagen, SAIC-GM-Wuling, and Chery Automobile have all reached cooperation with Zenutro in the field of ADAS.

In addition, Huawei, a representative of the lidar route, also launched the Huawei Kunpeng ADS SE Basic Edition last year. This version does not come with lidar and adopts a vision-based fusion perception solution, capable of achieving urban lane cruise assistance LCC+ and high-speed intelligent driving assistance NCA functions. It also provides parking assistance and all-round active safety features. It is reported that this version is mainly targeted at models in the 150,000 yuan price range, aiming to expand Huawei Kunpeng ADS's market share in mid-to-low-end models through a low-cost pure vision route. The 2026 Deep Blue L07, which was officially launched on August 13th, comes standard with the Huawei Kunpeng ADS SE Basic Edition across the board, with prices ranging from 135,900 to 155,900 yuan.

NIO has also adopted a pure vision solution on its sub-brand Land Rover. The Land Rover L90, which was launched in July, continues the pure vision solution of the L60 and is equipped with 7 8-megapixel cameras, 4 3-megapixel surround-view cameras, 1 4D imaging millimeter-wave radar, and 12 ultrasonic radars, enabling high-speed and urban assisted driving functions.

Xiaopeng has resolutely abandoned lidar and chosen to fully shift to a pure vision solution. Not only does the MONA M03, priced under 150,000 yuan, adopt a pure vision solution, but the newly updated all-new Xiaopeng P7 also adheres to this approach. Not long ago, several senior executives of Xiaopeng Auto firmly stated: "We will definitely not use lidar and will stick to the pure vision route."

Cost and safety remain the core points of contention

Previously, the industry's debate over the two technical routes of pure vision and lidar mainly centered on safety and cost. On the one hand, the camera sensors in pure vision solutions are relatively mature and inexpensive, making large-scale production feasible. On the other hand, due to the technical limitations of cameras, they cannot truly change focal length and depth of field like human eyes, and the recognition accuracy of pure vision solutions is limited in complex road conditions and weather environments. Although the multi-sensor fusion solution of "camera + radar" performs better than pure vision solutions in complex weather conditions, its high cost makes it difficult to popularize. As time has passed and technology has continued to mature and hardware costs have decreased, what has changed in the debate between the two?

According to Lu Wenliang, a specially-appointed researcher in the automotive industry at the Institute of Industrial Innovation of the Chinese Academy of Sciences' Strategic Consulting Institute, the main points of contention between the two technical routes remain cost and safety. Automakers that adopt pure vision solutions often focus on the cost-effectiveness of such solutions, which makes them easier to promote on a large scale. At the same time, with the support of advanced technologies such as BEV+Transformer and Occupancy Network Computing (OCC), the accuracy of pure vision technology has also been greatly enhanced.

Automakers that adopt lidar solutions prioritize safety in extreme environments, believing that lidar has stronger perception capabilities, is applicable to more scenarios, and is safer. It is worth noting that the cost of lidar has also dropped significantly. It is understood that the price of the main lidar used by leading lidar manufacturer Hesai Technology for L2-level intelligent assisted driving has dropped to around $200 (about 1,434 yuan), a decrease of more than 99.5% compared to 200,000 to 300,000 yuan three or four years ago.

Although both have improved, the supporters of the two technical routes still "dislike each other". Yuan Tingting, senior director of autonomous driving products at XPeng Motors, pointed out from three dimensions of physical characteristics, environmental adaptability and information processing efficiency that the claim that lidar can see far is actually a "false proposition". For example, in the identification of extreme weather, Yuan Tingting said that the actual test data shows that in heavy rain, the effective detection distance of lidar drops sharply to within 30 meters, and the near-field noise increases fivefold.

In addition, He Xiaopeng believes that the upper limit of pure vision solutions far exceeds that of lidar. "In the future, vision can identify nails on the road that may puncture tires and manhole covers that have been moved, which is very difficult for lidar to achieve," he said. In response, some industry insiders pointed out that lidar currently mainly addresses obstacle detection in wide fields of view, but through technical optimization, lidar can also meet the high-precision recognition requirements in specific scenarios.

Some industry insiders have pointed out that pure visual solutions still have limitations when cameras "cannot see" or "cannot see clearly", such as in cases of strong light exposure, weak light at night, or when foreground objects and background colors are the same, making it impossible to distinguish them. In such situations, there may be problems of untimely or incomplete obstacle recognition.

Meanwhile, Hu Jianyao, a chief researcher at the Fifth Research Institute of the Ministry of Industry and Information Technology, pointed out that while pure vision solutions are simpler and less costly in terms of hardware, they also have certain limitations. only through the iteration of computing power and the establishment of a data loop can these solutions largely make up for the hardware deficiencies. "The computing power demands of car manufacturers increase every year, and they constantly need to train on massive amounts of road condition data. This means that car manufacturers may have to spend a considerable amount to build supercomputing centers, and this indirect cost could be even higher," Hu Jianyao said.

Move forward in integration and develop in parallel

In He Xiaopeng's view, the poor performance of past visual solutions was due to insufficient computing power, lacking both adequate pixel arrays and frame rates, as well as spatio-temporal logic. With the rapid increase in computing power, the potential of pure visual solutions will be greatly unleashed, and this technical route will become the top choice for car manufacturers.

Lu Wenliang also believes that the current main constraint for pure vision solutions is computing power. "Camera technology is already quite mature. Besides cameras, pure vision solutions only have computing power and algorithms. Computing power and algorithms are interdependent. An increase in computing power can make up for the shortcomings of algorithms. Efficient algorithms can also reduce the demand for computing power," Lu Wenliang said. He Xiaopeng stated that the computing power for automotive assisted driving has increased tenfold compared to five years ago, and it will increase tenfold or even more in the next five to ten years. Nowadays, apart from Tesla, many automakers such as XPeng, Great Wall, Geely, and Li Auto, as well as suppliers like Huawei and SenseTime, have launched a computing power arms race, building their own supercomputing centers. This undoubtedly brings great benefits to the development of pure vision solutions.

However, when it comes to higher levels of autonomous driving, many still believe that lidar is a "must-have". Gu Jianmin, the Chief Technology Officer of Valeo China, stated that for L3 and above autonomous driving, lidar is indispensable, playing a role in perception fusion and safety redundancy. Jin Yuzhi, the CEO of Huawei's Intelligent Automotive Solutions BU, also believes that for the industry to move towards L3 and L4 autonomous driving, lidar is necessary.

For lidar solutions, the cost needs to be further reduced. Lu Wenliang said that the cost of the lidar route not only covers its own hardware cost, but also needs to be considered from the system-level cost. "Some models will be equipped with multiple lidars, and the cost of some high-performance lidars is still very high. In addition, after having lidars, high-performance processing chips, buses, domain controllers, etc. also need to be equipped," Lu Wenliang said. In this regard, Hu Jianyao believes that the cost of lidar will continue to decline with technological iteration and scale effect. In his view, the cost of lidar will drop below $80 in the future.

Regarding whether pure vision solutions can outperform lidar by 2027, Lu Wenliang believes that lidar's environmental perception capabilities in specific scenarios remain unmatched by cameras, so it is highly likely that both technical routes will develop in parallel. Hu Jianyao also stated that if costs converge in the future, the two technical routes can consider integrating and developing from different scenarios. "In urban commuting scenarios, pure vision solutions are already sufficient. Lidar can be used in complex road conditions to play a role in safety redundancy," Hu Jianyao said.