Car companies keep an eye on urban NOA, lidar will usher in a bright moment

Editor's note: Urban NOA, which stands for Urban Navigation Assisted Driving, means that after the user sets a destination in the navigation system, the vehicle can follow, overtake, change lanes, turn, and recognize and respond to traffic lights and other functions while driving.

The development of urban NOA is very important to increase the utilization rate of assisted driving among users. Data shows that urban scenes account for 90% of the driving time. At the same time, only 25% of users travel via expressways every day, compared with 100% on urban roads.

From the perspective of improving user frequency and user experience, urban NOA competition among car companies will also intensify.

In this imminent battle for Kaesong that will test technological reserves, AutoHeart will focus on the main line of "Urban NOA" and interview car companies, intelligent driving industry chains, technical experts, and in-depth users through text and video. Rationally and objectively deconstruct the key players and key technologies of urban NOA, and provide a clear "knowledge map" for the popularization of urban NOA.

Kaicheng, or Kaicheng.

Chinese car companies are keeping a close eye on Kaicheng, as if the urban NOA can capture many cities, and the winning rate will be a few points closer to winning the intelligent throne.

Xiaopeng announced that it will expand urban NGP to 50 cities by the end of 2023 and to 200 cities in 2024. Huawei, Ideal, and NIO have also launched actions. Huawei announced in a high-profile manner that the imageless version of ADS2.0 will be launched in 45 cities, allowing urban pilot assisted driving to be "available nationwide"; Ideal plans to expand the city NOA to 100 cities.

Li Xiang, CEO of Li Auto, believes: "In the future, in the mid- to high-end car market, if urban NOA cannot be provided, consumers will be the difference between buying and not buying."

At the same time, when technical keywords such as BEV+Transformer and high-precision maps emerged, Chinese car companies also became more confident because of the new technological wave.

Car companies are full of confidence, but urban NOA is a difficult battle to win.

The driving experience of the vehicle is smooth on the wide roads in the main city. Once it travels to the narrow and crowded roads in the city, it is often caught off guard by pedestrians and electric bicycles popping up from the side.

If they want to conquer urban NOA, car companies still need to face many challenges. Lidar is likely to be the most important piece of the puzzle for urban NOA perception capabilities.

In the past year, high-precision maps have become a trend, but lidar has not shown any weakness and has become a must-have in urban NOA.

We can also get a glimpse of the investment boom from lidar.

Although investment in the smart driving industry chain is currently stagnant, lidar company Discovery Technology is like a breath of fresh air and continues to be favored by capital.

After Tanwei Technology received over 100 million yuan in Series A+ financing in July, Tanwei Technology recently completed a new round of strategic financing led by Xiaomi Group, with participation from Jimu Capital, Suzhou InnoLight and other industrial capitals.

Why does capital favor lidar?

Jiang Wen, a partner in the strategic investment department of Xiaomi Group, believes that intelligent driving is being steadily commercialized, and lidar has room to show its capabilities, providing a strong perception foundation for autonomous driving to "land safely."

It seems that the answer to urban NOA has to be found in lidar.

01 Why has lidar become a necessity for urban NOA?

The technical characteristics of lidar are the last line of security defense for urban NOA.

In urban scenarios, no domestic car company has risked safety and radically removed lidar.

Lidar has become a necessity for car companies.

If you only focus on the new models in the two years of 2022 and 2023, you will find that the new models of these Kaicheng car companies are equipped with at least one lidar regardless of price.

Take the three car brands that are the fastest and most active in developing urban NOA as examples:

• 
  

• Xpeng G6/G9 is equipped with two lidars;

• 
  

• Avita 11 is equipped with three lidars;

• 
  

• Li LiDian L7/8/9 MAX is equipped with a lidar;

In the urban NOA scenario, why do car companies need lidar more?

Dr. Mei Jiawei, VP of Tanwei Technology Products, gave an explanation:

On the one hand, urban scenes are more complex. High-speed NOA only needs to pay attention to roads and traffic vehicles due to structured closed roads. However, urban roads must not only pay attention to vehicles and pedestrians, but also deal with busy traffic conditions: blind spots, pedestrians, and no markings. The roads, and even the sudden appearance of the delivery boy, have given the urban NOA a very high level of perception.

On the other hand, it lies in the unusual technical characteristics of lidar. Currently, pure vision or vision + millimeter wave radar sensing solutions are far behind lidar in terms of detection confidence, distance accuracy, and latency.

Measuring perceptual capabilities include key indicators such as semantic classification, dynamic target detection, contour detection, target detection, and distance measurement.

With the development of technology, all three sensors have the above capabilities, but lidar still has overwhelming advantages in key indicators such as target detection, distance measurement, and contour detection.

For example, millimeter wave radar has advantages in detection range, but has obvious disadvantages in detection accuracy and resolution. Compared with lidar, millimeter wave radar cannot clearly distinguish the specific characteristics of obstacles ahead.

For example, millimeter-wave radar can sense the existence of a roadblock and its general situation ahead, but it cannot further detect whether the obstacle is a road shoulder or a slope. Lidar will sense the precise shape down to the specific shape. After sensing it as a slope, vehicles can pass safely and avoid unnecessary obstacles. stop.

From the perspective of ranging, visual cameras can also use binocular or trinocular solutions to measure distance.

But the shortcomings are also very obvious.

For example, binocular cameras can calculate parallax and depth, but stereo matching in binocular vision systems is still a classic problem in computer vision: the width and length of the baseline can measure the distance of the target with different accuracy. Often the distance measurement for targets with a wide baseline is accurate, and the distance measurement for a short baseline target is accurate. There is no balance between the two. At the same time, the binocular system also has the problem of system drift after calibration.

The unsolved technical difficulties of the binocular vision system are also one of the manifestations of the technical insurmountability of the current visual perception solution.

If a purely visual perception solution wants to achieve a perception effect similar to the human eye's "1 billion pixel camera", there are many difficulties in terms of the number of image acquisition pixels, computer processing speed, and algorithms.

The uninterpretability of AI, limitations of a priori goals, and high requirements for data volume and scene richness all make it impossible to ensure high reliability of vision in urban NOA scenarios, and commercial implementation solutions are still in the minority.

The main advantages of lidar for urban NOA scenarios lie in three aspects: high detection confidence, high range accuracy, and low latency.

High detection confidence: Lidar can achieve extremely high angle and distance resolution. High resolution can create a 3D point cloud image with clear targets, allowing it to identify more complex objects in urban environments, thereby improving irregularity. Perceptual confidence of objects and corner cases.

High distance accuracy: Lidar can achieve high-precision ranging capabilities. Taking the Tempo series lidar of Xinwei Technology as an example, the longest detection distance can reach 300 meters, and the detection accuracy reaches ±3cm level.

Low latency: Lidar directly outputs distance information, allowing the perception system to more quickly identify and judge the target's motion status and intention, thereby making path planning in advance.

In the final analysis, the core test of urban NOA lies in the user driving experience. Through lidar, car companies can achieve the lowest possible takeover rate and withdrawal rate, thereby accelerating users' understanding of the safety and smoothness of autonomous driving, and ultimately forming intelligent The driving experience is cyclical.

Is lidar the only solution to urban NOA?

There are now companies exploring 4D imaging radar, which adds a vertical angle dimension to traditional millimeter-wave radar detection distance, speed and angle dimensions. However, 4D imaging radar still cannot replace lidar.

It needs to be clear that 4D imaging radar and lidar are not interchangeable.

First, its ranging accuracy and resolution are insufficient.

Secondly, lidar can increase the point cloud by raising the high beam, making it more accurate, which is difficult to achieve with 4D imaging radar.

Clearing the fog, lidar must be a must-have for urban NOA perception capabilities.

02 Product power leverages the “impossible triangle”

On the one hand, it is the urgent need for lidar by urban NOA, and on the other hand, it is the practical exploration of the lidar industry to crack the "impossible triangle".

Lidar has an "impossible triangle" problem, that is, there is a trade-off relationship between perception performance, vehicle stability and cost control, and the conditions cannot be met at the same time.

As a high-precision sensing device, lidar has advantages in performance, resolution and detection range. However, cost control is a challenge in maintaining vehicle-grade environmental stability and mass production delivery quality.