Is high-end intelligent driving a true embodiment of technological progress, or is it a word game when car companies advertise? What is the difference between 300,000 yuan and 100,000 yuan?
With the continuous advancement of technology, smart cars have gradually become a major trend in the automotive industry, and the closely related autonomous driving technology has also received widespread attention from the industry. With the increasing number of entrants, many smart car brands and models have emerged in the market. For example, NIO, Ideal, Zhiji, Zeekr, Xpeng, Chery iCAR, Baojun, etc., they are equipped with different intelligent driving systems, and there are significant differences in the prices of models. This triggered the reporter of "Automobile Vertical" to think deeply: what are the substantial differences between the models in the price range of 300,000 yuan and above and the models in the price range of 100,000 to 200,000 yuan?
What is Advanced Intelligent Driving?
First, let's dive into the concept of "advanced intelligent driving". As the name suggests, high-end intelligent driving refers to a car with a higher level of intelligent driving functions. According to the autonomous driving classification standard formulated by the Society of Automotive Engineers (SAE) (see Figure 1), high-level intelligent driving should be above Level 2, that is, conditional autonomous driving. In other words, compared with traditional Level 2 driver assistance systems, high-level intelligent driving should be able to handle more complex driving scenarios and tasks.
For example, conditional autonomous driving on highways, including automatic overtaking, automatic lane changing, automatic distance keeping and other functions. At the same time, the high-end intelligent driving system should also have higher environmental awareness and decision-making capabilities, and be able to more accurately identify road signs, traffic signals, obstacles and other information, and make more accurate driving decisions. However, although the international community has established clear hierarchical standards for intelligent driving, there are also differences between autonomous driving functions due to the uneven level of autonomous driving technology research and development of different manufacturers. Take the automatic parking function as an example, although most models are equipped with this function, the effect of automatic parking varies greatly from model to model. The automatic parking function of some models can only complete simple parallel parking, while some high-end models can realize a variety of complex functions such as vertical parking and automatic search for parking spaces. In addition, even if some models are equipped with high-level autonomous driving functions, such as high-speed NOA (Pilot Driver Assistance), the actual performance of these functions will be affected due to differences in the hardware configuration of the models. For example, some models may be equipped with high-performance sensors or more powerful algorithms, which make autonomous driving functions more stable and reliable. Others may have occasional delays or even misjudgments due to hardware configuration or algorithm defects. This discrepancy is not only confusing for consumers, but also blurs the definition of "high-end". However, most industry insiders hold a consensus: the definition of high-end intelligent driving should be dynamically adjusted. For example, two years ago, as long as it had a high-speed NOA function, it was considered a high-level intelligent driving; At present, high-speed NOA is no longer scarce, and urban NOA can be called "high-level", such as the identification of construction areas, anti-congestion, traffic light recognition, etc., so that more content will be taken out in the publicity to reflect the "sense of luxury" of intelligent driving. In addition, over the past few years, the industry has generally considered the most basic ADAS technology to be the initial stage of autonomous driving, so most automakers have chosen to leave this part of the technology to their suppliers. However, since the end of last year, companies such as Huawei and Xpeng have begun to pay attention to active safety features such as AEB, emphasizing that their AEB systems can achieve braking at higher speeds. This has affected consumer perception to a certain extent, so that the optimization of AEB technology by manufacturers is also regarded as one of the signs of "high level". However, there is also an argument that the definition of high-level, intermediate-level, and low-level should be from the perspective of the consumer – that is, the closer the operating style of an autonomous driving system is to the consumer's driving habits, the more it should be considered high-level. Of course, there is a more interesting statement: from the perspective of some consumers, only those with lidar are considered "high-end".
The higher the price, the higher the level of intelligent driving?
But is that really the case? According to the reporter of "Auto Vertical", there are two main routes among the manufacturers who claim to have the function of "high-end intelligent driving". One is a solution that relies on LiDAR as the core sensor to carry out vehicle positioning and perception and analysis of the surrounding environment; The other is a scheme that mainly relies on on-board cameras and computer vision algorithms to achieve the corresponding functions. First of all, from the hardware aspect, the intelligent driving system equipped with high-end models is usually equipped with more advanced sensors, processors and algorithms to achieve higher precision perception and decision-making. The cost of these hardware devices will naturally be higher, so the price of intelligent driving systems for high-end models will also increase accordingly. Secondly, from the perspective of software, the intelligent driving system of high-end models usually has richer functions and better performance. For example, pilot driving assistance for urban roads, multi-dimensional parking lots, and other scenarios. In addition, the intelligent driving system of high-end models may also have higher safety and stability, and can better cope with various complex road and traffic conditions. However, this does not mean that the intelligent driving system equipped with the lower-priced models does not have any advantages. With the advancement of technology and the reduction of costs, more and more low-end models have begun to be equipped with intelligent driving systems. While these systems may not be as functional and performant as high-end models in some respects, they can still improve the safety and convenience of driving to a certain extent. At the same time, for many consumers, the intelligent driving system equipped with lower-priced models is enough to meet daily needs. For example, most of the practical functions such as urban memory navigation, tracking parking, and high-speed NOA (including automatic lane change) can be realized by pure vision. On April 22, on the eve of the Beijing Auto Show, Baidu Apollo released the pure visual city pilot assisted driving product ANP3 Pro, claiming to pull the hardware cost of high-end intelligent driving into the era of 10,000 yuan. Coincidentally, at the recent press conference of car companies, more than one corporate executive publicly said: "We must bring down the price of high-end intelligent driving."
This may be a signal that today's "high-end intelligent driving" is moving firmly towards the "100,000 yuan era".
High-end intelligent driving ≠ 100% safe
"Models above 300,000 yuan may be equipped with more expensive sensors, such as liDAR, millimeter-wave radar, etc., which are annotated in 3D and then projected onto images; Models of 100,000-200,000 yuan often rely on cameras to obtain road information, but because the camera cannot obtain depth information (which can be simply understood as the information obtained by the camera is a 2D plane map), the algorithm is more tested. "A doctoral student at Beijing Jiaotong University who studies the direction of autonomous driving told the reporter that the gap between 300,000 intelligent driving and 100,000 intelligent driving is mainly reflected in stability, and this stability does not mean that the higher the price, the higher the stability, but both are unstable." There are inaccuracies in 3D annotation, and cameras are also susceptible to external environmental influences. The Ph.D. student added that algorithms are the soul of autonomous driving systems, and there are only a handful of companies that can do a good job of algorithms, and no one can guarantee that the 3D annotation of lidar will be accurate, and no one can guarantee that the camera's recognition is 100% correct. And this leads to a problem: even the so-called "high-level intelligent driving" cannot guarantee absolute safety. According to the established standards of the country and the industry, functions that cannot achieve autonomous driving are generally referred to as "assisted driving functions". Since it belongs to the category of "assisted driving function", human takeover is indispensable. This is also why no matter how powerful the high-end intelligent driving system launched by car companies at this stage, manufacturers still require drivers to be ready to take over the vehicle at all times. So, how to judge a company's autonomous driving capabilities to reach the "high level"? Once an industry insider gave his opinion: to see if car companies dare to take responsibility. As long as you don't dare to take responsibility, even if you claim that your "high-end intelligent driving" level has reached L2.9999 or L8 in the publicity, it is actually still L2. But despite this, there are still many people who ignore this and blindly believe in "high-level intelligent driving", which eventually led to the tragedy. Although high-end intelligent driving does not mean absolute safety, we have reason to believe that the development prospects of high-end intelligent driving are still broad. In recent years, with the rapid development of artificial intelligence, sensor technology, big data and other fields, the technical conditions required for high-end intelligent driving have gradually matured. In fact, as a technical outlet in the field of smart cars, high-end intelligent driving is now developing rapidly at a speed that exceeds most of our cognition.
For example, with high-precision maps and real-time traffic information, intelligent navigation systems can provide drivers with the best driving route to avoid congestion and dangerous areas. Autonomous driving technology, on the other hand, can perceive and judge the surrounding environment through sensors such as radar and cameras, further ensuring driving safety. It can not only improve the safety and convenience of driving, but also promote the transformation and upgrading of the automobile industry and inject new impetus into economic development. But at the same time, we look forward to the formation of more clear and unified high-end intelligent driving standards in the industry in the future, bringing consumers a safer, more convenient and more comfortable driving experience. Here is a quote from Wang Yunpeng, vice president of Baidu Group and vice president of the intelligent driving business group: "A good intelligent driving should be a reliable driver, who can drive on any road and park in any parking space, and the cost should be low enough." Not only should we let the car of more than 200,000 yuan be used, but we should also make the car of 150,000 yuan standard. "We expect "intelligent driving" to be more powerful, and we also expect "high-end intelligent driving" to become the standard configuration of all smart cars. But until then, we expect everyone to be able to face up to "high-end intelligent driving". Note: This article was first published in the June 2024 issue of Auto Magazine in the "Industry Report" column, so stay tuned.
Image: From the Internet
Article: Car Vertical
Typesetting: Car vertical